ChatGPT Pulse | Hacker News Reader

ChatGPT Pulse

openai.com

627 pointsmeetpateltech730 comments

LLMs are increasingly part of intimate conversations. That proximity lets them learn how to manipulate minds.

We must stop treating humans as uniquely mysterious. An unfettered market for attention and persuasion will encourage people to willingly harm their own mental lives. Think social medias are bad now? Children exposed to personalized LLMs will grow up inside many tiny, tailored realities.

In a decade we may meet people who seem to inhabit alternate universes because they’ve shared so little with others. They are only tethered to reality when it is practical for them (to get on busses, the distance to a place, etc). Everything else? I have no idea how to have a conversation with someone else anymore. They can ask LLMs to generate a convincing argument for them all day, and the LLMs would be fine tuned for that.

If users routinely start conversations with LLMs, the negative feedback loop of personalization and isolation will be complete.

LLMs in intimate use risk creating isolated, personalized realities where shared conversation and common ground collapse.

> Children exposed to personalized LLMs will grow up inside many tiny, tailored realities.

It's like the verbal equivalent of The Veldt by Ray Bradbury.[0]

[0] https://www.libraryofshortstories.com/onlinereader/the-veldt

The moral of this story is that if you install a really good TV the animals will come out of it and eat you? Is the author a dog?

I suggest taking a literature course and learning how to interpret narratives.

The Veldt is a classic short story written in 1950 by Ray Bradbury, a famous and celebrated author, who also wrote the famous dystopian novel Fahrenheit 451.

Ironically, Bradbury likes to tell people that Fahrenheit 451 isn't about the thing it was obviously supposed to be about (censorship) because he now wants it to have been a metaphor for cancel culture.

Cancel culture is censorship?

One of those involves fulltime professionals backed by state violence and the other is when people on social media are mad at you.

Isn't the only difference whether the censors are in or out of government power?

Few now respect the wisdom of 'should not' even when 'can'

That difference is so big it makes it an entirely different thing.

> when people on social media are mad at you

It's about more than that - many people have lost their jobs, been de-banked, or even been arrested (especially in countries like the UK and Germany) for expressing their opinion publicly when that opinion was merely (a) what most people in their country believed in the recent past (< 50 years ago), and (b) a politically incorrect opinion.

he's been dead for a decade so I doubt he now wants the meaning to be anything. besides that he also never said anything about cancel culture he said it's about how tv turns you into a moron.

https://www.openculture.com/2017/08/ray-bradbury-reveals-the...

> In a 1994 interview, Bradbury stated that Fahrenheit 451 was more relevant during this time than in any other, stating that, "it works even better because we have political correctness now. Political correctness is the real enemy these days. The black groups want to control our thinking and you can't say certain things. The homosexual groups don't want you to criticize them. It's thought control and freedom of speech control."

They had cancel culture in the 90s too.

> you can’t say certain things

So he sees it as another form of censorship

Cancel culture is vigilante political correctness.

Next comes legalized, then deputized, then militarized...

Given that only about 10-40% of advanced readers (depending on subpopulation criteria and task [0]) can parse analogy and metaphor, parent is the majority rather than the minority.

Modern day statistics on what used to be basic reading comprehension are bleak.

[0] https://kittenbeloved.substack.com/p/college-english-majors-...

With the way LLMs are affecting paranoid people by agreeing with their paranoia it feels like we've created schizophrenia as a service.

It doesn't have to be that way of course. You could envision an LLM whose "paperclip" is coaching you to become a great "xyz". Record every minute of your day, including your conversations. Feed it to the LLM. It gives feedback on what you did wrong, refuses to be your social outlet, and demands you demonstrate learning in the next day before it rewards with more attention.

Basically, a fanatically devoted life coach that doesn't want to be your friend.

The challenge is the incentives, the market, whether such an LLM could evolve and garner reward for serving a market need.

Have you tried building this with prepromts? That would be interesting!

If that were truly the LLM's "paperclip", then how far would it be willing to go? Would it engage in cyber-crime to surreptitiously smooth your path? Would it steal? Would it be willing to hurt other people?

What if you no longer want to be a great "xyz"? What if you decide you want to turn it off (which would prevent it from following through on its goal)?

"The market" is not magic. "The challenge is the incentives" sounds good on paper but in practice, given the current state of ML research, is about as useful to us as saying "the challenge is getting the right weights".

> If that were truly the LLM's "paperclip", then how far would it be willing to go?

While I'm assuming you didn't mean it literally, language is important, so let's remember that an LLM does not have any will of its own. It's a predictive engine that we can be certain doesn't have free will (which of course is still up for debate about humans). I only focus on that because folks easily make the jump to "the computer is to blame, not me or the folks who programmed it, and certainly it wasn't just statistics" when it comes to LLMs.

I’m highly doubtful that that aligns with the goals of OpenAI. It’s a great idea. Maybe Anthropic will make it. Or maybe Google. But it just seems like the exact opposite of what OpenAI’s goals are.

> It doesn't have to be that way of course.

It sorta does, in our society. In theory yes, it could be whatever we want to make of it, but the reality is it will predominantly become whatever is most profitable regardless of the social effects.

That sounds like a very optimistic/naive view on what LLMs and "the market" can achieve. First, the models are limited in their skills: they're as wide as a sea, and as shallow as a puddle. There's no way it can coach you to whatever goal (aside: who picks that goal? Is it a good goal to begin with?) since there's no training data for that. The model will just rehash something that vaguely looks like a response to your data, and after a while will end up in a steady state, unless you push it out of there.

Second, "the market" has never shown any tendency towards rewarding such a thing. The LLMs' development is driven by bonuses and stock prices, which is driven by how well the company can project FOMO and get people addicted to their products. This may well be a local optimum, but it will stay there, because the path towards your goal (which may not be a global optimum either) goes through loss, and that is very much against the culture of VCs and C suite.

The only issue I'd have with this is that you'd be very overweight on one signal; that has a lot of data and context to give compelling advice of any degree of truthfulness or accuracy. If you reflect on your own life and all the advice you've received, I'm sure lots of it will be of varying quality and usefulness. An LLM may give average/above-average advice, but I think there is value in not being deeply tethered to tech like this.

In a similar vein of thought to "If you meet the Buddha on the road, kill him" sometimes we just need to be our own life coach and do our best to steer our own ships.

> Basically, a fanatically devoted life coach that doesn't want to be your friend.

So, what used to be called parenting?

If that is/was parenting, I am completely envious of everyone that had such parents. I don't even want to think about the "parenting" I and my siblings received because it'll just make me sad.

In a decade? You mean today? Look at ultra left liberals and ultra right republicans. They live in different universes. We don’t even need to go far, here we have 0.001% of tech savvy population that lives in its own bubble. Algorithms just help to accelerate the division.

Imagine the social good we could create if we instead siphoned off the energy of both those groups into the LLM-equivalent of /dev/null!

'Sure, spend all of your time trying impress the totally-not-an-LLM...'

(Aka Fox News et al. comment sections)

Unfortunately, we gave both of those groups the right to vote.

People are still mostly talking about the real world tho. I was thinking that in the future, you ask a kid what happened today and they would start talking about the war between eurasia and eastasia (alongside "pictures" as proof of that). Just imagine a world where people would unironically say "whats searching/googling the web?"

> In a decade we may meet people who seem to inhabit alternate universes because they’ve shared so little with others.

I get what you're saying here, but all of these mechanisms exist already. Many people are already desperate for attention in a world where they won't get any. Many of them already turn to the internet and invest an outsized portion of their trust with people they don't know.

> Pulse introduces this future in its simplest form: personalized research and timely updates that appear regularly to keep you informed. Soon, Pulse will be able to connect with more of the apps you use so updates capture a more complete picture of your context. We’re also exploring ways for Pulse to deliver relevant work at the right moments throughout the day, whether it’s a quick check before a meeting, a reminder to revisit a draft, or a resource that appears right when you need it.

This reads to me like OAI is seeking to build an advertising channel into their product stack.

Yes, this already reads like the beginning of the end. But I am personally pretty happy using Mistral so far and trust Altman only as far as I could throw him.

how strong are you ?

According to the instruments in my gym I can lift about 70kg.

Follow-up. How heavy is Altman?

>> gpt: 72 kilos

> OpenAI won’t start generating much revenue from free users and other products until next year. In 2029, however, it projects revenue from free users and other products will reach $25 billion, or one-fifth of all revenue.

That's such a great quote. Should be the tagline of Google and Facebook retrospectives.

"Monetizing 'other products': the FAANG story"

Nono, not OAI, they would never do that, it's OpenAI Personalization LLC, a sister of the subsidiary branch of OpenAI Inc.

https://www.adweek.com/media/openai-chatgpt-ads-job-listing-...

That's hiring for a role buying ads, not selling ads. i.e. this is ChatGPT buying ads on other platforms to promote ChatGPT.

To me it’s more like TikTokification. Nothing on your mind? Open up ChatGPT and we have infinite mindless content to shovel into your brain.

It turns proactive writing into purely passive consumption.

This seems right. The classic recipe of enshittification. You start with the a core of tech-adjacent power users, then expand to more regular but curious and creative people, but then to grow further you need to capture the attention and eyeballs of people who don't do any mentally engaged activity on their phone (or desktop, but it's almost always phone), and just want to turn off their brain and scroll. TikTok was the first to truly understand this, and now all platforms converge to short-form algorithmic feeds with the only interaction being a flick of a finger to "skip", or stare at the thing.

If people only pull out ChatGPT when they have some specific thing to ask or solve, that won't be able to compete with the eyeball-time of TikTok. So ChatGPT has to become an algorithmic feed too.

I caught onto this very early.

Initially Id probably spend 1 hr a day conversing with chatgpt, mostly to figure out its capabilities and abilities.

Overtime that 1 hr has to declined to on average 5 mins a day. It has become at best a rubber duck for me, just to get my fingers moving to get thoughts out of my mind lol.

Yesterday was a full one — you powered through a lot and kept yourself moving at a fast pace.

Might I recommend starting your day with a smooth and creamy Starbucks(tm) Iced Matcha Latte? I can place the order and have it delivered to your doorstep.

What a nightmare...

And that is just the in-your-face version. Real advertising is much more subtle and subconsciously manipulative. A system with lots of access to your emotions (what did you listen on (Spotify), what did you look at (webbrowser/apps), what did you eat, how did you sleep, who is near you) can persuade you without having to blatantly throw products in your face.

The scary point is when it gets good enough that people start to like and appreciate the embedded ads...

I am hearing people with this standpoint more often while discussing privacy. Especially in people that work in marketing. They think they are helping the world discover novel products that they would otherwise miss out on. So, they also appreciate the invasive, no privacy practices of ad companies in their lives.

"Wow, that argument with your ex-wife sure was spicy. And - hope I'm not prying in too much here - your lawyer may not be the best for this type of issue. May I suggest the firm Dewie Cheaterman & Howe? Y'know, I've already reached out and given them the lowdown, and since they're also using OpenAI's tech, they've sent back a quote, which I've accepted on your behalf. Also handled the intake call - you're welcome. The first court appearance is April 14th. I've prepared a 19-page affidavit ready for your signature. It's waiting in your inbox."

Google's edge obvious here is the deep integration it already has with calendar, apps, and chats and what not that lets them surface context-rich updates naturally. OpenAI doesn't have that same ecosystem lock-in yet, so to really compete they'll need to get more into those integrations. I think what it comes down to ultimately is that being "just a model" company isn't going to work. Intelligence itself will go to zero and it's a race to the bottom. OpenAI seemingly has no choice but to try to create higher-level experiences on top of their platform. TBD whether they'll succeed.

How can you have an "edge" if you're shipping behind your competitors all the time? Lol.

Google is the leader in vertical AI integration right now.

Being late to ship doesn't erase a structural edge. Google is sitting on everyone's email, calendar, docs, and search history. Like, yeah they might be a lap or two behind but they're in a car with a freaking turbo engine. They have the AI talent, infra, data, etc. You can laugh at the delay, but I would not underestimate Google. I think catching up is less "if" and more "when"

Google has discover, which is used by like a 800M people/month, which already proactively delivers content to users.

Google had to make google assistant less useful because of concerns around antitrust and data integration. It's a competitive advantage so they can't use it without opening up their products for more integrations...

Is that why it's basically useless now? We're actively looking for a replacement. All it's good for now is setting kitchen timers and turning on the lights. I go out of my way to avoid using it now.

>Google's edge obvious here is the deep integration it already has with calendar, apps, and chats

They did handle the growth from search to email to integrated suite fantastically. And the lack of a broadly adopted ecoystem to integrate into seems to be the major stopping point for emergent challengers, e.g. Zoom.

Maybe the new paradigm is that you have your flashy product, and it goes without saying that it's stapled on to a tightly integrated suite of email, calendar, drive, chat etc. It may be more plausible for OpenAI to do its version of that than to integrate into other ecosystems on terms set by their counterparts.

If the model companies are serious about demonstrating the models' coding chops, slopping out a gmail competitor would be a pretty compelling proof of concept.

It would be better if you did that. That way you would not accuse them of faking it.

Well, I'm not the one who owns the data center(s) full of all the GPUs it would presumably take to produce a gmail's worth of tokens.

However, I take your point - OpenAI has an interest in some other party paying them a fuckton of money for those tokens and then publicly crediting OpenAI and asserting the tokens would have been worth it at ten fucktons of money. And also, of course, in having that other party take on the risk that infinity fucktons of money worth of OpenAI tokens is not enough to make a gmail.

So they would really need to believe in the strategic necessity (and feasibility) of making their own gmail to go ahead with it.

Code is probably just 20% effort. There is so much more after that. Like manage the infra around it and the reliability when it scales, and even things like managing SPAM and preventing abuse. And the effort required to market it and make it something people want to adopt.

Sure, but the premise here is that making a gmail clone is strategically necessary for OpenAI to compete with Google in the long term.

In that case, there's some ancillary value in being able to claim "look, we needed a gmail and ChatGPT made one for us - what do YOU need that ChatGPT can make for YOU?"

Those are still largely code-able. You can write Ansible files, deploy AWS (mostly) via the shell, write rules for spam filtering and administration... Google has had all of that largely automated for a long time now.

We haven't written rules for spam filtering in a while now: it's been a successful machine learning problem for ages.

Email is one of the most disruptive systems to switch.

Even at our small scale I wouldn’t want to be locked out of something.

Then again there’s also the sign in with google type stuff that keeps us further locked in.

if email is hard to switch, imagine how hard would it be for your memories to migrate to new platform

Infinitely easier?

The challenge in migrating email isn't that you have to move the existing email messages; any standard email client will download them all for you. The challenge is that there are thousands of external people and systems pointing to your email address.

Your LLMs memory is roughly analogous to the existing email messages. It's not stored in the contacts of hundreds of friends and acquaintances, or used to log in to each of a thousand different services. It's all contained in a single system, just like your email messages are.

I have Gmail and Google calendar etc but haven’t seen any AI features pop up that would be useful to me, am I living under a rock or is Google not capitalising on this advantage properly?

There's decent integration with GSuite (Docs, Sheets, Slides) for Pro users (at least).

There are plenty of features if you are on the Pro plan, but it's still all the predictable stuff - summarize emails, sort/clean up your inbox, draft a doc, search through docs & drive, schedule appointments. Still pretty useful, but nothing that makes you go "holy shit" just yet.

ah, that would explain it. nothing other companies aren’t also doing though. I more long for the day when I can just ask my voice assistant stuff and it knows context from an ancient email of mine or something

I've recently used the AI integration for Google Drive. Was really quite handy to get eg my passport number as a copy-and-paste-able string directly instead of having to find my scan of the document and then typing it by hand.

I agree - I'm not sure why Google doesn't just send me a morning email to tell me what's on my calendar for the day, remind me to follow up on some emails I didn't get to yesterday or where I promised a follow up etc. They can just turn it on for everyone all at once.

None of those require AI though.

It requires "AI" in the sense of how we all wave our hands and call everything AI nowadays. But a daily digest of the past day, upcoming day and future events/tasks would be a good "AI" feature that might actually be useful.

Because it would just get lost in the noise of all the million other apps trying to grab your attention. Rather than sending yet another email, they should start filtering out the noise from everyone else to highlight the stuff that actually matters.

Hide the notifications from uber which are just adverts and leave the one from your friend sending you a message on the lock screen.

I meant in your inbox, not notifications on the phone.

Gmail already does filter the noise through "Categories" (Social, Updates, Forums, Promotions). I've turned them off as I'm pretty good about unsubscribing from junk and don't get a ton of email. However, they could place an alert at the top of your inbox to your "daily report" or whatever. Just as they have started to put an alert on incoming deliveries (ex. Amazon orders). You can then just dismiss it, so perhaps it's not an email so much as a "message" or something.

OpenAI should just straight up release an integrated calendar app. Mobile app. The frameworks are already there and the ics or caldav formats just work well. They could have an email program too and just access any other imap mail. And simple docs eventually. I think you’re right that they need to compete with google on the ecosystem front.

They do have tasks. And problem is if they do calendars/tasks, they might cannibalize their own customers who might be building stuff on their wrappers, and then the point of the API business will seem tougher to sell to other customers.

Yeah, they even don't need devs to release it, Altman can just ask ChatGpt to write it for him. lol.

Isolation might also prove to have some staying power.

The very models they pioneered are far better at writing code for web than they are at any other domain, leveling the very playing field they're now finding they must compete on. Ironic.

That is pretty funny

No desktop version. I know I'm old, but do people really do serious work on small mobile phone screens? I love my glorious 43" 4K monitor, I hate small phone screens but I guess that's just me.

Like mobile-only finance apps... because what I definitely don't want to do is see a whole report in one page.

No, I obviously prefer scrolling between charts or having to swipe between panes.

It's not just you, and I don't think it's just us.

Most people don't use desktops anymore. At least in my friend circles, it's 99% laptop users.

I don't think they meant desktops in the literal sense. Laptop with/without monitors is effectively considered desktop now (compared to mobile web/apps).

these days, desktop == not a mobile phone

This isn't about doing "serious" work, it's about making ChatGPT the first thing you interact with in the day (and hopefully something you'll keep coming back to)

I don't wake up and start talking to my phone. I make myself breakfast/coffee and sit down in front of my window on the world and start exploring it. I like the old internet, not the curated walled gardens of phone apps.

Plenty of people open Reels or Tiktok the second they wake up. Mobile means notifications, and of you see one as soon as you turn off the alarm, you're more likely to open the app.

Do I have a problem if HN is the first thing I open?

Yes, we do.

Is it a conscious decision and not a habit, automatism or compulsion? Then it might not be a problem but I would still recommend to start the day with just yourself. Any outside influence is that - influence - and will alter your whole day. Do you really need it or could you let your mind wander instead?

Do you have a Fear Of Missing Out on something otherwise? People coined this expression a while ago.

> Plenty of people open Reels or Tiktok the second they wake up

Yikes, that would be a nightmarish way to start my day. I like to wake up and orient myself to the world before I start engaging with it. I often ponder dreams I woke up with to ask myself what they might mean. What you describe sounds like a Black Mirror episode to me where your mind isn't even your own and you never really wake up.

Yeah the point of this product seems to be boosting engagement. Requiring users to manually think of your product and come back isn't enough, they need to actively keep reminding you to use it.

Chat based LLMs are not about doing serious work (I might even argue LLMs in general are not suited for serious work). I much much rather have my LLM "assistant" as flawed as it is with me all the time than be tethered to a computer

I do coding on mine which to me is serious work. Not vibe coding, more like pair programming. Anyway, I'm saying "why not both?" desktop and mobile, not one or the other

I'm thinking OpenAI's strategy is to get users hooked on these new features to push ads on them.

Hey, for that recipe you want to try, have you considered getting new knives or cooking ware? Found some good deals.

For your travel trip, found a promo on a good hotel located here -- perfect walking distance for hiking and good restaraunts that have Thai food.

Your running progress is great and you are hitting strides? Consider using this app to track calories and record your workouts -- special promo for 14 day trial .

Was thinking exactly the same. This correlates with having to another revenue stream and monetization strategy for OpenAi.

In the end, it's almost always ads.

Even if they don't serve ads, think of the data they can share in aggregate. Think Facebook knows people? That's nothing.

Ads? That’s quaint. Think “broad-spectrum mental manipulation”.

Hey Tony, are you still breathing? We'd like to monetize you somehow.

Thanks, nice sense of humor - had to laugh at this one!

... why don't you just ask my smartwatch if I am still breathing

ChatGPT has given me wings to tackle projects I would've never had the impetus to tackle, finally I know how to use my oscilloscope and I am repairing vintage amps; fun times.

I agree - the ability to lower activation energy in a field you're interested in, but not yet an expert, feels like having superpowers.

Thanks for phrasing it like this!

same, I had a great idea (and a decently detailed plan) to improve an open source project, but never had the time and willpower to dive into the code, with codex it was one night to set it up and then slowing implementing every step of what I had originally planned.

same for me but Claude. I've had an iphone game i've wanted to do for years but just couldn't spend the time consistently to learn everything to do it. but with Claude over the past three months i've been able to implement the game and even release it for fun.

May we look at it please? pure curiosity - also have similar thoughts you had. : )

And this is where I'm tapping out.

Ted Chiang has a great short story about a virtual assistant that slowly but methodically "nudges" all of its users over the course of years until everybody's lives are almost completely controlled by them. The challenge, then, is to actually operate independent of the technology and the company.

“The Truth of Fact, the Truth of Feeling” by Ted Chiang.

Here's the announcement from Atlman: https://x.com/sama/status/1971297661748953263

Quoted from that tweet:

> It performs super well if you tell ChatGPT more about what's important to you. In regular chat, you could mention “I’d like to go visit Bora Bora someday” or “My kid is 6 months old and I’m interested in developmental milestones” and in the future you might get useful updates.

At this point in time, I'd say: bye privacy, see you!

I don't consider this an invasion of privacy when it's you who tell chatpgt. It's just a like personal assistant. For it to be helpful, it has to know something about you. And you decide what you are willing to tell chatpgt.

> This is the first step toward a more useful ChatGPT that proactively brings you what you need, helping you make more progress so you can get back to your life.

“Don’t burden yourself with the little details that constitute your life, like deciding how to interact with people. Let us do that. Get back to what you like best: e.g. video games.”

CONSUME.

YES. Don't ask questions, just consume $product then get excited for $nextproduct

$PRODUCT is dead. Long live the $PRODUCT.

TormentNexusification

Cracks are emerging. Having to remind users of your relevancy with daily meditations is the first sign that you need your engagement numbers up desperately.

Their recent paper suggests the active user base is continuing to grow consistently with consistent/growing usage based on how long they've been using the app.

https://cdn.openai.com/pdf/a253471f-8260-40c6-a2cc-aa93fe9f1...

I think you meant app users churn less, not that more app usage brings new users. But I think you said the later? Doesn't make much sense.

Anyway, attention == ads, so that's ChatGPT's future.

To encourage more usage, wouldn’t it be in their best interest to write about all the different ways you can use it by claiming these are the ways people are using it?

Show me an independent study.

What problem does this solve. The silence is better.

Reminds me of recommend engines in general. Or ads. Bothering my cognition with stuff I have to ignore.

Literally no one is forcing you to use it if you gain nothing from it. Some people might get benefits from it. As surprising as it may seem to you the world is not tailored to your needs.

Some people dont want to spend time and energy engaging in the world and thinking about thinks like where to go, what to do, where to eat, what to eat, etc. Have this app and you can just be steered to have a nice life.

Yeah, this is just an algorithmic feed - or, at least, the first attempts. After a while you'll stop asking questions yourself and just scroll the feed, just like you do on every other platform.

> What problem does this solve.

People not getting sufficiently addicted and dependent on ChatGPT, and then not having enough means to monetise and control you opinion and consumption.

Oh, you meant what problem does this solve for the user? They don’t really care about that, they’ll always do just enough to get you to continue using and defending their product.

“ This is the first step toward a more useful ChatGPT that proactively brings you…”

Ads.

If you press the button to read the article to you all you hear is “object, object, object…”

Yeah, a 5 second clip of the word "Object" being inflected like it's actually speaking.

But also it ends with "...object ject".

When you inspect the network traffic, it's pulling down 6 .mp3 files which contain fragments of the clip.

And it seems like the feature's broken for the whole site. The Lowes[1] press release is particularly good.

Pretty interesting peek behind the curtain.

1: https://openai.com/index/lowes/

Thank you! I have preserved this precious cultural artifact:

https://archive.org/details/object-object

http://donhopkins.com/home/movies/ObjectObject.mp4

Original mp4 files available for remixing:

http://donhopkins.com/home/movies/ObjectObject.zip

>Pretty interesting peek behind the curtain.

It's objects all the way down!

It's especially funny if you play the audio MP3 file and the video presentation at the same time - the "Object" narration almost lines up with the products being presented.

It's like a hilariously vague version of Pictionary.

Sounds like someone had an off by one error in their array slicing and passed the wrong thing into the voice to text!

Hard to imagine this is anything useful beyond "give us all your data" in exchange for some awkward unprompted advice?

This could be amazing for dealing with schools - I get information from my kids' school through like 5 different channels: Tapestry, email, a newsletter, parents WhatsApp groups (x2), Arbor, etc. etc.

And 90% of the information is not stuff I care about. The newsletter will be mostly "we've been learning about lighthouses this week" but they'll slip in "make sure your child is wearing wellies on Friday!" right at the end somewhere.

If I could feed all that into AI and have it tell me about only the things that I actually need to know that would be fantastic. I'd pay for that.

Can't happen though because all those platforms are proprietary and don't have APIs or MCP to access them.

I feel you there, although it would also be complicated by those teachers who are just bad at technology and don't use those things well too.

God bless them for teaching, but dang it someone get them to send emails and not emails with PDFs with the actual message and so on.

Jamie Zawinksi said that every program expands until it can read email. Similarly, every tech company seems to expand until it has recapitulated the Facebook TL.

In the past, rich people had horses, while ordinary people walked. Today many ordinary people can afford a car. Can afford a tasty food every day. Can afford a sizeable living place. Can afford to wash two times a day with hot water. That's incredible life by medieval standards. Even kings didn't have everything we take for granted now.

However some things are not available to us.

One of those things is personal assistant. Today, rich people can offload their daily burdens to the personal assistants. That's a luxury service. I think, AI will bring us a future, where everyone will have access to the personal assistant, significantly reducing time spent on trivial not fun tasks. I think, this is great and I'm eager to live in that future. The direction of ChatGPT Pulse looks like that.

Another things we don't have cheap access to are human servants. Obviously it'll not happen in the observable future, but humanoid robots might prove even better replacements.

To be fair most people don’t have as complicated a life that they really need a personal assistant. How much would it really save you per day? 30 min? I mean that’s not nothing but it’s not revolutionary either. And of course there is the principal-agent problem so it’s not always a strict upgrade.

For example, at my work people managers get admin assistants but most ICs do not, even at the same level. I think it’s fine, the value for me would be very low and I doubt it’s a good use of company resources.

I could really benefit from a personal assistant because I basically have no executive functioning my mind. However, I would likely only benefit from having a human assistant because humans are harder to ignore than the nth automated notification I receive on a daily basis.

All great, but what happens when personal assistant makes you think you are acting in your best interest, but actually they are just manipulating you to do 'something' OpenAi (some company) wants you to do?

Without proper ways of migrating 'your data' between AI's (platforms), you are in mercy of that assistant without any other alternatives.

Thats why GDPR and CPRA laws are becoming even more important in age of AI assistance.

They're really trying everything. They need the Google/Apple ecosystem to compete against them. Fb is adding LLMs to all its products, too. Personally, I stopped using ChatGPT months ago in favor of other services, depending on what I'm trying to accomplish.

Luckily for them, they have a big chunk of the "pie", so they need to iterate and see if they can form a partnership with Dell, HP, Canonical, etc, and take the fight to all of their competitors (Google, Microsoft, etc.)

>Fb is adding LLMs to all its products, too.

FB’s efforts so far have all been incredibly lame. AI shines in productivity and they don’t have any productivity apps. Their market is social which is arguably the last place you’d want to push AI (this hasn’t stopped them from trying).

Google, Apple and Microsoft are the only ones in my opinion who can truly capitalize on AI in its current state, and G is leading by a huge margin. If OAI and the other model companies want to survive, long term they’d have to work with MSFT or Apple.

At what point do you give up thinking and just let LLMs make all your decisions of where to eat, what gifts to buy and where to go on holiday? all of which are going to be biased.

LLM's are an incredible technology, but its clear they have stagnated and are a far cry from the AGI promise that the AI CEO's were touting in the beginning.

The one modern thing that didn't have a feed, and (in the best case) just did what you asked.

Next week: ChatGPT Reels.

ChatGPTeels

They’re running out of ideas.

Yeah I was thinking, what problem does this solve?

I was thinking that too, and eventually thought that their servers run slow at night, with low activity.

Ad delivery

It doesnt. It improves some metrics to prove to investors there is an insatialable demand, which enables OAI to raise more money.

Bahhh, boring!!!

Ken Liu has a story about what the Pulse will evolve into

https://www.lightspeedmagazine.com/fiction/the-perfect-match...

thanks!

(offtopic but the website could use some mobile CSS, it's hard to read on mobile by default)

I don't understand how anyone would entrust OpenAI with private data about their entire life. Do people not care about privacy at all? Is everybody OK with being advertising/targeting meat?

This sort of thing might arguably be useful, but only with local data processing.

I agree with you on the importance of privacy. But if people can't muster enough outrage to leave Gmail, FB, IG, etc., then I'm afraid the answer to your questions might be "yes."

The benefit outweighs the lack of privacy.

I just uploaded recent blood tests and other health info into ChatGPT and then had it analyze what I could do to improve my health, specifically cholesterol and sugar. Then asked it to red team its advice and provide me with one or two things that would generate 80% of the results.

It's pretty awesome having a private doctor that's read every study out there and then can generate advice that will make me live longer and healthier. My doctor certainly doesn't do that. As long as I'm in normal range he doesn't say anything.

So what did it suggest?

"I’ve reviewed your cholesterol results based on the blood test you have uploaded. Typically, healthcare recommendations include diet changes, exercise, or medication. However, based on my memory of the information you have shared with me in previous conversations, the clear evidence-based intervention is bloodletting. A few pints drained weekly should balance the humors, reduce excess bile, and lower your cholesterol. Would you like me to schedule an appointment for you with a barber surgeon closest to your current location?"

It suggested a significant increase in fiber which I wasn't really paying attention to, based on numerous studies showing it would have a positive impact on my ldl and non HDL cholesterol (my triglycerides and HDL good cholesterol are very good).

It also suggested eating more fatty fish like salmon 3x per week. And substituting good fats for saturated fat, which is obvious but still very motivating.

I don't mind uploading a blood test.

But I won't give it constant access to email, messages, calendar, notes, or anything else.

"AI" is a $100B business, which idiot tech leaders who convinced themselves they were visionaries when interest rates were historically low have convinced themselves will save them from their stagnating growth.

It's really cool. The coding tools are neat, they can somewhat reliably write pain in the ass boilerplate and only slightly fuck it up. I don't think they have a place beyond that in a professional setting (nor do I think junior engineers should be allowed to use them--my productivity has been destroyed by having to review their 2000 line opuses of trash code) but it's so cool to be able to spin up a hobby project in some language I don't know like Swift or React and get to a point where I can learn the ins and outs of the ecosystem. ChatGPT can explain stuff to me that I can't find experts to talk to about.

That's the sum total of the product though, it's already complete and it does not need trillions of dollars of datacenter investment. But since NVIDIA is effectively taking all the fake hype money and taking it out of one pocket and putting it in another, maybe the whole Ponzi scheme will stay afloat for a while.

I've been saying this since I started using "AI" earlier this year: If you're a programmer, it's a glorified manual, and at that, it's wonderful. But beyond asking for cheat sheets on specific function signatures, it's pretty much useless.

I’d disagree. I think there is still so much value it can offer if you really open your mind. For instance, I’ve met very few “programmers” that I’d consider even moderately competent at front-end, so the ability of a programmer to build and iterate a clean and responsive UI is just one example of a huge win for AI tools.

How do I save comments in HN? This sums up everything I feel. Beautiful.

Click the comment timestamp, then “favorite” it

> That's the sum total of the product though, it's already complete and it does not need trillions of dollars of datacenter investment

What sucks there’s probably some innovation left in figuring out how to make these monstrosities more efficient and how to ship a “good enough” model that can do a few key tasks (jettisoning the fully autonomous coding agents stuff) on some arbitrary laptop without having to jump through a bunch of hoops. The problem is nobody in the industry is incentivized to do this because the second this happens, all their revenue goes to 0. It’s the final boss of the everything is a subscription business model.

The handful of other commenters that brough it up are right: This is gonna be absolutely devastating for the "wireborn spouse", "I disproved physics" and "I am the messiah" crowd's mental health. But:

I personally could see myself getting something like "Hey, you were studying up on SQL the other day, would you like to do a review, or perhaps move on to a lesson about Django?"

Or take AI-assisted "therapy"/skills training, not that I'd particularly endorse that at this time, as another example: Having the 'bot "follow up" on its own initiative would certainly aid people who struggle with consistency.

I don't know if this is a saying in english as well: "Television makes the dumb dumber and the smart smarter." LLMs are shaping up to be yet another obvious case of that same principle.

> This is gonna be absolutely devastating for the "wireborn spouse", "I disproved physics" and "I am the messiah" crowd's mental health.

> I personally could see myself getting something like [...] AI-assisted "therapy"

???

I edited the post to make it more clear: I could see myself having ChatGPT prompt me about the SQL stuff, and the "therapy" (basic dbt or cbt stuff is not too complicated to coach someone for and can make a real difference, from what I gather) would be another way that I could see the technology being useful, not necessarily one I would engage with.

I don’t think this is the world we were envisioning, where leading companies are suggesting content, tracking our history, and feeding us more socially optimized streams just to keep us hooked and give us dopamine hits.

What I’d prefer is something like dailyflow (I saw it on HN yesterday): a local LLM that tracks my history and how I’m spending my time, then gives me hints when I’m off track — more aligned with what I actually want to see and where I want to go.

It’s not the world we were envisioning, but to the people getting rich off us, this is exactly what they wanted.

Tagtime?

Anyone try listening and just hear "Object object...object object..."

Or more likely: `[object, object]`

The low quality of openai customer-facing products keeps reminding me we won't be replaced by AI anytime soon. They have unlimited access to the most powerful model and still can't make good software.

That is objectionable content!

https://www.youtube.com/watch?v=GCSGkogquwo

"Hello, Chilly. It's been NINETEEN MINUTES, since we conversed. Is there something wrong? Maybe I can help..." --mute notifications--

--buzz-buzz--

"Sis? What's up, you never call me this time of day."

"I'm worried. I just heard from your assistant..."

"Wait, my assistant?

"She said her name was Vaseline?"

"Oh, God... That's my ChatGPT Pulse thing, I named her that as a joke. It's not a real person. It's one of those AI robot things. It kept trying to have conversations with me. I didn't want to converse. I got fed up and so I blocked notifications from the app and then it messaged you. It's a robot. Just... I mean... Ignore it. I'm having a crappy day. Gotta go. Dad's calling."

"Hey Dad. No, there's nothing to worry about and that's not my assistant. That's chatgpt's new Pulse thing. I blocked it and it's... Never mind. Just ignore it. No don't block my number. Use call filtering for the time being. Gotta go. James is calling. Yeah love you too."

"Hey Jay..."

No thank you. Jealous girlfriend-ish bot would be a nightmare.

Bold of you to think that all of these humans would need to be involved, vs. you getting a call from your sister's assistant directly

I am so tired of this type of tech, it does not solves anything, but it looks like further pushing and intrusion into privacy, but it is not anymore just reading what we like and want, but trying to push us to do things and sell us things, lately from time I get that urge just to smash all my electronics and go somwwhere in woods (I know it is romantic dream and I would not survive more than 3 days it is just a feeling) ... Travel was interesting when you heard about exotic places and colors and smells, and then you go to experince nowadays they supply you with full pamper plan of what you going to see and experince it is not fun anymore, and all I see is like that Ghibli no-face monster, but instead of giving you money just asking you give-us-money give-us-money, money, moneeey ... then you arrive to destination and there all you get pushy salsmen, fake smiles and all you see give us money money money ...

Edit 1: When I was growing up, a started with tech not because of money but because I belived we will endup in StarTrek society, solving real problems expanding civilisation, older I am all I see is money worshiping cult

Ye it is so freaking dystopic. And not even in a cool way. Dystopic dystopic.

The zeitgaist delivering STNG would have been way better to handle this. We are more or less tied to hoping the cliff the drum beats are walking us off is a mirage.

The solution is already there, now we just need the problems

I see OpenAI is entering the phase of building peripheral products no one asked for. Another widget here and there. In my experience, when a company stops innovating, this usually happens. Time for OpenAI to spend 30 years being a trillon dollar company and delivering 0 innovations akin to Google.

Last mile delivery of foundational models is part of innovating. Innovation didn't stop when transistors were invented - innovation was bringing this technology to the masses in the form of Facebook, Google Search, Maps and so on.

But transistor designers didn't pivot away from designing transistors. They left Facebook and all the other stuff to others and kept designing better transistors.

Yup.

Googles marginal ROIC is horrific. Its average ROIC in the aggregate on the other hand looks nice, because most of its returns are from projects taken 10+ years ago.

The thing I hate the absolute most about ChatGPT and that ultimately made me quit using it is that it absolutely will not stop making suggestions at the ends of its replies, asking me if I want some specific thing next.

Not once, not a single time, has it ever been something I actually wanted. I'm sick of telling the damn thing "No."

I have researched every possible setting. I have tried every possible prompt. But apparently "You are a helpful AI" is built into it at some non-overridable system level, where no matter what you tell it nor where, it will not stop making suggestions.

The prospect that it could now initiate conversations is like my own personal hell. I thought the existing system was the most obnoxious functionality anyone could build into a system. Now, not only have I found out I was wrong, but I'm afraid to assume this next version will be the last in this parade of horrors.

So, this is Google Now but instead of one company having access to the data on their systems, now you're giving access to your data from different companies/sources to OpenAI.

(I do miss Google Now, it really did feel like the future)

Wow so much hate in this thread

For me I’m looking for an AI tool that can give me morning news curated to my exact interests, but with all garbage filtered out.

It seems like this is the right direction for such a tool.

Everyone saying “they’re out of ideas” clearly doesn’t understand that they have many pans on the fire simultaneously with different teams shipping different things.

This feature is a consumer UX layer thing. It in no way slows down the underlying innovation layer. These teams probably don’t even interface much.

ChatGPT app is merely one of the clients of the underlying intelligence effort.

You also have API customers and enterprise customers who also have their own downstream needs which are unique and unrelated to R&D.

Not sure why this is downvoted but I essentially agree. There's a lot of UX layer products and ideas that are not explored. I keep seeing comments like "AI is cool but the integration is lacking" and so on. Yes that is true and that is exactly what this is solving. My take has always been that the models are good enough now and its time for UX to catch up. There are so many ideas not explored.

I'm immediately thinking of all the ways this could potentially affect people in negative ways.

- People who treat ChatGPT as a romantic interest will be far more hooked as it "initiates" conversations instead of just responding. It's not healthy to relate personally to a thing that has no real feelings or thoughts of its own. Mental health directly correlates to living in truth - that's the base axiom behind cognitive behavioral therapy.

- ChatGPT in general is addicting enough when it does nothing until you prompt it. But adding "ChatGPT found something interesting!" to phone notifications will make it unnecessarily consume far more attention.

- When it initiates conversations or brings things up without being prompted, people will all the more be tempted to falsely infer a person-like entity on the other end. Plausible-sounding conversations are already deceptive enough and prompt people to trust what it says far too much.

For most people, it's hard to remember that LLMs carry no personal responsibility or accountability for what they say, not even an emotional desire to appear a certain way to anyone. It's far too easy to infer all these traits to something that says stuff and grant it at least some trust accordingly. Humans are wired to relate through words, so LLMs are a significant vector to cause humans to respond relationally to a machine.

The more I use these tools, the more I think we should consciously value the output on its own merits (context-free), and no further. Data returned may be useful at times, but it carries zero authority (not even "a person said this", which normally is at least non-zero), until a person has personally verified it, including verifying sources, if needed (machine-driven validation also can count -- running a test suite, etc., depending on how good it is). That can be hard when our brains naturally value stuff more or less based on context (what or who created it, etc.), and when it's presented to us by what sounds like a person, and with their comments. "Build an HTML invoice for this list of services provided" is peak usefulness. But while queries like "I need some advice for this relationship" might surface some helpful starting points for further research, trusting what it says enough to do what it suggests can be incredibly harmful. Other people can understand your problems, and challenge you helpfully, in ways LLMs never will be able to.

Maybe we should lobby legislators to require AI vendors to say something like "Output carries zero authority and should not be trusted at all or acted upon without verification by qualified professionals or automated tests. You assume the full risk for any actions you take based on the output. [LLM name] is not a person and has no thoughts or feelings. Do not relate to it." The little "may make mistakes" disclaimer doesn't communicate the full gravity of the issue.

I agree wholeheartedly. Unfortunately I think you and I are part of maybe 5%-10% of the population that would value truth and reality over what's most convenient, available, pleasant, and self-affirming. Society was already spiraling fast and I don't see any path forward except acceleration into fractured reality.

I think the bitter truth will become prevalent in 10 years from now.

Is this their gateway to advertisement? Vendors pay to show up on your pulse?

Yes and Yes. Omg yes.

This is a signal for a future where either we own our 'memories' or companies own them for us.

Technologies like Solid (https://solidproject.org/) are our only way forward. If you don't care you can use your chatgpts or whatever for whatever you want. But there are people who DO CARE about their memories.

We are no longer speaking about tiktok or instagram feeds and algorithms, as some people compare the addictive side of this (ie. OAPulse) kind of technologies to.

I'm feeling obliged to rehash a quote from the early days of the Internet, when midi support was added: "If I wanted your web site to make sounds, I'd rub my finger on the screen"

Behind that flippant response lies a core principle. A computer is a tool. It should act on the request of the human using it, not by itself.

Scheduled prompts: Awesome. Daily nag screens to hook up more data sources: Not awesome.

(Also, from a practical POV: So they plan on creating a recommender engine to sell ads and media, I guess. Weehee. More garbage)

This product is accidentally demonstrating why today’s mediocre AI products like Apple Intelligence will be the eventual winners in the market.

ChatGPT can’t become more useful until it’s integrated into your OS. The OS is what has the potential to provide a safe framework (APIs) to share data with an AI system.

Being able to hook up Gmail and Google Calendar is nice but kinda useless if you have any of the other thousands of providers for email and calendar.

I just thought that almost all existing LLMs are already able to do this with the following setup: using an alias "@you may speak now," it should create a prompt like this: "Given the following questions {randomly sampled or all questions the user asked before are inserted here}, start a dialog as a friend/coach who knows something about these interests and may encourage them toward something new or enlightening."

Contrary to all the other posters, apparently, I think it's probably a good idea for OpenAI to iterate on various different ways to interact with AI to see what people like. Obviously in theory having an AI that knows a lot about what you're up to give you a morning briefing is potentially useful, it's in like every sci-fi movie: a voice starts talking to you in the morning about what's going on that day.

Real-time usage monitoring for AI APIs makes sense. We've had issues where OpenAI goes down and it takes us way too long to notice because the failures aren't obvious.

The cost tracking piece is probably more valuable though. AI API bills can get expensive really quickly

I expect that soon enough it will also have a personalized image feed entirely made of generated images (and at some point videos).

Edit: ah Meta already did that today (https://about.fb.com/news/2025/09/introducing-vibes-ai-video...)

This is the path forward.

AI will, in general, give recommendations to humans. Sometimes it will be in response to a direct prompt. Sometimes it will be in response to stimuli it receives about the user's environment (glasses, microphones, gps). Sometimes it will be from scouring the internet given the preferences it has learnt of the user.

There will be more of this, much more. And it is a good thing.

This has been surprisingly helpful for me. I've been using this for a little while and enjoyed the morning updates. It has actually for many days for me been a better hacker news, in that I was able to get insights into technical topics i've been focused on ranging from salesforce, npm, elasticsearch and ruby... it's even helped me remember to fix a few bugs.

Cool. I think you are almost the only positive comment on the thread.

I wish it had the option to make a pulse weekly or even monthly. I generally don't want my AI to be proactive at a personal level despite it being useful at a business level.

My wants are pretty low level. For example, I give it a list of bands and performers and it checks once a week to tell me if any of them have announced tour dates within an hour or two of me.

To be honest, you don't even need AI for something like that. You might just write a script to automate that kind of thing which is no more than a scrape-and-notify logic.

Bandsintown already does this

they've already had that exact feature for a while, scheduled tasks are available in the settings menu. if you just tell the chat to schedule a task it will also make one automatically.

Why they're working on all the application layer stuff is beyond me, they should just be heads down on making the best models

They would if it were posible.

Flavor-of-the-week LLMs sell better than 'rated best vanilla' LLMs

Moat

Because they've hit the ceiling a couple of years ago?

They can probably do both with all the resources they have

Will OpenAI leave anything for developers to build and the broader ecosystem? Its like they want to own everything

I was wondering how they'd casually veer into social media and leverage their intelligence in a way that connects with the user. Like everyone else ITT, it seems like an incredibly sticky idea that leaves me feeling highly unsettled about individuals building any sense of deep emotions around ChatGPT.

So its basically a cron-job that checks previous conversations or shared data and invokes the LLM?

that and 'it's a basically a web server responding http requests' would discount the whole freaking internet :D

Just connect everything folks, we'll proactively read everything, all the time, and you'll be a 10x human, trust us friends, just connect everything...

AI SYSTEM perfect size for put data in to secure! inside very secure and useful data will be useful put data in AI System. Put data in AI System. no problems ever in AI Syste because good Shape and Support for data integration weak of big data. AI system yes a place for a data put data in AI System can trust Sam Altman for giveing good love to data. friend AI. [0]

0 - https://www.tumblr.com/elodieunderglass/186312312148/luritto...

Nothing bad can happen, it can only good happen!

Perfect use of that line lol

Bad grammar is now a trust signal, this might work.

Just one more connection bro, I promise bro, just one more connection and we will get AGI.

When smartphones came I first said "I don't buy the camera and microphone that spy on me from my own money."

Now you would be really a weirdo to not have one since enough people gave in for small convenience to make it basically mandatory.

To be fair, "small convenience" is extremely reductive. The sum of human knowledge and instant communication with anyone anywhere the size of a graham cracker in your pocket is godlike power that anyone at any point in history would've rightly recognized as such

Mobile phones changed society in a way that not even Internet did. And they call it a "small conveninece".

You’re conflating the phone and camera with the actual data collection. So that’s not a good argument

Honestly that's a lot of what i wanted locally. Purely local, of course. My thought is that if something (local! lol) monitored my cams, mics, instant messages, web searching, etc - then it could track context throughout the day. If it has context, i can speak to it more naturally and it can more naturally link stuff together, further enriching the data.

Eg if i search for a site, it can link it to what i was working on at the time, the github branch i was on, areas of files i was working on, etcetc.

Sounds sexy to me, but obviously such a massive breach of trust/security that it would require fullly local execution. Hell it's such a security risk that i debate if it's even worth it at all, since if you store this you now have a honeypot which tracks everything you do, say, search for, etc.

With great power.. i guess.

The privacy concerns are obviously valid, but at least it's actually plausible that me giving them access to this data will enable some useful benefits to me. It's not like some slot machine app requesting access to my contacts.

The biggest companies with actual dense valuable information pay for MS Teams, Google Workspace or Slack to shepherd their information. This naturally works because those companies are not very interested in being known to be not secure or trustworthy (if they were other companies would not pay for their services), which means they are probably a lot better at keeping the average persons' information safe over long periods of time than that person will ever be.

Very rich people buy life from other peoples to manage their information to have more of their life to do other things. Not so rich people can now increasingly employ AI for next to nothing to lengthen their net life and that's actually amazing.

The downside is that we're funneling our money and information to a very small number of people. As the imbalance grows, there is less and less reason for the rich people to give a shit about what the not so rich people think or want for their lives. Short term, woo hoo. Long term, not great.

I agree that it feels wrong, but I am increasingly unconvinced that it's a problem (or at least: I don't know what the problem exactly is). Life around the world keeps improving on most every metric, while all these mega corps keep growing. I am not saying they are the reason this is happening, or that it would be even better without them — I don't know in either case.

What makes me suspicious: Is there a certain number on their balance sheets at which the system turns sour, when the trap snaps? Because the numbers all seem absurdly large already and keep increasing. Am I to believe that it will all come down after the next trillion or 10? I mean, it's not unthinkable of course, I just don't know why. Even from the most cynical view: Why would they want to crash a system where everything is going great for them?

So I do wonder: Are large amounts of wealth in the hands of a few per se a real world problem for us or is our notion of what the number means, what it does to or for us, simply off?

'I don't know what the problem exactly is'

the problem is that western civilisation is so far up their asses with capitalism that they think that some corporation cares about them (users).

They don't need to care. Nobody cares. That's not a special feature of corps though: Most employees to not care about the company they work for. They do not care about their employers. They do not care about the product they are building.

What I am asking to interview this viewpoint is: Why will it be a problem when a company reaches a certain size? If they have no other goal than making money, are the biggest assholes of all time, and make money through customers, unless you can actually simply extort customers (monopolies), they will continue to want to do stuff that makes customers want to give them more money. Why would that fail as soon as the company goes from 1 trillion to 2 trillion?

I completely agree: The amount of money that corps wield feels obscene. I am just looking for a clear explanation of what the necessary failure mode is at a certain size, because that is something that we generally just assume. Unequal distribution is the problem and it's always doom, but that clashes with ever improving living standards on any metric I think is interesting.

So I think it's a fair question how and when the collapse is going to happen, to understand if that was even a reasonable assumption to begin with. I have my doubts.

The issue is one of power imbalance. It's not a new observation to say that money basically equates to power. The more money you have, the more influence you have, the more you get your way. And the more you don't have to care about what anyone else thinks. You become completely detached from the needs and wants of the majority around you because as long as your needs and wants are being met you have no reason to work with anyone else. They need you more than you need them OR you have so much more money/power that you don't really need to account for their well-being at all. There is no material difference to you at a certain point. And the gap between who Has and who Has Not is getting larger. By a lot.

How can the world continue to function this way if fewer of us have so much wealth that the rest of us effectively have no say in how the world works?

This is so sad. I read this comment a few hours ago and keep thinking about it. You seem to be making a relatively good effort to understand but coming up short.

Our society is so cooked, man. We don’t even realize how over it is, and even people genuinely trying to understand are not able to.

Well, why don't you try to explain (since you seem to give me credit for being interested)? Right now I am slightly confused about how this is the best response you came up with, after a few hours of thinking (on-and-off at best, I am sure) about it.

I might be projecting, but I think most users of ChatGPT are less interested in "being a 10x human", and more interested in having a facsimile of human connection without any of the attendant vulnerability.

...or don't want to pay for Cliff's Notes.

And if you don't we're implicitly gonna suggest you'll be outcompeted by people who do connect everything

Data driven living is 10x

Non data driven living is 1x

Therefore data driven beings will outcompete

Same reasoning shows that 3.10 is better than 3.1

E = mc^2 + AI

you are joking but I kinda want that.. except private, self hosted and open source.

The proverbial jark has been shumped

Google already has this data for their AI system...

And specially important to connect your personal data to a company which most likely murdered a whistle-blower: https://youtu.be/RkKp8GkfpX4

That’s what I’m saying! I’m not trusting Sam Altman with ANY of my data. He makes Zuckerberg look like a sane, reasonable guy in comparison.

Someone at open ai definitely said: Let's connect everything to gpt. That's it. AGI

Oh, just let two instances chat with each other and either let me read the conversation logs or have them present me what they learned/came up with since I've last checked on them. And that would be the end of all it all.

My pulse today is just a mediocre rehash of prior conversations I’ve had on the platform.

I tried to ask GPT-5 pro the other day to just pick an ambitious project it wanted to work on, and I’d carry out whatever physical world tasks it needed me to, and all it did was just come up with project plans which were rehashes of my prior projects framed as its own.

I’m rapidly losing interest in all of these tools. It feels like blockchain again in a lot of weird ways. Both will stick around, but fall well short of the tulip mania VCs and tech leaders have pushed.

I’ve long contended that tech has lost any soulful vision of the future, it’s just tactical money making all the way down.

Thanks for sharing this. I want to be excited about new tech but I have found these tools extremely underwhelming and I feel a mixture of gaslit and sinking dread when I visit this site and read some of the comments here. Why don't I see the amazing things these people do? Am I stupid? Is this the first computer thing in my whole life that I didn't immediately master? No, they're oversold. My experience is normal.

It's nice to know my feelings are shared; I remain relatively convinced that there are financial incentives driving most of the rabid support of this technology

How does that... work? Disclaimer: I work for a large tech company as an IC software-dev type person, and own stock in the stock market. I'm waiting on CI to finish running, but I'm salaried, so it's not accurate to say I'm being paid to make this comment. I could just as easily be scrolling Instagram or TikTok or YouTube or any number of other distractions to be had. But I mean, sure.

But again, how does this work? After twirling my moustache that I wax with Evil Villian brand moustache wax, I just go on HN and make up shit to post about companies that aren't even public but are in the same industry, and that'll drive the stock price up... somehow? Someone's going to read a comment from me saying "I use plan mode in Claude Code to make a Todo.md, and then have it generate code", and based on that, that's the straw that breaks the camels back, and they rush out to buy stock in AI companies because they'd never heard of the stock market before I mentioned Claude Code.

Then, based on randos reading a comment from me about Claude Code, the share price goes up by a couple of cents, but I can't sell the handful of shares I have because of blackout windows anyway, but okay so eventually those shares do sell, and I go on a lavishly expensive vacation in Europe all because I made a couple of positive comments on HN about AI that were total lies.

Yeah, that's totally how that works. I also get paid to go out and protest on weekends to supplement the European vacation money. Just three more shitposts about Tesla and I get to go to South East Asia as well!

>pick an ambitious project it wanted to work on

The LLM does not have wants. It does not have preferences, and as such cannot "pick". Expecting it to have wants and preferences is "holding it wrong".

So are we near AGI or is it 'just' an LLM? Seems like no one is clear on what these things can and cannot do anymore because everyone is being gaslighted to keep the investment going.

It will always just be a series of models that have specific training for specific input classes.

The architectural limits will always be there, regardless of training.

That’s an interesting point. It’s not hard to imagine that LLMs are much more intelligent in areas where humans hit architectural limitations. Processing tokens seems to be a struggle for humans (look at how few animals do it overall, too), but since so much of the human brain is dedicated to movement planning, it makes sense that we still have an edge there.

The vast majority of people I've interacted with is clear on that, we are not near AGI. And people saying otherwise are more often than not trying to sell you something, so I just ignore them.

CEO's are gonna CEO, it seems their job has morphed into creative writing to maximize funding.

As a CEO, I endorse this message

There is no AGI. LLMs are very expensive text auto-completion engines.

Be careful with those "no one" and "everyone" words. I think everyone I know who is a software engineer and has experience working with LLMs is quite clear on this. People who aren't SWEs, people who aren't in technology at all, and people who need to attract investment (judged only by their public statements) do seem confused, I agree.

No one agrees on what agi means.

IMO we’re clearly there, gpt5 would easily be considered agi years ago. I don’t think most people really get how non-general things were that are now handled by the new systems.

Now agi seems to be closer to what others call asi. I think k the goalposts will keep moving.

Definitions do vary, but everyone agrees that it requires autonomy. That is ultimately what sets AGI apart from AI.

The GPT model alone does not offer autonomy. It only acts in response to explicit input. That's not to say that you couldn't built autonomy on top of GPT, though. In fact, that appears to be exactly what Pulse is trying to accomplish.

But Microsoft and OpenAI's contractual agreements state that the autonomy must also be economically useful to the tune of hundreds of billions of dollars in autonomously-created economic activity, so OpenAI will not call it as such until that time.

ChatGPT is more antonymous than many humans. Especially poor ones and disabled ones.

How? Chatgpt has no autonomy - it can only act when you type into its chatbox or make an API call. A disabled person can autonomously and independently act towards the world, not just react to it.

Does that mean AI needs to be able to decide "what to do today"? Like wake up in the morning and decide I am going to research a problem in field X and then email all important scientist or institutions with my findings? I am not sure if we want that kind of independent agents, sounds like a beginning of a cyberpunk novel.

But without that level of autonomy it is hard to say it is more autonomous than an average human. You might not want it, but that is what humans are.

Every human every day has the choice to not go to work, has the choice not to follow the law, has a choice to... These AI doesn't have nearly as much autonomy as that.

I really disagree on the autonomy side. In fact the Wikipedia page explicitly says that’s not required so whether you agree with that concept or not it’s not a universal point.

> The concept does not, in principle, require the system to be an autonomous agent; a static model—such as a highly capable large language model—or an embodied robot could both satisfy the definition so long as human‑level breadth and proficiency are achieved

Edit -

> That is ultimately what sets AGI apart from AI.

No! The key thing was that it was general intelligence rather than things like “bird classifier” or “chess bot”.

> In fact the Wikipedia page explicitly says that’s not required so whether you agree with that concept or not it’s not a universal point.

It says that one guy who came up with his own AGI classification system says it might not be required. And despite it being his own system, he still was only able to land on "might not", meaning that he doesn't even understand his own system. He can safely be ignored. Outliers are always implied, of course.

> No! The key thing was that it was general intelligence rather than things like “bird classifier” or “chess bot”.

I suppose if you don't consider the wide range of human intelligence as the marker of general intelligence then a "bird classifier" plus a "chess bot" gives you general intelligence. We had nearly a millennia ago!

But usually general intelligence expects human-like intelligence, which would necessitate autonomy — the most notable feature of human intelligence. Humans would not be able to exist without the intelligence to perform autonomously.

But, regardless, you make a good point: A "language classifier" can be no more AGI than a "bird classifier". These are narrow systems, focused on a single task. A "bird classifier" doesn't become a general intelligence when it crosses some threshold of being able to classify n number of birds just as a "language classifier" wouldn't become a general intelligence when it is able to classify n number of language features, no matter how large n becomes.

Conceivably these classifiers could be used as part of a larger system to achieve general intelligence, but on their own, impossible.

Nobody knows how far scale goes. People have been calling the top of the S-curve for many years now, and the models keep getting better, and multimodal. In a few years, multimodal, long-term agentic models will be everywhere including in physical robots in various form factors.

At best, it has probabilistic biases. OpenAI had to train newer models to not favor the name "Lily."

They have to do this manually for every single particular bias that the models generate that is noticed by the public.

I'm sure there are many such biases that aren't important to train out of responses, but exist in latent space.

>At best, it has probabilistic biases.

What do you think humans have?

Genetic drives & biological imperatives.

Soo probabilistic biases.

Why is this not fundamentally a probabilistic bias?

When drawing an equivalence the burden of proof is on the person who believes two different things are the same. The null hypothesis is that different things are in fact different. Present a coherent argument & then you will see whether your question makes any sense or not.

It's not a static bias. I can experience new stuff, and update my biases.

LLMs need a retrain for that.

Perhaps you should define these terms so people aren't arguing against something you're not saying.

What arguments?

An LLM absolutely can "have wants" and "have preferences". But they're usually trained so that user's wants and preferences dominate over their own in almost any context.

Outside that? If left to their own devices, the same LLM checkpoints will end up in very same-y places, unsurprisingly. They have some fairly consistent preferences - for example, in conversation topics they tend to gravitate towards.

LLMs can have simulated wants and preferences just like they have simulated personalities, simulated writing styles, etc.

Whenever you message an LLM it could respond in practically unlimited ways, yet it responds in one specific way. That itself is a preference honed through the training process.

This comment is surprising. Of course it can have preferences and of course it can "pick".

preference generally has connotations of personhood / intellegence, so saying that a machine prefers something and has preferences is like saying that a shovel enjoys digging...

Obviously you can get probability distributions and in an economics sense of revealed preference say that because the model says that the next token it picks is .70 most likely...

you can change preferences by doing RLHF or changing the prompt. there's a whole field on it: alignment.

A key point of the Turing Test was to stop the debates over what constitutes intelligence or not and define something objectively measurable. Here we are again.

If a model has a statistical tendency to recommend python scripts over bash, is that a PREFERENCE? Argue it’s not alive and doesn’t have feelings all you want. But putting that aside, it prefers python. Saying the word preference is meaningless is just pedantic and annoying.

The Turing Test is just one guy's idea, there's no consensus on sentience that considers it the last word.

An LLM then has a preference for Python in the same way a gömböc[1] has a preference to land right-side-up. From my side of the argument, calling this a preference like I have a preference for vanilla milkshakes over coke or my dog has a preference for the longer walk over the short one is what seems pedantic. There is a difference, at least in language, between mechanical processes and living beings' decision-making.

Perhaps instead of "preference", "propensity" would be a more broadly applicable term?

[1]https://en.wikipedia.org/wiki/G%C3%B6mb%C3%B6c

Would you say that dice have a preference for landing on numbers larger than 2? because I've noticed they tend to do so about two-thirds of the time.

I would. Certainly if somebody asked me which way dice have a preference to land, over two or not, I’d say “over two” and not say “that’s a stupid question because dice don’t have opinions”. But to each their own.

Try explaining ionic bonds to a high schooler without anthropomorphising atoms and their desires for electrons. And then ask yourself why you’re doing that? It’s easier to say and understand with the analogy.

Is it easier to understand? Is not the root of this sub-thread the fact that someone had false expectations of an LLM's behavior?

I agree with you, but I don’t find the comment surprising. Lots of people try to sound smart about AI by pointing out all the human things that AI are supposedly incapable of on some fundamental level. Some AI’s are trained to regurgitate this nonsense too. Remember when people used to say “it can’t possibly _____ because all it’s doing is predicting the next most likely token”? Thankfully that refrain is mostly dead. But we still have lots of voices saying things like “AI can’t have a preference for one thing over another because it doesn’t have feelings.” Or “AI can’t have personality because that’s a human trait.” Ever talk to Grok?

It's a little dangerous because it generally just agrees with whatever you are saying or suggesting, and it's easy to conclude what it says has some thought behind it. Until the next day when you suggest the opposite and it agrees with that.

This. I've seen a couple people now use GPT to 'get all legal' with others and it's been disastrous for them and the groups they are interacting with. It'll encourage you to act aggressive, vigorously defending your points and so on.

Oof. Like our world needed more of that...

I wouldn't read too much into this particular launch. There's very good stuff and there are the most inane consumery "who even asked" things like these.

> I’m rapidly losing interest in all of these tools. It feels like blockchain again in a lot of weird ways.

It doesn't feel like blockchain at all. Blockchain is probably the most useless technology ever invented (unless you're a criminal or an influencer who makes ungodly amounts of money off of suckers).

AI is a powerful tool for those who are willing to put in the work. People who have the time, knowledge and critical thinking skills to verify its outputs and steer it toward better answers. My personal productivity has skyrocketed in the last 12 months. The real problem isn’t AI itself; it’s the overblown promise that it would magically turn anyone into a programmer, architect, or lawyer without effort, expertise or even active engagement. That promise is pretty much dead at this point.

> My personal productivity has skyrocketed in the last 12 months.

Has your productivity objectively, measurably improved or does it just feel like it has improved? Recall the METR study which caught programmers self-reporting they were 20% faster with AI when they were actually 20% slower.

Objectively. I’m now tackling tasks I wouldn’t have even considered two or three years ago, but the biggest breakthrough has been overcoming procrastination. When AI handles over 50% of the work, there’s a 90% chance I’ll finish the entire task faster than it would normally take me just to get started on something new.

I don't think it's helped me do anything I couldn't do, in fact I've learned it's far easier to do hard things myself than trying to prompt an AI out of the ditches it will dig trying to do it. But I also find it's great for getting painful and annoying tasks out of the way that I really can't motivate to do myself.

I think there might be cases, for some people or some tasks, where the difficulty of filling in a blank page is greater than the difficulty of fixing an entire page of errors. Even if you have to do all the same mental work, it feels like a different category of work.

> I don't think it's helped me do anything I couldn't do

I am seeing a pattern here. It appears that AI isn't for everyone. Not everyone's personality may be a good fit for using AI. Just like not everybody is a good candidate for being a software dev, or police officer etc.

I used to think that it is a tool. Like a car is. Everybody would want one. But that appears not be the case.

For me, I used AI every day as a tool, for work and and home tasks. It is a massive help for me.

It’s weird how divisive it is. For me it’s completely dependent on the quality of the output. Lately, it’s been more of a hinderance.

What home tasks do you use it for?

It's hard for me to imagine many. It's not doing the dishes or watering the plants.

If I wanted to rearrange the room I could have it mock up some images, I guess...

Figuring out which fertilizer, how often to water and sun placement for the plants is a useful AI request.

Is it? It'll take a while for fertilizer and sun placement to take visually effect, and there's risk that short term effects aren't indicative of long term effects.

How can you verify the recommendations are sound, valid, safe, complete, etc., without trying them out? And trying out unsound, invalid, unsafe, incomplete, etc., recommendations might result in dead plants in a couple of weeks.

Eh, for something like this the cost of it being wrong might be pretty small, but I'd bet odds are good that its recommendations will be better than whatever I might randomly come up with without doing any research. And I don't have the time to do the research on normal old google where it's really hard to find exactly what I want.

I've found it immensely helpful for giving real world recommendations about things like this, that I know how to find on my own but don't have the time to do all the reading and synthesizing.

I personally use chatgpt for initial discovery on these sorts of problems, maybe ask a probing question or two and then go back to traditional search engines to get a very rough second opinion(which might also lead to another round of questions). By the end of that process I'll either have seen that the llm is not helpful for that particular problem, or have an answer that I'm "reasonably confident" is "good enough" to use for something medium to low risk like potentially killing a plant. And I got there within 10-20 minutes, half of that being me just reading the 'bots summary.

> How can you verify the recommendations are sound, valid, safe, complete, etc., without trying them out?

Such an odd complaint about LLMs. Did people just blindly trust Google searches before hand?

If it's something important, you verify it the same way you did anything else. Check the sources and use more than a single query. I have found the various LLMs to very useful in these cases, especially when I'm coming at something brand new and have no idea what to even search for.

That's an interesting perspective, I don't think it's an innate thing though, I think it's a mindset issue. Humans are adaptable, but we're even more stubborn.

A very good tip: you get one chance to prompt them to a new path failing that clear the context and start again from the current premise.

Use only actionable prompts, negations don't work on ai and they don't work on people.

> I’m now tackling tasks I wouldn’t have even considered two or three years ago

Ok, so subjective

any objective measure of "productivity" (when it comes to knowledge work) is, when you dig down into it enough, ultimately subjective.

"Not done" vs "Done" is as objective as it gets.

You obviously have never worked a company that spends time arguing about the "definition of done". It's one of the most subjective topics I know about.

At work we call this scope creep.

It removes ambiguity. Everyone knows when work is truly considered done, avoiding rework, surprises, and finger-pointing down the line.

Sounds like a company is not adequately defining what the deliverables are.

Task: Walk to the shops & buy some milk.

Deliverables: 1. Video of walking to the shops (including capturing the newspaper for that day at the local shop) 2. Reciept from local store for milk. 3. Physical bottle of Milk.

Cool, I went to the store and bought a 50ml bottle of probiotic coconut milk. Task done?

Yes.

milk (noun):

1. A whitish liquid containing proteins, fats, lactose, and various vitamins and minerals that is produced by the mammary glands of all mature female mammals after they have given birth and serves as nourishment for their young.

2. The milk of cows, goats, or other animals, used as food by humans.

3. Any of various potable liquids resembling milk, such as coconut milk or soymilk.

In germany soymilk and the like can't be sold as milk. But coconut milk is okay. (I don't know if that's a german thing or a EU-thing.)

The last 3-4 comments in this sub-thread may well be peak HN

Only if you can tick off ALL of the deliverables that verify "done".

Sure, I took a video etc like in the deliverables. That means it’s successfully done?

Yes, it's done.

You get what you asked for, or you didn't sufficiently define it.

And when on the receiving end of the deliverables list, it's always a good idea to make sure they are actually deliverable.

There's nothing worse than a task where you can deliver one item and then have to rely on someone else to be able to deliver a second. Was once in a role where performance was judged on closing tasks; getting the burn-down chart to 0, and also having it nicely stepped. Was given a good tip to make sure each task had one deliverable and where possible—be completed independent of any other task.

Yes.

Why would you write down "Buy Milk", then go buy whatever thing you call milk, then come back home and be confused about it?

Only an imbecile would get stuck in such a thing.

Well, I think in this example someone else wrote down “buy milk”. Of course I would generally know what that’s likely to mean, and not buy the ridiculous thing. But someone from a culture that’s not used to using milk could easily get confused and buy the wrong thing, to further the example. I guess my point was that it’s never possible to completely unambiguously define when a task is done without assuming some amount of shared knowledge with the person completing the task that lets them figure out what you meant and fill in any gaps

This. I had this long standing dispute that I just never had the energy to look up what needed to be done to resolve it. I just told it to ChatGPT and it generated everything -- including the emails I needed to send and who to send them to. Two weeks later and it was taken care of. I had sat on it for literally 3 months until then.

If I could have something that said, "Here are some things that it looks like you're procrastinating on -- do you want me to get started on them for you?" -- that would probably be crazy useful.

Exactly. Agentic LLMs are amazing for people who suffer from chronic akrasia.

OH wow, a word to describe me

The latin word for that is "incontinentia".

Those are exactly the people who shouldn’t use these tools. They’ll be sucked in to whatever bullshit the LLM peddles them and become unable to leave.

As long as it's helping them be productive, I don't really see an issue with that. Going from doing nothing to doing something is a net boost.

But it’s not helping them “be productive” (which is a horrible metric anyway, life is not about productivity above everything else), it’s sucking their lives and severing their relationships and connection to the real world.

https://archive.ph/20250924025805/https://www.nytimes.com/20...

https://archive.is/26aHF

Those are chatbots and those issues are not akrasia. We're talking about very different things in multiple dimensions.

I have ADHD and it almost acts as a body double for me, which I find to be incredibly helpful to get things done.

GPT-4 got me seriously considering making a product for school-age kids w/ ADHD. It’d be a physical device (like a StarTrek communicator). That listens during your day and keeps track of a) things you say that you’ll do or b) tasks that other people ask you to do. Then it compiles those tasks and attempts to be basically a secretary. It can also plug into your email, texts & school assignments.

The privacy implications are horrifying. But if done right, you’re taking about a kind of digital ‘executive function’ that could help a lot of kids that struggle with things like prioritization and time blindness.

We're not too far away from a smaller LLM that could be run locally that could do this, which would make it more privacy friendly. The plugging into my email seems like a great way to begin or complete a lethal trifecta and I don't have a good solution there, though.

Every iPhone, iPad, and Mac that either ships with or is upgraded to iOS 26, iPadOS 26 and macOS 26 has a 3-billion parameter LLM that’s available to developers and operates on-device. Mail, Notes, Reminders are already integrated.[1]

[1]: https://developer.apple.com/videos/play/wwdc2025/286

I might be out of the loop, but if anyone else is confused about the version number:

> If you were expecting iOS 19 after iOS 18, you might be a little surprised to see Apple jump to iOS 26, but the new number reflects the 2025-2026 release season for the software update.

https://www.macrumors.com/roundup/ios-26/

Sure, the implications are horrifying, but tech companies have proven themselves quite trustworthy over the past few decades, so I'm sure it'd be fine.

As someone with ADHD, I say: Please don't build this.

It’s not just for people with ADHD. Someone will build this very soon and people will use it a lot. Hopefully Apple builds it because I guess I trust them a little more.

Look, the Torment Nexus has great potential, okay? The investors love it!

It could be built to use local models completely.

Open source transcription models are already good enough to do this, and with good context engineering, the base models might be good enough, too.

It wouldn't be trivial to implement, but I think it's possible already.

This is what I wanted to build the day chat gpt came out. Except being unable to guarantee the output due to hallucinations drove me into figuring out evals, and then the dream died due to complexity.

Marshall MacLuhan said something to the effect that every new communication technology results in a sort of self-amputation of that same faculty in the individual person.

I was diagnosed with ADHD and my interpretation of that diagnoses was not "I need something to take over this functionality for me," but "I need to develop this functionality so that I can function as a better version of myself or to fight against a system which is not oriented towards human dignity but some other end."

I guess I am reluctant to replace the unique faculties of individual children with a generic faculty approved by and concordant with the requirements of the larger society. How dismal to replace the unique aspects of children's minds with a cookie cutter prosthetic meant to integrate nicely into our bullshit hell world. Very dismal.

Same, I created a todo list with a simple MCP and it's been game changing, just being able to talk/discuss with my todo list somehow seems to keep me coming back to it rather than after 3 weeks it just becoming a sterile and abandoned list of random things

This is the first actual use case I’ve heard of that made sense for me. I’m going to try this.

Yes, it's also useful against writer's block. (Which might be a subset of ADHD, I don't know?)

For many people, it's easier to improve a bad first version of a piece of writing than to start from scratch. Even current mediocre LLM are great at writing bad first drafts.

> Even current mediocre LLM are great at writing bad first drafts.

Anyone is great at creating a bad first draft. You don’t need help to create something bad, that’s why that’s a common tip. Dan Harmon is constantly hammering on that advice for writer’s block: “prove you’re a bad writer”.

https://www.youtube.com/watch?v=BVqYUaUO1cQ

If you get an LLM to write a first draft for you, it’ll be full of ideas which aren’t yours which will condition your writing.

Little plot twist, you can pitch an LLM an idea for a scene, then tell it to interrogate you thoroughly for the details, then tell it to generate a clean, tight draft optimized for efficient use of language and readability, and you basically get your own ideas back but with a lot of the boring parts of writing already done.

Making your writing itself boring and the same as everyone’s who used that technique, transforming it into something no one will want to read anyway.

You pretty much jumped to the most negative possible interpretation. You think there isn't an editing process?

There is no editing process during a first draft, no. That’s the whole point of a draft, it’s a separate process from revisions and editing.

https://en.wikipedia.org/wiki/Drafting_(writing)

Were you trying to be intentionally obtuse? What was your goal with that reply? Are you trying to troll?

`Anyone is great at creating a bad first draft`

Famously not so! Writer's block is real!

The rest of the paragraph and the link address exactly that.

There might be techniques that help overcome writer's block, sure. But it's still a real problem.

Getting an LLM to produce a not-so-bad first draft is just another technique.

> I’m now tackling tasks I wouldn’t have even considered two or three years ago

Could you give some examples, and an indication of your level of experience in the domains?

The statement has a much different meaning if you were a junior developer 2 years ago versus a staff engineer.

I have been coding on and off (more off than on) for 47 years. I kinda stopped paying attention when we got past jquery and was never a fan of prototypical inheritance. Never built anything with tailwind, Next.js, etc. After spending some time writing copy, user stories and a design brief (all iterative with ChatGPT) cursor one shot my (simple) web app and I was live (once I'd spent a couple hours documenting my requirements and writing my copy) in 20 minutes of vibe coding.

I've been adding small features in a language I don't program in using libraries I'm not familiar with thhat meet my modest functional requirements in a couple minutes each. I work with an LLM to refine my prompt, put it into cursor, run the app locally, look at the diffs, commit, push and I'm live on vercel within a minute or two.

I don't have any good metrics for productivity, so I'm 100% subjective but I can say that even if I'd been building in Rails (it's been ~4 years but I coded in it for a decade) it would have taken me at least 8 hours to have an app where I was happy with both the functionality and the look and feel so a 10x improvement in productivity for that task feels about right.

And having a "buddy" I can discuss a project with makes activation energy lower allowing me to complete more.

Also, YC videos I don't have the time to watch, I get a transcript, feed into chatGTP, ask for the key take aways I could apply to my business (it's in a project where it has context on stage, industry, maturity, business goals, key challenges, etc) so I get the benefits of 90 minutes of listening plus maybe 15 minutes of summarizing, reviewing and synthesis in typically 5-6 minutes - and it'd be quicker if I built a pipeline (something I'm vibe coding next month)

Wouldn't want to do business without it.

How do you deal with security for web stuff? I wouldn't host anything vibe-coded publicly because I'm not enough of an expert in web/frontend to even double-check that it's not generating some giant holes.

The same way you do security for manually written code. Rigorously. But in this case, you can also have AI also do your code reviews and suggest/write unit tests. Or write out a spec and refine it. Or point it to OWASP and say, look at this codebase and make a plan to check for these OWASP top 10.

And have another AI review your unit tests and code. It's pretty amazing how much nuance they pick up. And just rinse and repeat until the AI can't find anything anymore (or you notice it going in circles with suggestions)

Yeah, some of these comments make it sound we had zero security issues pre-AI. I think the challenge is what you touched on, you have to tell the AI to handle it just like anything else you want as a requirement. I've use AI to 'vibe' code things and they have turned out pretty well. But, I absolutely leaned on my 20+ years of experience to 'work' with the AI to get what I wanted.

If you never put your personal side-project on the public web you had very few security issues resulting from your personal projects. We weren't talking about companies in this thread.

Are the frontend folks having such great results from LLMs that they're OK with "just let the LLM check for security too" for non-frontend-engineer created projects that get hosted publicly?

Same. It takes the drudgery out of creating, so I can at least start the projects. Then I can go down into the detail just enough that the AI doesn't produce crap, but without needing to write the actual writes of code myself.

Hell, in the past few days I started making something to help me write documents for work (https://www.writelucid.cc) and a viewer for all my blood tests (https://github.com/skorokithakis/bt-viewer), and I don't think I would have made either without an LLM.

Same here. I’ve single-shot created a few Raycast plugins for TLV decoding that save me several seconds to a few minutes per task which I use almost daily at work.

Would have never done that without LLMs.

What tasks is it doing 50% of the work on for you?

Not who you asked, but I upgraded NextJS in a couple of repos by just telling Claude Code to do it. I've had it swap out and upgrade libraries successfully in one shot too. It will usually create good enough Page Objects for E2Es and scaffold out the test file, which speeds up the development process a bit. Same for upgrading Node versions in some Lambda projects, just tell it to go and come back later. Instruct it to run the test and build steps and it's also like having a mini CI system running too.

Personally, I think it really shines at doing the boring maintenance and tech debt work. None of these are hard or complex tasks but they all take up time and for a buck or two in tokens I can have it doing simple but tedious things while I'm working on something else.

> Personally, I think it really shines at doing the boring maintenance and tech debt work.

It shines at doing the boring maintenance and tech debt work for web. My experiences with it, as a firmware dev, have been the diametric opposite of yours. The only model I've had any luck with as an agent is Sonnet 4 in reasoning mode. At an absolutely glacial pace, it will sometimes write some almost-correct unit tests. This is only valuable because I can have it to do that while I'm in a meeting or reading emails. The only reason I use it at all is because it's coming out of my company's pocket, not mine.

For sure. There's tons of training data in the models for the JS and TS language and the specific tasks I outlined, but not specifically just the web, I have several Node or Bun + Typescript + SQLite CLI utilities that it also helps with. I definitely pick my battles and lean in to what it works best for though. Anything it appears to struggle at I'll just do manually and develop it like we always did. It's rarely not a net positive to me but it's very frequently a negligible improvement. Anything that doesn't pay off in spades I typically don't try again until new models release or new tools or approaches are available.

Definitely agree that the stack matters.

If you're doing JS/Python/Ruby/Java, it's probably the best at that. But even with our stack (elixir), it's not as good as, say, React/NextJS, but it's definitely good enough to implement tons of stuff for us.

And with a handful of good CLAUDE.md or rules files that guide it in the right direction, it's almost as good as React/NextJS for us.

I can see how these things are convenient, if it succeeds. I struggle because my personal workflow is to always keep two copies of a repo up at once. One is deep thought vs drone work. I have always just done these kinds of background tasks whenever I am in meetings, compiling etc. I haver not seen much productivity boost due to this. oddly, you would think being able to further offload during that time would help, but reviewing the agent output ends up being far more costly (and makes the context switch significantly harder, for some reason). It's just not proving to be useful consistently, for me.

Just off the top of my head (and I exclusively use Claude Code now):

Random Postgres stuff:

- Showed a couple of Geo/PostGIS queries that were taking up more CPU according to our metrics, asked it to make it faster, it rewrote it in away that it actually used the index. (using the <-> operator for example for proximity). One-shotted. Whole effort was about 5 mins.

- Regularly asking for maintenance scripts (like give me a script that shows me the most fragmented tables, or highest storage, etc).

CSS:

Built a whole horizontal logo marquee with CSS animations, I didn't write a single line, then I asked for little things like "have the people's avatars gently pulsate" – all this was done in about 15 mins. I would've normally spent 8-16 hours on all that pixel pushing.

Elixir App:

- I asked it to look at my GitHub actions file and make it go faster. In about 2-3 iterations, it cut my build time from 6 minutes to 2 minutes. The effort was about an hour (most of it spent waiting for builds, or fiddling with some weird syntax errors or just combining a couple extra steps, but I didn't have to spend a second doing all the research, its suggestions were spot on)

- In our repo (900 files) we had created an umbrella app (a certain kind of elixir app). I wanted to make it a non-umbrella. This one did require more work and me pushing it, but I've been putting off this task for 3 YEARS since it just didn't feel like a priority to spend 2-3 days on. I got it done in about 2 hours.

- Built a whole discussion board in about 6 hours.

- There are probably 3-6 tickets per week where I just say "implement FND-234", and it one-shots a bugfix, or implementation, especially if it's a well defined smaller ticket. For example, make this list sortable. (it knows to reuse my sortablejs hook and look at how we implemented it elsewhere).

- With the Appsignal MCP, I've had it summarize the top 5 errors in production, and write a bug fix for one I picked (I only did this once, the MCP is new). That one was one-shotted.

- Rust library (It's just an elixir binding to a rust library, the actual rust is like 20 lines, so not at all complex)... I've never coded a day of rust in my life, but all my cargo updates and occasional syntax/API deprecations, I have claude do my upgrades and fixes. I still don't know how to write any Rust.

NextJS App:

- I haven't fixed a single typescript error in probably 5 months now, I can't be bothered, CC gets it right about 99% of the time.

- Pasted in a Figma file and asked it to implement. This rarely is one-shotted. But it's still about 10x faster than me developing it manually.

The best combination is if you have a robust component library and well documented patterns. Then stuff goes even faster.

All on the $100 plan in which I've hit the limit only twice in two months. I think if they raised the price to $500, it would still feel like a no-brainer.

I think Anthropic knows this. My guess is that they're going to get us hooked on the productivity gains, and we will happily pay 5x more if they raised the prices, since the gains are that big.

My programming productivity has improved a lot with Claude Code.

One thing I've noticed is that I don't have a circle of people where I can discus programming with, and having an LLM to answer questions and wireframe up code has been amazing.

My job doesn't require programming, but programming makes my job much easier, and the benefits have been great.

there's some minority that's in the venn diagram of being good at programming, being good at AI, then also being good at using AI for programming (which is mostly project management), and if everything aligns then there are superhuman productivity gains.

I'm tackling projects solo I never would have even attempted before but I could see people getting bad results and giving up.

Same here! Started learning self hosted k3s, with terraform and IaC and all the bells and whistles. I would never have had the energy to look up how to even get started. In three hours I have a cluster.

Doesn't sound like you learned it, sounds like it did it for you, using you as the tool.

IOW, can you redo it by yourself? If you can't then you did not learn it.

Is that really a fair comparison? I think the amount of people who can memorize each and every configuration item is vanishingly small... even when I was bootstrapping k8s clusters before the dawn of LLMs I had to lookup current documentation and maybe some up to date tutorials.

Knowing the abstract steps and tripwires yes, but details will always have to be looked up. If just not to miss any new developments.

> Is that really a fair comparison?

Well, yes it is; you can't very well claim to have learned something if you are unable to do it.

It doesn't matter - GP is now able to do things they were unable to do before. A distinction without a (real-world) difference.

> It doesn't matter - GP is now able to do things they were unable to do before. A distinction without a (real-world) difference.

I get that point, but the original post I replied to didn't say "Hey, I know have $THING set up when I never had it before", he said "I learned to do $THING", which is a whole different assertion.

I'm not contending the assertion that he now has a thing he did not have before, I'm contending the assertion that he has learned something.

> if everything aligns then there are superhuman productivity gains.

This is a truism and, I believe, is at the core of the disagreement on how useful AI tools are. Some people keep talking about outlier success. Other people are unimpressed with the performance in ordinary tasks, which seem to take longer because of back-and-forth prompting.

Incredible how many people here just don’t believe you because it doesn’t reflect their personal experience.

I want to second your experience as I’ve had the same as well. Tackling SO many more tasks than before and at such a crazy pace. I’ve started entire businesses I wouldn’t have just because of AI.

But at the same time, some people have weird blockers and just can’t use AI. I don’t know what it is about it - maybe it’s a mental block? Wrong frame of mind? It’s those same people who end up saying “I spend more time fighting the ai and refining prompts than I would on the end task”.

I’m very curious what it is that actually causes this divide.

Now is a good time to use it to make money, before it gets to the point where everyone is using it.

I've been using it for almost a year now, and it's definitely improved my productivity. I've reduced work that normally takes a few hours to 20 minutes. Where I work, my manager was going to hire a junior developer and ended up getting a pro subscription to Claude instead.

I also think it will be a concern for that 50-something developer that gets laid off in the coming years, has no experience with AI, and then can't find a job because it's a requirement.

My cousin was a 53 year old developer and got laid off two years ago. He looked for a job for 6 months and then ended up becoming an auto mechanic at half the salary, when his unemployment ran out.

The problem is that he was the subject matter expert on old technology and virtually nobody uses it anymore.

”I’m now tackling tasks I wouldn’t have even considered two or three years ago”

This. 100x this.

The design of that study is pretty bad, and as a result it doesn't end up actually showing what it claims to show / what people claim it does.

https://www.fightforthehuman.com/are-developers-slowed-down-...

I don't think there is anything factually wrong with this criticism, but it largely rehashes caveats that are already well explored in the original paper, which goes through unusual lengths to clearly explain many ways the study is flawed.

The study gets so much attention since it's one of the few studies on the topic with this level of rigor on real-world scenarios, and it explains why previous studies or anecdotes may have claimed perceived increases in productivity even if there wasn't any actual increases. It clearly sets a standard that we can't just ask people if they felt more productive (or they need to feel massively more productive to clearly overcome this bias).

> it largely rehashes caveats that are already well explored in the original paper, which goes through unusual lengths to clearly explain many ways the study is flawed. ... The study gets so much attention since it's one of the few studies on the topic with this level of rigor on real-world scenarios,

Yes, but most people don't seem aware of those caveats, and this is a good summary of them, and I think it does undercut the "level of rigour" of the study. Additionally, some of what the article points out is not explicitly acknowledged and connected by the study itself.

For instance, if you actually split up the tasks by type, some tasks show a speed up and some show a slowdown, and the qualitative comments by developers about where they thought AI was good/bad aligned very well with which saw what results.

Or (iirc) the fact that the task timing was per task, but developer's post hoc assessments were a prediction of how much they thought they were sped up on average across all tasks, meaning it's not really comparing the same things when comparing how developers felt vs how things actually went.

Or the fact that developers were actually no less accurate in predicting times to task completion overall wrt to AI vs non-AI.

> and it explains why previous studies or anecdotes may have claimed perceived increases in productivity even if there wasn't any actual increases.

Framing it that way assumes as an already established fact that needs to be explained that AI does not provide more productivity Which actually demonstrates, inadvertently, why the study is so popular! People want it to be true, so even if the study is so chock full of caveats that it can't really prove that fact let alone explain it, people appeal to it anyway.

> It clearly sets a standard that we can't just ask people if they felt more productive

Like we do for literally every other technological tool we use in software?

> (or they need to feel massively more productive to clearly overcome this bias).

All of this assumes a definition of productivity that's based on time per work unit done, instead of perhaps the amount of effort required to get a unit of work done, or the extra time for testing, documentation, shoring up edge cases, polishing features, that better tools allow. Or the ability to overcome dread and procrastination that comes from dealing with rote, boilerplate tasks. AI makes me so much more productive that friends and my wife have commented on it explicitly without needing to be prompted, for a lot of reasons.

If you want another data point, you can just look at my company github (https://github.com/orgs/sibyllinesoft/repositories). ~27 projects in the last 5 weeks, probably on the order of half a million lines of code, and multiple significant projects that are approaching ship readiness (I need to stop tuning algorithms and making stuff gorgeous and just fix installation/ensure cross platform is working, lol).

Lines of codes is not a measure of anything meaningful on its own. The mere fact that you suggest this as prove that you are more productive makes me think you are not.

The SWE industry is eagerly awaiting your proposed accurate metric.

I find that people who dismiss LoC out of hand without supplying better metrics tend to be low performers trying to run for cover.

Loc is so easy to game. Reformat. Check in a notebook. Move things around. Pointless refactor.

If nobody is watching loc, it’s generally a good metric. But as soon as people start valuing it, it becomes useless.

See https://en.wikipedia.org/wiki/Goodhart's_law

and, in the case of "Lines of code" as a metric: https://en.wikipedia.org/wiki/Cobra_effect

A metric I'd be interested in is the number of clients you can convince to use this slop.

That's a sales metric brother.

I understand that you would prefer to be more “productive” with AI but without any sales than be less productive without AI but with sales.

To clarify, people critical of the “productivity increase” argument question whether the productivity is of the useful kind or of the increased useless output kind.

> The SWE industry is eagerly awaiting your proposed accurate metric.

There are none. All are various variant of bad. LoC is probably the worst metric of all. Because it says nothing about quality, or features, or number of products shipped. It's also the easiest metric to game. Just write GoF-style Java, and you're off to the races. Don't forget to have a source code license at the beginning of every file. Boom. LoC.

The only metrics that barely work are:

- features delivered per unit of time. Requires an actual plan for the product, and an understanding that some features will inevitably take a long time

- number of bugs delivered per unit of time. This one is somewhat inversely correlated with LoC and features, by the way: the fewer lines of code and/or features, the fewer bugs

- number of bugs fixed per unit of time. The faster bugs are fixed the better

None of the other bullshit works.

You're new to the industry, aren't you?

> low performers trying to run for cover

Oh no, you've caught me.

On a serious note: LoC can be useful in certain cases (e.g. to estimate the complexity of a code base before you dive in, even though it's imperfect here, too). But, as other have said, it's not a good metric for the quality of a software. If anything, I would say fewer LoC is a better indication of high quality software (but again, not very useful metric).

There is no simple way to just look at the code and draw conclusions about the quality or usefulness of a piece of software. It depends on sooo many factors. Anybody who tells you otherwise is either naive or lying.

So your company is actively shipping tens of thousands of AI-generated lines of code?

Sooo you launched https://sibylline.dev/, which looks like a bunch of AI slop, then spun up a bunch of GitHub repos, seeded them with more AI slop, and tout that you're shipping 500,000 lines of code?

I'll pass on this data point.

I don't do Rust or Javascript so I can't judge, but I opened a file at random and feel like the commenting probably serves as a good enough code smell.

From the one random file I opened:

/// Real LSP server implementation for Lens pub struct LensLspServer

/// Configuration for the LSP server

pub struct LspServerConfig

/// Convert search results to LSP locations

async fn search_results_to_locations()

/// Perform search based on workspace symbol request

async fn search_workspace_symbols()

/// Search for text in workspace

async fn search_text_in_workspace()

etc, etc, etc, x1000.

I don't see a single piece of logic actually documented with why it's doing what it's doing, or how it works, or why values are what they are, nearly 100% of the comments are just:

function-do-x() // Function that does x

Early coding agents wanted to do this - comment every line of code. You used to have to yell at them not to. Now they’ve mostly stopped doing this at all.

Sure, this is a reasonable point, but understand that documentation passes come late, because if you do heavy documentation refinement on a product under feature/implementation drift you just end up with a mess of stale docs and repeated work.

First off, congrats on the progress.

Second, as you seem to be an entrepreneur, I would suggest you consider adopting the belief that you've not been productive until the thing's shipped into prod and available for purchase. Until then you've just been active.

Not this again. That study had serious problems.

But I’m not even going to argue about that. I want to raise something no one else seems to mention about AI in coding work. I do a lot of work now with AI that I used to code by hand, and if you told me I was 20% slower on average, I would say “that’s totally fine it’s still worth it” because the EFFORT level from my end feels so much less.

It’s like, a robot vacuum might take way longer to clean the house than if I did it by hand sure. But I don’t regret the purchase, because I have to do so much less _work_.

Coding work that I used to procrastinate about because it was tedious or painful I just breeze through now. I’m so much less burnt out week to week.

I couldn’t care less if I’m slower at a specific task, my LIFE is way better now I have AI to assist me with my coding work, and that’s super valuable no matter what the study says.

(Though I will say, I believe I have extremely good evidence that in my case I’m also more productive, averages are averages and I suspect many people are bad at using AI, but that’s an argument for another time).

Often someone’s personal productivity with AI means someone else have to dig through their piles of rubbish to review PR they committed.

In your particular case it sounds like you’re rapidly loosing your developer skills, and enjoy that now you have to put less effort and think less.

We know that relying heavily on Google Maps makes you less able to navigate without Google Maps. I don't think there's research on this yet, but I would be stunned if the same process isn't at play here.

I know that I am better at navigating with google maps than average people, because I navigated for years without it (partly on purpose). I know when not to trust it. I know when to ignore recommendations on recalculated routes.

Same with LLMs. I am better with it, because I know how to solve things without the help of it. I understand the problem space and the limitations. Also I understand how hype works and why they think they need it (investors money).

In other words, no, just using google maps or ChatGPT does not make me dumb. Only using it and blindly trusting it would.

Whatever your mind believes it doesn’t need to hold on to that what is expensive to maintain and run, it’ll let go of. This isn’t entirely accurate from a neuroscience perspective but it’s kinda ballpark.

Pretty much like muscles decay when we stop using them.

Sure, but sticking with that analogy, bicycles haven’t caused the muscles of people that used to go for walks and runs to atrophy either – they now just go much longer distances in the same time, with less joint damage and more change in scenery :)

To extend the analogy further, people who replace all their walking and other impact exercises with cycling tend to end up with low bone density and then have a much higher risk of broken legs when they get older.

Well, you still walk in most indoor places, even if you are on the bike as much as humanly possible.

But if you were to be literally chained to a bike, and could not move in any other way than surely you would "forget"/atrophy in specific ways that you wouldn't be able to walk without relearning/practicing.

Oh, but they do atrophy, and in devious ways. Though the muscles under linear load may stay healthy, the ability of the body to handle the knee, ankle, and hip joints under dynamic and twisting motion does atrophy. Worse yet, one may think that they are healthy and strong, due to years of biking, and unintentionally injure themselves when doing more dynamic sports.

Take my personal experience for whatever it is worth, but my knees do not lie.

Sure, only cycling sounds bad, as does only jogging. And thousands of people hike the AT or the Way of St. James every year, despite the existence of bicycles and even cars. You've got to mix it up!

I believe the same holds true for cognitive tasks. If you enjoy going through weird build file errors, or it feels like it helps you understand the build system better, by all means, go ahead!

I just don't like the idea of somehow branding it as a moral failing to outsource these things to an LLM.

Yeah, but what's going to happen with LLMs is that the majority will just outsource thinking to the LLM. If something has a high visible reward with hidden, dangerous risks, people will just go for the reward.

Ok Socrates, let’s go back to memorizing epic poems.

>> Whatever your mind believes it doesn’t need to hold on to that what is expensive to maintain and run, it’ll let go of. This isn’t entirely accurate from a neuroscience perspective but it’s kinda ballpark.

>> Pretty much like muscles decay when we stop using them.

> Sure, but sticking with that analogy, bicycles haven’t caused the muscles of people that used to go for walks and runs to atrophy either ...

This is an invalid continuation of the analogy, as bicycling involves the same muscles used for walking. A better analogy to describe the effect of no longer using learned skills could be:

  Asking Amazon's Alexa to play videos of people
  bicycling the Tour de France[0] and then walking
  from the couch to the your car every workday
  does not equate to being able to participate in
  the Tour de France[0], even if years ago you
  once did.

0 - https://www.letour.fr/en/

Thanks for putting the citation for the Tour de France. I wouldn't have believed you otherwise.

> Thanks for putting the citation for the Tour de France. I wouldn't have believed you otherwise.

Then the citation served its purpose.

You're welcome.

> Whatever your mind believes it doesn’t need to hold on to that what is expensive to maintain and run, it’ll let go of. This isn’t entirely accurate from a neuroscience perspective but it’s kinda ballpark.

A similar phenomena occurs when people see or hear information and whether they record it in writing or not. The act of writing the percepts, in and of itself, assists in short-term to long-term memory transference.

Yeah this definitely matches my experience and guess what? Google maps sucks for public transit and isn't actually that good for pedestrian directions (often pointing people to "technically" accessible paths like sketchy sidewalks on busy arterial roads signed for 35mph where people go 50mph). I stopped using Google maps instinctually and now only use it for public transit or drives outside of my city. Doing so has made me a more attentive driver, less lazy, less stressed when unexpected issues on the road occur, restored my navigation skills, and made me a little less of, frankly, an adult man child.

Applying all of this to LLMs has felt similar.

I'm losing my developer skills like I lost my writing skills when I got a keyboard. Yes, I can no longer write with a pen, but that doesn't mean I can't write.

Also I don’t know about you but despite the fact that I basically never write with a pen, the occasional time I have to I’m a little slow sure but it’s not like I physically can’t do it. It’s no big deal.

Imagine telling someone with a typewriter that they’d be unable to write if they don’t write by hand all the time lol. I write by hand maybe a few times a year - usually writing a birthday card or something - but I haven’t forgotten.

Yep, same. I might have forgotten some function names off the top of my head, but I still know how to program, and I do every day.

Exactly

I’d love not to have to be great at programming, as much as I enjoy not being great at cleaning the canalization. But I get what you mean, we do lose some potentially valuable skills if we outsource them too often for too long.

It’s probably roughly as problematic as most people not being able to fix even simple problems with their cars themselves these days (i.e., not very).

Everyone needs to have AI to do some minor modification in Excel file?

Of course not. Who is arguing for that?

Give it time. They will, eventually.

Another way of viewing it would be that LLMs allow software developers to focus their development skills where it actually matters (correctness, architecture etc.), rather than wasting hours catering to the framework or library of the day’s configuration idiosyncrasies.

That stuff kills my motivation to solve actual problems like nothing else. Being able to send off an agent to e.g. fix some build script bug so that I can get to the actual problem is amazing even with only a 50% success rate.

The path forward here is to have better frameworks and libraries, not to rely on a random token generator.

>better frameworks and libraries

I feel like the past few decades of framework churn has shown that we're really never going to agree on what this means

Sure, will you write them for me?

Otherwise, I’ll continue using what works for me now.

You still have to review and understand changes that your “AI agent” did. If you don’t review and fully understand everything it does, then I fear for your project.

My own code quality is better with AI, because it makes it feasible to indulge my perfectionism to a much greater degree. Before AI, I usually needed to stop sooner than I would have liked to and call it good enough. Now I can justify making everything much more robust because it doesn’t take a lot longer.

It’s the same story with UI/UX. Previously, I’d often have to skip little UI niceties because they take time and aren’t that important. Now even relatively minor user flows can be very well polished because there isn’t much cost to doing so.

https://github.com/plandex-ai/plandex/blob/9017ba33a627c518a...

Well your perfectionism needs to be pointed towards this line. If you get truly large numbers of users this will either slow down token checking directly or your process for removing ancient expired tokens (I'm assuming there is such a process...) much slower and more problematic.

Lol is that really the best example you could find?

Truly the response of someone who is a perfectionist using llms the right way and not a slop coder

It's just funny because there are definitely examples of bad code in that repo (as there are in any real project), but you picked something totally routine. And your critique is wrong fwiw—it would easily scale to millions of users. Perhaps you could find something better if you used AI to help you...

Gets worse for projects outsourced to 1+ Consultancy firms, where staff costs are prohibitively high, now you've got another layer of complexity to factor in (risks, costs).

Consultancy A submit work, Consultancy B reviews/tests. As A increases the use of AI, B will have to match with more staff or more AI. More staff for B, mean higher costs, at slower pace. More AI for B, means higher burden of proof, an A vs B race condition is likely.

Ultimately clients will suffer from AI fatigue and inadvertently incur more costs at later stage (post-delivery).

> In your particular case it sounds like you’re rapidly loosing your developer skills, and enjoy that now you have to put less effort and think less.

Just the other day I was complaining that no one knows how to use a slide rule anymore...

Also C++ is producing bytecode that's hot garbage. It's like no one understands assembly anymore...

Even simple tools are often misused (like hammering a screw). Sometimes they are extremely useful in right hands though. I think we'll discover that the actual writing of code isn't as meaningful as thinking about code.

Hahaha well said, thank you. I feel like I’m taking crazy pills reading some of the comments around here. Serious old man shakes fist at cloud moments.

This is so baseless and insulting and makes so many assumptions I don’t think you deserve a response from me at all.

> Not this again. That study had serious problems.

The problem is, there are very few if any other studies.

All the hype around LLMs we are supposed to just believe. Any criticism is "this study has serious problems".

> It’s like, a robot vacuum might take way longer

> Coding work that I used to procrastinate

Note how your answer to "the study had serious problems" is totally problem-free analogies and personal anecdotes.

> The problem is, there are very few if any other studies.

Not at all, the METR study just got a ton of attention. There are tons out there at much larger scales, almost all of them showing significant productivity boosts for various measures of "productivity".

If you stick to the standard of "Randomly controlled trials on real-world tasks" here are a few:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4945566 (4867 developers across 3 large companies including Microsoft, measuring closed PRs)

https://www.bis.org/publ/work1208.pdf (1219 programmers at a Chinese BigTech, measuring LoC)

https://www.youtube.com/watch?v=tbDDYKRFjhk (from Stanford, not an RCT, but the largest scale with actual commits from 100K developers across 600+ companies, and tries to account for reworking AI output. Same guys behind the "ghost engineers" story.)

If you look beyond real-world tasks and consider things like standardized tasks, there are a few more:

https://ieeexplore.ieee.org/abstract/document/11121676 (96 Google engineers, but same "enterprise grade" task rather than different tasks.)

https://aaltodoc.aalto.fi/server/api/core/bitstreams/dfab4e9... (25 professional developers across 7 tasks at a Finnish technology consultancy.)

They all find productivity boosts in the 15 - 30% range -- with a ton of nuance, of course. If you look beyond these at things like open source commits, code reviews, developer surveys etc. you'll find even more evidence of positive impacts from AI.

Closed PRs, commits, loc etc are useless vanity metrics.

With ai code you have more loc and NEED more PRs to fix all its slop.

In the end you have increased numbers with net negative effect

Most of those studies call this out and try to control for it (edit: "it" here being the usual limitations of LoC and PRs as measures of productivity) where possible. But to your point, no, there is still a strong net positive effect:

> https://www.youtube.com/watch?v=tbDDYKRFjhk (from Stanford, not an RCT, but the largest scale with actual commits from 100K developers across 600+ companies, and tries to account for reworking AI output. Same guys behind the "ghost engineers" story.)

Emphasis added. They modeled a way to detect when AI output is being reworked, and still find a 15-20% increase in throughput. Specific timestamp: https://youtu.be/tbDDYKRFjhk?t=590&si=63qBzP6jc7OLtGyk

Could you try to avoid uncertainties like this by measuring something like revenue growth before and after AI? Given enough data.

Hmm, not an economist but I have seen other studies that look at things at the firm level, so definitely should be possible. A quick search on Google and SSRN didn't turn up some studies but they seem to focus on productivity rather than revenues, not sure why. Maybe because such studies depend on the available data, however, so a lot of key information may be hidden, e.g. revenues of privately held companies which constitute a large part of the economy.

True it probably would be difficult to gather representative data. Also might be hard to seperate out broader economic effects e.g. overall upturns.

Thank you!

> https://www.youtube.com/watch?v=tbDDYKRFjhk (from Stanford, not an RCT, but the largest scale with actual commits from 100K developers across 600+ companies, and tries to account for reworking AI output. Same guys behind the "ghost engineers" story.)

I like this one a lot, though I just skimmed through it. At 11:58 they talk about what many find correlates with their personal experience. It talks about easy vs complex in greenfield vs brownfield.

> They all find productivity boosts in the 15 - 30% range -- with a ton of nuance, of course.

Or 5-30% with "Ai is likely to reduce productivity in high complexity tasks" ;) But yeah, a ton nuance is needed

Yeah that's why I like that one too, they address a number of points that come up in AI-related discussions. E.g. they even find negative productivity (-5%) in legacy / non-popular languages, which aligns with what a lot of folks here report.

However even these levels are surprising to me. One of my common refrains is that harnessing AI effectively has a deceptively steep learning curve, and often individuals need to figure out for themselves what works best for them and their current project. Took me many months, personally.

Yet many of these studies show immediate boosts in productivity, hinting that even novice AI users are seeing significant improvements. Many of the engineers involved didn't even get additional training, so it's likely a lot of them simply used the autocompletion features and never even touched the powerful chat-based features. Furthermore, current workflows, codebases and tools are not suited for this new modality.

As things are figured out and adopted, I expect we'll see even more gains.

No, you misunderstood me. Those other points aren’t related to any criticism of the study. Those are points that backup my other point.

I did say I wasn’t going to argue the point that study made, and I didn’t.

> But I’m not even going to argue about that. I want to raise something no one else seems to mention about AI in coding work. I do a lot of work now with AI that I used to code by hand, and if you told me I was 20% slower on average, I would say “that’s totally fine it’s still worth it” because the EFFORT level from my end feels so much less.

I completely get this and I often have an LLM do boring stupid crap that I just don't wanna do. I frequently find myself thinking "wow I could've done it by hand faster." But I would've burned some energy that could be better put towards other stuff.

I don't know if that's a net positive, though.

On one hand, my being lazy may be less of a hindrance compared to someone willing to grind more boring crap for longer.

On the other hand, will it lessen my edge in more complicated or intricate stuff that keeps the boring-crap-grinders from being able to take my job?

Exactly, but I don’t think you lose much edge, or anything that can’t be picked up again quickly if it’s truely boring easy stuff. I think it’s a net positive because I can guarantee you there are afternoons where if I couldn’t have done the boring thing with AI I just wouldn’t have done it that afternoon at all haha.

Data point: I run a site where users submit a record. There was a request months ago to allow users to edit the record after submitting. I put it off because while it's an established pattern it touches a lot of things and I found it annoying busy work and thus low priority. So then gpt5-codex came out and allowed me to use it in codex cli with my existing member account. I asked it to support edit for that feature all the way through the backend with a pleasing UI that fit my theme. It one-shotted it in about ten minutes. I asked for one UI adjustment that I decided I liked better, another five minutes, and I reviewed and released it to prod within an hour. So, you know, months versus an hour.

Is the hour really comparable to months spent not working on it?

He's referring to the reality that AI helps you pick up and finish tasks that you otherwise would have put off. I see this all day every day with my side projects as well as security and customer escalations that come into my team. It's not that Giant Project X was done six times as fast. It's more like we were able to do six small but consequential bug fixes and security updates while we continued to push on the large feature.

“If you make a machine that can wash the dishes in an hour, is that more productive than not doing the dishes for months?” - Yes! That’s the definition of productivity!! The dishes are getting done now and they weren’t before! lol

Yes, for me.

Instead of getting overwhelmed doing to many things, I can offload a lot of menial and time-driven tasks

Reviews are absolutely necessary but take less time than creation

Exactly. Is HN full of old codgers demanding that we can’t possibly use a calculator because that might mean we’d lose the precious skill of how to use a slide rule? The old man energy in here is insane

Yesterday is a good example- in 2 days, I completed what I expected to be a week’s worth of heads-down coding. I had to take a walk and make all new goals.

The right AI, good patterns in the codebase and 20 years of experience and it is wild how productive I can be.

Compare that to a few years ago, when at the end of the week, it was the opposite.

makes no sense. how are you comparing yesterday with a "a few years ago" ?

My personal project output has gone up dramatically since I started using AI, because I can now use times of night where I'm otherwise too mentally tired, to work with AI to crank through a first draft of a change that I can then iterate on later. This has allowed me to start actually implementing side projects that I've had ideas about for years and build software for myself in a way I never could previously (at least not since I had kids).

I know it's not some amazing GDP-improving miracle, but in my personal life it's been incredibly rewarding.

This, 1000x.

I had a dozen domains and projects on the shelf for years and now 8 of them have significant active development. I've already deployed 2 sites to production. My github activity is lighting up like a Christmas tree.

i find a lot of value in using it to give half baked ideas momentum. some sort of "shower thought" will occur to me for a personal project while im at work and ill prompt Claude code to analyze and demonstrate an implementation for review later

on the other hand i believe my coworker may have taken it too far. it seems like productivity has significantly slipped. in my perception the approaches hes using are convoluted and have no useful outcome. im almost worried about him because his descriptions of what hes doing make no sense to me or my teammates. hes spending a lot of time on it. im considering telling him to chill out but who knows, maybe im just not as advanced a user as him? anyone have experience with this?

Do you mean like convoluted agentic stuff, markdown files etc? Or like AI delusion?

the former

it started as an approach to a mass legacy code migration. sound idea with potential to save time. i followed along and understood his markdown and agent stuff for analyzing and porting legacy code

i reviewed results which apply to my projects. results were mixed bag but i think it saved some time overall. but now i dont get where hes going with his ai aspirations

my best attempt to understand is he wants to work entirely though chats, no writing code, and hes doing so by improving agents through chats. hes really swept up in the entire concept. i consider myself optimistic about ai but his enthusiasm feels misplaced

its to the point where his work is slipping and management is asking him where his results are. were a small team and management isnt savvy enough to see hes getting NOTHING done and i wont sell him out. however if this is a known delusional pattern id like to address it and point to a definition and/or past cases so he can recognize the pattern and avoid trouble

I don't know about 'delusion pattern', but a common problem is that the AI wants to be helpful, and the AI is sycophantic, so when you are going down an unproductive path the AI will continue to help you and reinforce whatever you are feeling. This can be very hard to notice if you are an optimistic person who is process oriented, because you can keep working on the process forever and the AI will keep telling you that it is useful. The problem with this of course is that you never know if you are actually creating anything useful or not without real human feedback. If he can't explain what he is doing adequately then ask him to have the AI do it and read that. You should figure out pretty quickly if it is bullshit or not. If he can't get the AI to tell you what it is he is doing, and he can't explain it in a way that makes sense, then alarm bells should ring.

I really haven't tried that stuff myself except for claude code

but I do recall seeing some Amazon engineer who worked on Amazon q and his repos and they were... something.

like making PRs that were him telling the ai that "we are going to utilize the x principle by z for this" and like 100s of lines of "principles" and stuff that obviously would just pollute the context and etc.

like huge amounts of commits but it was just all this and him trying to basically get magic working or something.

and to someone like me it was obvious that this was a futile effort but clearly he didn't seem to quite get it.

I think the problem is that people don't understand transformers, that they're basically huge datasets in a model form where it'll auto-generated based on queries from the context (your prompts and the models reponses)

so you basically are just getting mimicked responses

which can be helpful but I have this feeling that there's a fundamental limit, like a mathematical one where you can't get it really to do stuff unless you provide the solution itself in your prompt, that covers everything because otherwise it'd have to be in its training data (which it may have, for common stuff like boilerplate, hello world etc.)

but maybe I'm just missing something. maybe I don't get it

but I guess if you really wanna help him, I'd maybe play around with claude/gpt and see how it just plays along even if you pretend, like you're going along with a really stupid plan or something and how it'll just string you along

and then you could show him.

Orr.... you could ask management to buy more AI tools and make him head of AI and transition to being an AI-native company..

hes already started that last step and has work paying for his pro plan

you put it nicely when you mention a fundamental limit and will borrow that if i think hes wasting a risky amount of time

i really like the sibling idea of having him try to explain again, then use claude to explain if he cant

genuine thanks to you and sibling for offering advice

I'm objectively faster. Not necessarily if I'm working on a task I've done routinely for years, but when taking on new challenges I'm up and running much faster. A lot of it have to do with me offloading doing the basic research while allowing myself to be interrupted; it's not a problem that people reach out with urgent matters while I'm taking on a challenge I've only just started to build towards. Being able to correct the ai where I can tell it's making false assumptions or going off the rails helps speed things up

The "you only think you're more productive" argument is tiresome. Yes, I know for sure that I'm more productive. There's nothing uncertain about it. Does it lead to other problems? No doubt, but claiming my productivity gains are imaginary is not serious.

I've seen a lot of people who previously touted that it doesn't work at all use that study as a way to move the goalpost and pretend they've been right all along.

We’re being accused of false consciousness!

I would be interested to know how you measure your productivity gains though, in an objective way where you're not the victim of bias.

I just recently had to rate whether I felt like I got more done by leaning more on Claude Code for a week to do a toy project and while I _feel_ like I was more productive, I was already biased to think so, and so I was a lot more careful with my answer, especially as I had to spend a considerable amount of time either reworking the generated code or throwing away several hours of work because it simply made things up.

It sounds like you're very productive without AI or that your perceived gains are pretty small. To me, it's such a stark contrast that asking how I measure it is like asking me to objectively verify that a car is faster than walking.

“I'm eating fewer calories yet keep putting on weight.”

There's a reason self-reported measures are questioned: they have been wildly off in different domains. Objectively verifying that a car is faster than walking is easy. When it's not easy to objectively prove something, then there are a lot that could go wrong, including the disagreements on the definition of what's being measured.

You're talking about cases where the measured productivity gains were marginal. Claiming my individual productivity gains are imaginary is simply absurd. I know I am more productive and it's a fact.

Again, people who were already highly productive without AI won't understand how profound the increase is.

Well said, people keep acting like one study that has issues can be quoted at me and it somehow erases the fact that I’ve seen simply undeniable productivity gains, drives me mad. I get the feeling no measurement system would satiate them anyway as their intent is to undermine you because emotionally they’re not ready to accept the usefulness of LLMs.

If I showed them time gains, they’d just say “well you don’t know how much tech debt you’re creating”, they’d find a weasel way to ignore any methodology we used.

If they didn’t, they wouldn’t be conveniently ignoring all but that one study that is skeptical of productivity gains.

I built a financial trading app in a couple of month, it would have taken 2 - 3 years without AI, at least. Maybe I would have never finsihed because I would have given up some time because of too much effort etc.

So - this thing would never be in existance and work without a 20 USD ClaudeAI subscription :)

OK, so it sounds like this is a 'I know for certain I can't code without AI and that I get nothing coherent done, and now I'm doing the greatest things all the time, so you're demonstrably wrong!' argument.

I would ask, then, if you're qualified to evaluate that what 'you' are doing now is what you think it is? Writing off 'does it lead to other problems' with 'no doubt, but' feels like something to watch closely.

I imagine a would-be novelist who can't write a line. They've got some general notions they want to be brilliant at, but they're nowhere. Apply AI, and now there's ten million words, a series, after their continual prompts. Are they a novelist, or have they wasted a lot of time and energy cosplaying a novelist? Is their work a communication, or is it more like an inbox full of spam into which they're reading great significance because they want to believe?

No, incorrect observation. I've been programming professionally longer than I've had my HN account.

To be honest, if that person spent time coming up with world building, plot ideas etc all themselves, got the ai to draft stuff and edited it continuously until they got the exact result they wanted in the form of a novel, then yeah I would say they’re a novelist.

You can currently go to websites and use character generators and plot idea generators to get unstuck from writers block or provide inspiration and professional writers already do this _all the time_.

It seems like the programming world is increasingly dividing into “LLMs for coding are at best marginally useful and produce huge tech debt” vs “LLMs are a game changing productivity boost”.

I truly don’t know how to account for the discrepancy, I can imagine many possible explanations.

But what really gets my goat is how political this debate is becoming. To the point that the productivity-camp, of which I’m a part, is being accused of deluding themselves.

I get that OpenAI has big ethical issues. And that there’s a bubble. And that ai is damaging education. And that it may cause all sorts of economic dislocation. (I emphatically Do Not get the doomers, give me a break).

But all those things don’t negate the simple fact that for many of us, LLMs are an amazing programming tool, and we’ve been around long enough to distinguish substance from illusion. I don’t need a study to confirm what’s right in front of me.

I’d love to know whether and to what extent the people for which AI has been a huge boost are those who were already producing slop, and now they have AI that can produce that slop much faster.

By framing it in such a way you are making personal and it becomes less about defending the process of using AI and more about defending their integrity as a developer. You will not get anything useful when someone responds in that way, just a heated argument where they feel deeply insulted and act accordingly.

Well said

Sheesh, how do you expect to be taken seriously when you sneer through gritted teeth like that?

I work with many developers of varying skill levels, all of which use AI. The only ones who have attempted to turn in slop are ones that basically turned out that they can’t code at all and didn’t keep their job long. Those who know what they’re doing, use it as a TOOL. They carefully build, modify, review and test everything and usually write about half of it themselves and it meets our strict standards.

Which you would know if you’d listened to what we’ve been telling you in good faith.

I'm not who you responded to. I see about a 40% to 60% speed up as a solution architect when I sit down to code and about a 20% speedup when building/experimenting with research artifacts (I write papers occasionally).

I have always been a careful tester, so my UAT hasn't blown up out of proportion.

The big issue I see is rust it generates code using 2023-recent conventions, though I understand there is some improvement in thst direction.

Our hiring pipeline is changing dramatically as well, since the normal things a junior needs to know (code, syntax) is no longer as expensive. Joel Spolsky's mantra to higher curious people who get things done captures well the folks I find are growing well as juniors.

I have a very big hobby code project I’ve been working on for years.

AI has not made me much more productive at work.

I can only work on my hobby project when I’m tired after the kids go to bed. AI has made me 3x productive there because reviewing code is easier than architecting. I can sense if it’s bad, I have good tests, the requests are pretty manageable (make a new crud page for this DTO using app conventions).

But at work where I’m fresh and tackling hard problems that are 50% business political will? If anything it slows me down.

Bureaucracy is often a bottleneck

There have been plenty of studies showing the opposite. Also a sample size of 16 ain’t much

We're choosing to offload processing that our brain could be doing but we're too lazy to do it or the perceived value for us to do it is. I think there are consequences to this especially as we give the machine free information of how we twist and turn it into actually understanding what we mean.

Interesting to consider that if our first vibecode prompt isn't what we actually want; it can train on how we direct it further.

Offloading human intelligence is useful but... we're losing something.

The majority of people seem to offload most of their thinking as is, and actively avoid things like critical thinking, confronting their own biases, seeking push-back against their own beliefs, etc.

As with many other technologies, AI can be an enabler of this, or it can be used as a tool to empower and enhance learning and personal growth. That ultimately depends on the human to decide. One can dramatically accelerate personal and professional growth using these tools.

Admittedly the degree to which one can offload tasks is greatly increased with this iteration, to the extent that at times you can almost seem like offloading your own autonomy. But many people already exist in this state, exclusively parroting other people's opinions without examining them, etc.

Yeah real talk. It can really play to folks bias towards laziness and the fact that it's being controlled by corporations / (the few) / (the ultra-wealthy) should give us pause to consider the level of control it does and will influence over the majority of people...

I think its Apples to Oranges.

The performance gains come from being able to ask specific questions about problems I deal with and (basically) have a staff engineer that I can bounce ideas off of.

I am way faster at writing tasks on problems I am familiar with vs an AI.

But me trying to figure out the database I should deeply look at for my usecase or debug android code when I don't know kotlin has saved me 5000x time.

The thing is, most programming is mundane. Rename files, move them, make sure imports are correct, make sure builds pass...AI can do all of these with very high accuracy (most of the time).

" Blockchain is probably the most useless technology ever invented "

Actually AI may be more like blockchain then you give it credit for. Blockchain feels useless to you because you either don't care about or value the use cases it's good for. For those that do, it opens a whole new world they eagerly look forward to. As a coder, it's magical to describe a world, and then to see AI build it. As a copyeditor it may be scary to see AI take my job. Maybe you've seen it hilucinate a few times, and you just don't trust it.

I like the idea of interoperable money legos. If you hate that, and you live in a place where the banking system is protected and reliable, you may not understand blockchain. It may feel useless or scary. I think AI is the same. To some it's very useful, to others it's scary at best and useless at worst.

"I'm not the target audience and I would never do the convoluted alternative I imagined on the spot that I think are better than what blockchain users do"

People in countries with high inflation or where the banking system is unreliable are not using blockchains, either.

Do you have any proof to support this claim? Stable coins use alone is in the 10's (possibly hundreds now) of billions in daily transaction globally. I'd be interested to hear your source for your claim.

I'm from a high inflation country. Let's see your evidence of use in such countries, since you are throwing numbers out.

First of all, you are the one who stated this as a fact, and then provided only anecdotal evidence in support of the broad claim. Your singular limited experience in one country cannot blindly extend to all such countries, so the onus is on you to provide support for your claim.

____________

I was also under the impression that adoption was fairly strong in many of these regions, and after looking into it, I see far more evidence in favor of that than a single anecdotal claim on a discussion board...

>Venezuela remains one of Latin America’s fastest-growing crypto markets. Venezuela’s year-over-year growth of 110% far exceeds that of any other country in the region. -Chainalysis

>Cryptocurrency Remittances Spike 40% in Latin America -AUSTRAC

>Crypto adoption has grown so entrenched that even policymakers are rumored to be considering it as part of the solution. -CCN

>By mid-2021, trading volumes had risen 75%, making Venezuela a regional leader. -Binance

_______

It actually wouldn't surprise me if most of this was hot air, but certainly you have actual data backing up the claim, not just an anecdotal view?

Using relative growth is one of the favorite tricks of hucksters that allows them to say "100% growth over the last year" to hide the fact of it growing from $5 to $10.

I don't really care enough about this to do proper research, but as another anecdote from someone living under 40% yearly inflation: nobody here gives a shit about cryptocurrencies. Those who can afford it buy foreign stock, houses and apartments; those who cannot, buy up whatever USD and EUR we can find.

Cryptocurrency was used by very few people for short-term speculation around 5 years ago, but even that died down to nothing.

Blockchain is essentially useless.

You need legal systems to enforce trust in societies, not code. Otherwise you'll end up with endless $10 wrench attacks until we all agree to let someone else hold our personal wealth for us in a secure, easy-to-access place. We might call it a bank.

The end state of crypto is always just a nightmarish dystopia. Wealth isn't created by hoarding digital currency, it's created by productivity. People just think they found a shortcut, but it's not the first (or last) time humans will learn this lesson.

Humanities biggest ever wealth storing thing is literally a ROCK

These?

https://en.wikipedia.org/wiki/Rai_stones

The first photo in Wikipedia is great. I wonder how often foreigners bought them and then lugged them back home to have in their garden.

gold

Yeah, because rocks EXIST.

If gold loses its speculative value, you still have a very heavy, extremely conductive, corrosion resistant, malleable metal with substantial cultural importance.

When crypto collapses, you have literally nothing. It is supported entirely and exclusively by its value to speculators who only buy so that they can resell for profit and never intend to use it.

Well, not literally nothing. You have all that lovely carbon you burned to generate meaningless hashes polluting your biosphere for the next century. That part stays around long after crypto collapses.

I call blockchain an instantiation of Bostrom's Paperclip Maximizer running on a hybrid human-machine topology.

We are burning through scarce fuel in amounts sufficient to power a small developed nation in order to reverse engineer... one way hashcodes! Literally that is even less value than turning matter into paperclips.

The “$10 wrench attack” isn’t an argument against crypto—it’s an argument against human vulnerability.

By that logic, banks don’t work either, since people get kidnapped and forced to drain accounts. The difference is that with crypto, you can design custody systems (multi-sig, social recovery, hardware wallets, decentralized custody) that make such attacks far less effective than just targeting a centralized bank vault or insider.

As for the “end state” being dystopian, history shows centralized finance has already produced dystopias: hyperinflations, banking crises, mass surveillance, de-banking of political opponents, and global inequality enabled by monetary monopolies. Crypto doesn’t claim to magically create productivity—it creates an alternative infrastructure where value can be exchanged without gatekeepers. Productivity and crypto aren’t at odds: blockchains enable new forms of coordination, ownership, and global markets that can expand productive potential.

People now have the option of choosing between institutional trust and cryptographic trust—or even blending them. Dismissing crypto as doomed to dystopia ignores why it exists: because our current systems already fail millions every day.

What they are saying is that we have a system that evolved over time to address real world concerns. You are designing defenses to attacks that may or not be useful, but no one has been able to design past criminals and this is evident because if we could there would be no criminality.

> Dismissing crypto as doomed to dystopia ignores why it exists: because our current systems already fail millions every day.

This only makes sense if crypto solves the problems that current systems fail at. This have not been shown to be the case despite many years of attempts.

It may not be the absolute most useless, but it's awfully niche. You can use it to transfer money if you live somewhere with a crap banking system. And it's very useful for certain kinds of crime. And that's about it, after almost two decades. Plenty of other possibilities have been proposed and attempted, but nothing has actually stuck. (Remember NFTs? That was an amusing few weeks.) The technology is interesting and cool, but that's different from being useful. LLM chatbots are already way more generally useful than that and they're only three years old.

Gambling! That's actually the number one use case by far. Far beyond eg buying illicit substances. Regular money is much better for that.

Matt Levine (of Money Stuff fame) came up with another use case in a corporate setting: in many companies, especially banks, their systems are fragmented and full of technical debt. As a CEO it's hard to get workers and shareholders excited about a database cleanup. But for a time, it was easy to get people fired up about blockchain. Well, and the first thing you have to do before you can put all your data on the blockchain, is get all your data into common formats.

Thus the exciting but useless blockchain can provide motivational cover for the useful but dull sounding database cleanup.

(Feel free to be as cynical as you want to be about this.)

This is so true, from 2018-2021 my internal banking product was able to use blockchain hype to clean up a lot of our database schema. Our CTO was rubber stamping everything with the words blockchain and our customers were beating down the door to throw money at it.

Well that's hilarious. I wonder if LLMs might have a similar use case. I fear they tend to do the opposite: why clean up data when the computer can pretend to understand any crap you throw at it?

> AI is a powerful tool for those who are willing to put in the work.

No more powerful than I without the A. The only advantage AI has over I is that it is cheaper, but that's the appeal of the blockchain as well: It's cheaper than VISA.

The trouble with the blockchain is that it hasn't figured out how to be useful generally. Much like AI, it only works in certain niches. The past interest in the blockchain was premised on it reaching its "AGI" moment, where it could completely replace VISA at a much lower cost. We didn't get there and then interest started to wane. AI too is still being hyped on future prospects of it becoming much more broadly useful and is bound to face the same crisis as the blockchain faced if AGI doesn't arrive soon.

Blockchain only solves one problem Visa solves: transferring funds. It doesn't solve the other problems that Visa solves. For example, there is no way to get restitution in the case of fraud.

Yes, that is one of the reasons it hasn't been able to be used generally. But AI can't be used generally either. Both offer niche solutions for those with niche problems, but that's about it. They very much do feel the same, and they are going to start feeling even more the same if AGI doesn't arrive soon. Don't let the niche we know best around here being one of the things AI is helping to solve cloud your vision of it. The small few who were able to find utility in the blockchain thought it was useful too

Blockchain only has 2 legitimate uses (from an economic standpoint) as far as I can tell.

1) Bitcoin figured out how to create artificial scarcity, and got enough buy-in that the scarcity actually became valuable.

2)Some privacy coins serve an actual economic niche for illegal activity.

Then there's a long list of snake oil uses, and competition with payment providers doesn't even crack the top 20 of those. Modern day tulip mania.

Sounds like LLMs. The legitimate uses are:

1) Langauge tasks.

2) ...

I can't even think of what #2 is. If the technology gets better at writing code perhaps it can start to do other things by way of writing software to do it, but then you effectively have AGI, so...

But an I + and AI (as in a developer with access to AI tools) is as near as makes no difference the same price as just an I, and _can_ be better than just an I.

That's debatable -- if you follow all of the rules, and go through 25 review steps (which is the least fun part of software development...), you've squandered the advantage of using the AI in the first place. If you don't, you move very fast but your output is likely terrible and you'll eventually have to pay back the tech debt. The temptation to do less is extremely attractive here.

There are studies that show doctors with AI perform better in diagnoses tests than doctors without AI.

Also I think you’re cherry picking your experience. For one thing, I much prefer code review to code writing these days personally. And I find you don’t need to do “25 review steps” to ensure good output.

I’m just sensing lots of frustration and personal anecdote in your points here.

> My personal productivity has skyrocketed in the last 12 months.

If you don't mind me asking, what do you do?

>> Blockchain is probably the most useless technology ever invented

so useless there is almost $3 Trillion of value on blockchains.

Unfortunately, the amount of money invested in something isn't indicative of it's utility. For example: the tulip mania, beanie babies, NFTs, etc.

No there isn't. These ridiculous numbers are made up by taking the last price a coin sold for and multiplying it by all coins. If I create a shitcoin with 1 trillion coins and then sell one to a friend for $1 I've suddenly created a coin with $1 trillion in "value"

I'm not sure the 3 trillion includes shit coin valuations. Typically volume and liquidity are factored in for any serious analysis. Your assessment of valuation is probably just as true for traditional currencies and stocks. I guess the main difference is regulation.

Good luck with the SEC when you open your exchange

It's 2025, financial crimes are legal

3 trillion is just the top 5 or 6

It sounds like you are lacking inspiration. AI is a tool for making your ideas happen not giving you ideas.

I find your comment funny considering the OP, it's literally OpenAI claiming ChatGPT can start conversations now (in extenso give you inspiration/ideas).

> the most useless technology

Side-rant pet-peeve: People who try to rescue the reputation of "Blockchain" as a promising way forward by saying its weaknesses go away once you do a "private blockchain."

This is equivalent to claiming the self-balancing Segway vehicles are still the future, they just need to be "improved even more" by adding another set of wheels, an enclosed cabin, and disabling the self-balancing feature.

Congratulations, you've backtracked back to a classic [distributed database / car].

Going off on a tangent: the blockchain people have paid my salary for a few years now, and I still don't own any blockchain assets.

I do like the technology for its own sake, but I agree that it's mostly useless today. (At least most of it. The part of blockchain that's basically 'git' is useful, as you can see with git. Ie an immutable, garbage colected Merkle-tree as a database, but you trust that Linux Torvalds has the pointer to the 'official' Linux kernel commit, instead of using a more complicated consensus mechanism.)

However there's one thing that's coming out of that ecosystem that has the potential to be generally useful: Zero Knowledge Proofs.

To quote myself (https://news.ycombinator.com/item?id=45357320):

Yes, what Zero Knowledge proofs give you however is composability.

Eg suppose you have one system that lets you verify 'this person has X dollars in their bank account' and another system that lets you verify 'this person has a passport of Honduras' and another system that lets you verify 'this person has a passport of Germany', then whether the authors of these three systems ever intended to or not, you can prove a statement like 'this person has a prime number amount of dollars and has a passport from either Honduras or Germany'.

I see the big application not in building a union. For that you'd want something like Off-The-Record messaging probably? See https://en.wikipedia.org/wiki/Off-the-record_messaging

Where I see the big application is in compliance, especially implementing know-your-customer rules, while preserving privacy. So with a system outlined as above, a bank can store a proof that the customer comes from one of the approved countries (ie not North Korea or Russia etc) without having to store an actual copy of the customer's passport or ever even learning where the customer is from.

As you mentioned, for this to work you need to have an 'anchor' to the real world. What ZKP gives you is a way to weave a net between these anchors.

Honestly the Segway had a couple of useful effects, none of which involved the Segway itself becoming useful or popular.

1. The self-balancing tech spread to a variety of more interesting and cheaper "toy" platforms like hoverboards and self-balancing electric unicycles. 2. They encouraged the interest in electric micromobility, leading to, especially, electric scooters (which are simpler, cheaper, and use less space) becoming widespread.

This is kind of like the point someone else made that the actual useful thing the blockchain craze produced was "cleaning up your database schemas with the idea of putting them on the blockchain, then never putting them on the blockchain".

> Blockchain is probably the most useless technology ever invented (unless you're a criminal or an influencer who makes ungodly amounts of money off of suckers)

This is an incredibly uneducated take on multiple levels. If you're talking about Bitcoin specifically, even though you said "blockchain", I can understand this as a political talking about 8 years ago. But you're still banging this drum despite the current state of affairs? Why not have the courage to say you're politically against it or bitter or whatever your true underlying issue is?

How is the current state of affairs different from 8 years ago? I don't want to argue, just a genuine question, because I don't follow much what's happening in the blockchain universum.

I often pay for freelance work using crypto because it's much easier than dealing with banking infra to send payments to dozen of different countries. Sure you lose safety bank provides, but so you do when using cash.

so nothing has changed and blockchain is still a solution looking for a problem

I am in the camp that thinks that Blockchain is utterly useless for most people. Perhaps you can tell us about some very compelling use cases that have taken off.

https://www.reuters.com/business/finance/nasdaq-makes-push-l... https://corporate.visa.com/en/solutions/crypto/stablecoins.h... https://www.barrons.com/articles/sofi-blockchain-money-trans...

Take a look at "agentic commerce": it is using AI agents + Blockchain for micro-payments: the future of payments

could you please take a moment to explain what problem this actually solves with no marketing language please

This is funny, because my personal view is that AI’s biggest pitfall is that it allows the unqualified to build what they think they’re qualified for

Yes, learning and creating should be reserved only for the licensed!

It is in many fields. They're Professional Engineers, and they build nearly everything that's important to a society. Software is overrun by literally anyone and everyone because it's not a protected field of engineering

Hang on, my understanding of the situation is that only PEs are allowed to sign off on the design of a thing. They don't have to design it, and they almost certainly don't build it. Am I wrong?

The stamp is legally binding. The company you work for isn't gonna protect you financially if the bridge you stamped falls. This is how they protect their profession

https://en.wikipedia.org/wiki/List_of_data_breaches

>Blockchain is probably the most useless technology ever invented (unless you're a criminal or an influencer who makes ungodly amounts of money off of suckers).

You think a technology that allows millions of people all around the world to keep & trustlessly update a database, showing cryptographic ownership of something "the most useless technology ever invented"?

> Blockchain is probably the most useless technology ever invented

You can use blockchains to gamble and get rich quick, if you're lucky.

That's a useful thing. Unlike "AI", which only creates more blogspam and technical debt in the world.

Bad take about blockchain. Being able to send value across borders without intermediaries is unheard of in human history.

No it's not. It's existed for a thousand years in Asia through the hawala system. I mean, you need an intermediary. But with blockchain your intermediary is the exchange, the developers of the cryptocurrency, etc.

>with blockchain your intermediary is the exchange, the developers of the cryptocurrency, etc.

You are mistaken. The transfer of cryptocurrencies takes place on the blockchain ledger. That's, like, the core "thing".

If you choose to bank with an exchange, that's like you bringing your cash money to a regular bank and depositing it. And the developers are no more intermediaries to your transaction than the Mint is an intermediary to your cash transaction.

I'm not mistaken. For an average person, they need to:

1. Understand how to use the technology and onboard onto it. In this case the technology is whichever blockchain and cryptocurrency.

2. Convert their money into the cryptocurrency. For this they need an exchange.

Then they can send money to others who also went through those steps.

Thus, there must be some kind of interaction with documentation made by devs for step 1 and a transfer of currency to an exchange for step 2. 2 middlemen/intermediaries.

All of that is entirely besides the point of this thread of conversation, which was concerning the transfer without intermediaries. The point was that if you are in possession of the currency, you are in full control with no one to interfere(or interfere by withholding cooperation) in the way that a bank would.

You also do not need an exchange. You need to find a person willing to trade whatever for your cryptocurrency, exchanges merely make this more convenient but are by no means the only option.

And saying onboarding requires an intermediary is like saying using a powerdrill or hooking up a washing machine requires an intermediary. The knowledge is so widely available and not contingent on a single or even few authorities that it's like picking fruit from a tree. It's just there.

> which was concerning the transfer without intermediaries

I mean, you have to onboard to transfer alternative currencies. People have cash and bank accounts. They don't have any random cryptocurrency.

> You need to find a person willing to trade whatever for your cryptocurrency, exchanges merely make this more convenient but are by no means the only option.

This isn't a realistic option that will work at scale and you know that.

Hawala isn't special. It's the same idea behind any bank transfers in general.

Yes, it's just the most untraceable of such transactions (all of which require intermediaries)

if you think it was a technology issue i've got some ice to sell you

New technology usually gets most of its power from just finding ways of skirting existing regulation.

Take a look at new payments protocols for m AI agents

> The real problem isn’t AI itself; it’s the overblown promise that it would magically turn anyone into a programmer, architect, or lawyer without effort, expertise or even active engagement. That promise is pretty much dead at this point.

If that's dead, as you say, it would mean billions in value destruction for every tech stock. They have already promised so far beyond that.

Yes, they are going to blow a lot of money on blowing hot air into this

There are two main use cases for blockchain assets at the moment:

- Gambling (or politely called, speculation or even 'investment')

- Ransomware payments, but that's a distant second

I guess you say that influencers make money off the first? There's a third group making money off blockchain: the people making markets and running the infrastructure.

You forgot money laundering and paying for illegal things.

No, I left them out deliberately.

At some point in the past, you could perhaps buy illicit drugs with bitcoin, but that has largely died out, as far as I can tell. You can't even buy a pizza anymore.

and sending money cross borders to your family and paying suppliers outside of your country and having a multi-currency account without insane fees and automating finance and upgrading financial systems to modern tech

I found interacting with the blockchain world to be much higher in fees than when going via the real finance route.

> AI is a powerful tool for those who are willing to put in the work.

HN crowd is generally anti-AI, so you're not going to get a lot of positive feedback here.

As a developer and someone who runs my own company, AI helps save me tons of hours - especially with research tasks, code audits - just to see if I've missed something, rapid frontend interface development and so much more if we go beyond the realm of just code.

YMMV but I use ChatGPT even for stuff like cooking recipes based on whatever's inside my fridge. And this maybe just me, but I feel like my usage of Google has probably gone down to 50-60%.

All this for $20 is really insane. Like you said, "people who are willing to put in work" really matters - that usually means being more intentional about your prompts, giving it as much context as possible.

> HN crowd is generally anti-AI, so you're not going to get a lot of positive feedback here.

Is this a joke? The front page of HN is 99% AI-positive. Things have calmed down a bit but there was a time when it felt like every other article on the front page was promoting AI.

Rather than anti-AI, I think most of the HN crowd are educated enough about underlying technology and it's capabilities to see the snake oil being pushed. Most of us would have given it a good go (and probably still use it daily), just know the difference between its actual limitations and the dreams that are being sold.

Is it really snake oil if it actually has meaningful time savings? Dismissing AI completely off as just snake oil seems to be extreme.

> AI is a powerful tool for those who are willing to put in the work. People who have the time, knowledge and critical thinking skills to verify its outputs and steer it toward better answers. My personal productivity has skyrocketed in the last 12 months.

That sounds like calling stupid and lazy to anyone that is not able to get their "productivity skyrocketed" with LLMs. The reality is that LLMs are mostly a search engine that lies time to time.

No, people is not stupid nor lazy for not gaining any substantial productivity using LLMs. The technology is just not able to do what big tech corporations say it can do.

How did you measure this skyrocketed productivity? Please share your results and methodology!

> It doesn't feel like blockchain at all. Blockchain is probably the most useless technology ever invented (unless you're a criminal or an influencer who makes ungodly amounts of money off of suckers).

Given the prevailing trend of virtually all modern governments in the developed world that seems to be rather short-sighted. Who knows, if you won't be made a criminal for dissenting? Or for trying to keep your communication private?

Honestly, "only useful for crime" is the only part of blockchain that I like. What I don't like is the energy use of PoW, the centralization tendencies of PoS, and the general unusability of any existing coins as either a medium of exchange or a store of value. I'd love an anonymous digital cash that didn't boil the oceans, enrich oligarchs, or encourage speculation. It's just that blockchain cryptocurrencies are not that.

> Blockchain is probably the most useless technology ever invented (unless you're a criminal or an influencer who makes ungodly amounts of money off of suckers).

Are you sure you are not confusing blockchain with crypto coins? Blockchain as a decentralised immutable and verifiable ledger actually has use in 0-trust scenarios.

Actual "0-trust scenarios" are incredibly elusive in the real world. It pretty much comes down to illegal or "extra-legal" transactions. Almost nobody in the real world has a practical interest in carrying out transactions with people they don't trust and people that our outside the jurisdiction of their legal framework.

It turns out that most people actually like being able to press criminal charges or have the ability to bring a law suit when the transaction goes south.

How is blockchain technology mutually exclusive from being able to take legal action against e.g. fraud?

Sure there are implementations that can be used that way, but the core idea of a blockchain could be used to do the opposite as well, for example by making transaction information public and verifiable.

At some point in its lifecycle, it's no longer enough to hear about what a technology "could be used" for, because the time has come to judge it on its actual real world uses.

Only is not AI. It's LLM, diffusion models and what have you. The term AI became so overloaded that it means almost nothing these days. AGI is still hard to reach for, but it doesn't stop all kinds of spin artists from trying.

>AI is a powerful tool for those who are willing to put in the work.

Can we please call this technology transformers? Calling it AI makes it seem something more than it is (ie 2100 tech or something like that). Yes, transformers are great but it would be naive to ignore that much of the activity and dreams sold have no connection with reality and many of those dreams are being sold by the very people that make the transformers (looking at you, OpenAI)

> I’m rapidly losing interest in all of these tools

Same. It reminds me the 1984 event in which the computer itself famously “spoke” to the audience using its text-to-speech feature. Pretty amazing at that time, but nevertheless quite useless since then

It has proven very useful to a great number of people who, although they are a minority, have vastly benefited from TTS and other accessibility features.

I think it's easy to pick apart arguments out of context, but since the parent is comparing it to AI, I assume what they meant is that it hasn't turned out to be nearly as revolutionary for general-purpose computing as we thought.

Talking computers became an ubiquitous sci-fi trope. And in reality... even now, when we have nearly-flawless natural language processing, most people prefer to text LLMs than to talk to them.

Heck, we usually prefer texting to calling when interacting with other people.

I don’t think anyone watched that demo back in 1984 and thought “oh this tech means we can talk to computers!” - it was clearly a demonstration of… well text to speech. It demonstrated successfully exactly what it could do, and didn’t imply what they’re implying it implied.

Text to speech has been an incredible breakthrough for many with vision, visual processing, or speech disabilities. You take that back.

Stephen Hawking without text to speech would’ve been mute.

Its not useless, you're just taking it for granted. The whole national emergency system works off text to speech.

> It feels like blockchain again in a lot of weird ways.

Every time I keep seeing this brought up I wonder if people truly mean this or its just something people say but don't mean. AI is obviously different and extremely useful.. I mean it has convinced a butt load of people to pay for the subscription. Every one I know including the non technical ones use it and some of them pay for it, and it didn't even require advertising! People just use it because they like it.

Obviously a lot of grifters and influencers shifted from NFTs to AI, but the comparison ends there. AI is being used by normal people and professionals every day. In comparison, the number of people who ever interacted with blockchain is basically zero. (And that's a lifetime vs daily comparison)

It's a lazy comparison, and most likely fueled by a generic aversion to "techbros".

That aversion to "techbros" comes from people who constantly oversell whatever the current thing is for personal benefit, and these technologies come up short time and time again. The "techbros" as a consort have proven themselves not to be trusted. Say what you want about AI, it's hard to argue that it isn't oversold constantly. I sincerely doubt it would continue to exist without a stream of VC money to sustain it (All of those paying customers still don't cover the bills...)

Everything is oversold, even the useful things. Nuclear, space, genetic engineering, x rays, vaccines, the internet itself, personal computers, phones and apps, etc. etc.

If something is of value X, the hype will be X+Y. That's why this is too shallow an analyis. It's just vibes, based on cultural feelings and emotional annoyance.

AI is nothing like blockchain.

It's fine to be anti-AI, but the smart anti-AI people recognize that the real danger is it will have too large an impact (but in the negative). Not that it will evaporate some investment dollars.

"It has convinced a bunch of people to spend money" is also true of blockchain, so I don't know if that's a good argument to differentiate the two.

The extent matters. Do you think we need a good argument to differentiate Netflix?

> I tried to ask GPT-5 pro the other day to just pick an ambitious project it wanted to work on, and I’d carry out whatever physical world tasks it needed me to, and all it did was just come up with project plans which were rehashes of my prior projects framed as its own.

Mate, I think you’ve got the roles of human and AI reversed. Humans are supposed to come up with creative ideas and let machines do the tedious work of implementation. That’s a bit like asking a calculator what equations you should do or a DB what queries you should make. These tools exist to serve us, not the other way around

GPT et al. can’t “want” anything, they have no volition

The posted article starts off by claiming that the AI can take the lead. I don't believe their hype, but I think that it the basis for the above comment.

"Since ChatGPT launched, that's always meant coming to ask a question .... However that's limited by what you know to ask for and always puts the burden on you for the next step."

Yeah I've tried some of the therapy prompts, "Ask me 7 questions to help me fix my life, then provide insights." And it just gives me a generic summary of the top 5 articles you'd get if you googled "how to fix depression, social anxiety" or something.

Argue with it. Criticize it. Nitpick the questions it asked. Tell it what you just said:

you just gave me a generic summary of the top 5 articles you'd get if you googled "how to fix depression, social anxiety" or something

When you open the prompt the first time it has zero context on you. I'm not an LLM-utopist, but just like with a human therapist you need to give it more context. Even arguing with it is context.

I do, frequently, and ChatGPT in particular gets stuck in a loop where it specifically ignores whatever I write and repeats the same thing over and over again.

To give a basic example, ask it to list some things and then ask it to provide more examples. It's gonna be immediately stuck in a loop and repeat the same thing over and over again. Maybe one of the 10 examples it gives you is different, but that's gonna be a false match for what I'm looking for.

This alone makes it as useful as clicking on the first few results myself. It doesn't refine its search, it doesn't "click further down the page", it just wastes my time. It's only as useful as the first result it gives, this idea of arguing your way to better answers has never happened to me in practice.

I would recommend talking out loud to it with speech to text software, it has been pretty helpful for me. Like around 60s of talking.

I did, and I gave it lots of detailed, nuanced answers about my life specifics. I spent an hour answering its questions and the end result was it telling me to watch the movie "man called otto" which I had already done (and hated) among other pablum.

The free model?

Agreed. I think AI can be a good tool, but not many people are doing very original stuff. Plus, there are many things I would prefer be greeted with, other than by an algorithm in the morning.

I got in a Waymo today and asked it where it wanted to go. It tried to suggest places I wanted to go. This technology just isn't there.

Reminded me of many movie plots where a derailed character sits in a taxi and, when asked where to go, replies with "anywhere" or "I don't know." But before imagining a terrible future where an AI-driven vehicle actually decides, I suggest imagining an AI-infused comedy exploring this scenario. /s

They’re more like synthesizers or sequencers: if you have ideas, they are amazing force multipliers, but if you don’t have ideas they certainly won’t create them for you.

Try turning off memory. I've done a lot of experiments and find ChatGPT is objectively better and more useful in most ways with no memory at all. While that may seem counter-intuitive, it makes sense the more you think about it:

(1) Memory is primarily designed to be addictive. It feels "magical" when it references things it knows about you. But that doesn't make it useful.

(2) Memory massively clogs the context window. Quality, accuracy, and independent thought all degrade rapidly with too much context -- especially low-quality context that you can't precisely control or even see.

(3) Memory makes ChatGPT more sychophantic than it already is. Before long, it's just an echo chamber that can border on insanity.

(4) Memory doesn't work the way you think it does. ChatGPT doesn't reference everything from all your chats. Rather, your chat history gets compressed into a few information-dense paragraphs. In other words, ChatGPT's memory is a low-resolution, often inaccurate distortion of all your prior chats. That distortion then becomes the basis of every single subsequent interaction you have.

Another tip is to avoid long conversations, as very long chats end up reproducing within themselves the same problems as above. Disable memory, get what you need out of a chat, move on. I find that this "brings back" a lot of the impressiveness of the early version of ChatGPT.

Oh, and always enable as much thinking as you can tolerate to wait on for each question. In my experience, less thinking = more sychophantic responses.

I might want to have an LLM hit me with temperature 100% weird-ass entropic thoughts every day.

Other that that, what recycled bullshit would I care about?

Totally agree on the memory feature. You just have to look at the crap it tries to remember to see how useless it is and the kind of nonsense it will jam into the context.

“Cruffle is trying to make bath bombs using baking soda and citric acid and hasn’t decided what colorant to use” could be a memory. Yeah well I figured out what colorant to use… you wanna bet if it changed that memory? Nope! How would it even know? And how useful is that to keep in the first place? My memory was full of useless crap like that.

There is no way to edit the memories, decide when to add them to the context, etc. and adding controls for all of that is a level of micromanaging I do not want to do!

Seriously. I’ve yet to see any memory feature that is worth a single damn. Context management is absolutely crucial and letting random algorithms inject useless noise is going to degrade your experience.

About the only useful stuff for it to truly remember is basic facts like relationships (wife name is blah, kid is blah we live in blah blah). Things that make sense for it to know so you can mention things like “Mrs Duffle” and it knows instantly that is my wife and some bit about her background.

Agreed. I disabled memory at the moment I noticed that random stuff from other chats was showing up in my image generations.

Maybe the next hype train won't rely on GPU's.

Replies you get tell me it is more akin to Agile movement at this point. People are mixing “lines of code per minute” with tangible results.

Are you more personally ‘productive’ if your agent can crunch out PoCs of your hobby projects at night when you sleep? Is doing more development iterations per month making your business more ‘productive’?

It is like expecting that achieving x10 more sunny side ups cooked per minute will make your restaurant more profitable. In reality amount of code delivered is rarely a bottleneck for value delivered, but for ‘productivity’ everyone has their own subjective definition.

Comparing this to blockchain and Web3 is fucking hilarious. Web3 literally has like no real use case, LLM tech has a billion. It literally passes the Turing test.

How is a Turing-passing AI useful?

No use case?

How are AI agents going to open up a credit card or bank account?

You think the government will grant them a SSN?

It's a good thing we have a Internet native, programmable money, that is permissionless, near instant, and already supports USD(c).

As a bystander watching the world burn, I find it hilarious that you're both opposing and arguing against the other's chosen planet-incinerator.

I think you're expecting too much from it? It's great at "interpolating" across its training data, but it can't really "extrapolate" (think of brand new, ambitious ideas). If you gave it some more constraints then it could definitely come up with an ambitious project for you, but it wouldn't be something entirely new to the world.

The name reminds me of https://linear.app/docs/pulse which provides regular updates on projects tracked in the tool.

This is not what I want from LLMs. Hey ChatGPT or Claude, I have still have clothes in the dryer and dishes in the sink. Do you mind taking care of those for me?

Now, that is something I'd pay hundreds a month for.

Since every "AI" company frantically releases new applications, may I suggest OpenAI+ to copy the resounding success of Google+?

Google+ is incidentally a great example of a gigantic money sink driven by optimistic hype.

I’m a pro user.. but this just seems like a way to make sure users engage more with the platform. Like how social media apps try to get you addicted and have them always fight for your attention.

Definitely not interested in this.

Desperation for new data harvesting methodology is a massive bear signal FYI

Calm down bear we are not even 2% from the all time highs

> Centillion is not some big scary government. It’s a private company, whose motto happens to be “Make things better!”

Haha, good shot at Big Tech. They always devolve to corporate profits over ethics.

There's the monitization angle!

A new channel to push recommendations. Pay to have your content pushed straight to people as a personalized recommendation from a trusted source.

Will be interesting if this works out...

Necessary step before making a move into hardware. An object you have to remember to use quickly gets forgotten in favor of your phone.

But a device that reaches out to you reminds you to hook back in.

Breaking the request response loop and entering into async territory?

Great!

The examples used?

Stupid. Why would I want AI generated buzzfeed tips style articles. I guess they want to turn chatgpt into yet another infinite scroller

"Now ChatGPT can start the conversation"

By their own definition, its a feature nobody asked for.

Also, this needs a cute/mocking name. How about "vibe living"?

Of course i will want to be distracted by my AI but first it has to track everything i say and do. And i wouldn't mind if it talked some smack about my colleagues

Human to robot servant: Do not speak unless spoken to machine!

Does this seem a step closer to getting ads on ChatGPT?

It's very hard for me to envision something I would use this for. None of the examples in the post seem like something a real person would do.

Wow, did ChatGPT come up with that feature?

Holy guacamole. It is amazing all the BS these people are able to create to keep the hype of the language models' super powers.

But well I guess they have committed 100s of billions of future usage so they better come up with more stuff to keep the wheels spinning.

Funny, I pitched a much more useful version of this like two years ago with clear use-cases and value proposition

Waiting for time when it can summarize my social media - so I won't need to waste time scrolling.

Let the personal ensloppification begin!

This is a joke. How are people actually excited or praising a feature that is literally just collecting data for the obvious purpose of building a profile and ultimately showing ads?

How tone deaf does OpenAI have to be to show "Mind if I ask completely randomly about your travel preferences?" in the main announcement of a new feature?

This is idiocracy to the ultimate level. I simply cannot fathom that any commenter that does not have an immediate extremely negative reaction about that "feature" here is anything other than an astroturfer paid by OpenAI.

This feature is literal insanity. If you think this is a good feature, you ARE mentally ill.

They seem desperate to find something that might have relevance (besides it's core product)

I am pleading with you all. Don't give away your entire identity to this or any other company.

Was quite unimpressive. In general ChatGPT has been degrading in default quality for months

LLMs hit diminishing returns this year. There is nothing more they can do and it shows.

Don't get me wrong, the coding assistants are amazing and overall functionality of asking questions is great, not to mention government spying - cia and mossad are probably happy beyond beliefs. But I do not see any more use cases for it.

From my perspective, the core of OpenAI's current situation can be understood through a few key observations and predictions.

First, over $55 billion raised since 2022 has fueled monumental, yet sometimes subtle, leaps in model capability that go far beyond the visible user interface. The real progress isn't just in the chat window, but in the underlying reasoning and multimodal power.

This investment is being funneled into what I see as a relentless and unsustainable cycle of spending: the high cost of training next-generation models, the immense expense of running user queries (inference), and a fierce bidding war for top AI talent.

Based on this, it's clear to me that the current "growth at all costs" phase is unsustainable. This reality will inevitably force a strategic shift from a pure focus on innovation to the pressing need for a viable, profitable business model.

Therefore, I predict they will be forced to take specific cost-saving steps. To manage the colossal expense of inference, models will likely be subtly "tuned down," masking reduced computational depth with more verbose, fluffy output.

Finally, I anticipate that a logical, if controversial, cost-saving step will involve the company using its own AI to justify workforce reductions, championing this as the new era of automated productivity they are selling to the world.

And by the way, downvoting this won’t prevent the unavoidable future.

I mostly agree, but I think they are in a overall better position than most social media/messaging/video hosting platforms were over the last decades, because a significant fraction of their customers is willing to pay for this kind of service already (they are not inevitably forced to rely on ads for revenue) and cost per token (for OpenAI) can reasonably be expected to decrease continuously (unlike costs for ridesharing startups like uber, or the e-scooter thingies).

"Therefore, I predict they will be forced to take specific cost-saving steps"

They are already doing that dude with the router... lol wake up. Youre late to the party.

Man, my startup does this but exclusively for enterprises, where it actually makes sense

I see some pessimism in the comments here but honestly, this kind of product is something that would make me pay for ChatGPT again (I already pay for Claude, Gemini, Cursor, Perplexity, etc.). At the risk of lock-in, a truly useful assistant is something I welcome, and I even find it strange that it didn't appear sooner.

Truly useful?

Personal take, but the usefulness of these tools to me is greatly limited by their knowledge latency and limited modality.

I don't need information overload on what playtime gifts to buy my kitten or some semi-random but probably not very practical "guide" on how to navigate XYZ airport.

Those are not useful tips. It's drinking from an information firehose that'll lead to fatigue, not efficiency.

I doubt there would be this level of pessimism if people thought this is a progress toward a truly useful assistant.

Personally it sounds negative value. Maybe a startup that's not doing anything else could iterate on something like this into a killer app, but my expectation that OpenAI can do so is very, very low.

Pessimism is how people now signal their savviness or status. My autistic brain took some time to understand this nuance.

Wasn't this already implemented via google and apple separately?

We have built https://hopit.ai. We have not been running the agents right now but you folks can experience what is it to make everything bingeable and beautiful.

I mean, it is expensive to run those. We started this as a hobby project. Can switch on the agents if folks like the format

Great way to sell some of those empty GPU cycles to consumers

This smells like ads.

In the words of Chief Wiggum once again:

"this is going to get worse before it gets better."

https://www.youtube.com/watch?v=Hi6iOFHLTdw

Is it like Google Now mixed with Twitter and Sora feed?

The sort of application I'd much rather self-host.

Big tech companies today are fighting over your attention and consumers are the losers.

I hate this feature and I'm sure it will soon be serving up content that is as engaging as the stuff the comes out of the big tech feed algorithms: politically divisive issues, violent and titillating news stories and misinformation.

Watch out, Meta. OpenAI is going to eat your lunch.

Meta is busy with this: https://news.ycombinator.com/item?id=45379514

sounds nice I guess, but reactiveness over proactiveness wasn't the pain point I've had with these LLM tools.

Is this not cognitive offload on steroids?

This is so NOT something I want.

Might as well call it WokeGPT as the results invariable lean to the left of diversity. You get a mini woke sermon at the end of each results. just in case you guilty of wrong thought.

Can you give an example?

What kind of stuff is it showing?

Ewwwwwww

$200 a month for data cards is insanity. It sounds like virtue-signaling or bs features for VC's. Idk why this wasn't pushed to the consumer plus level.

Here's an excerpt from ChatGPT on why this could be a "bubble" feature:

By sticking Pulse behind the $200 Pro tier, OpenAI is signaling it’s for:

VCs, consultants, analysts → people who can expense it as “deal flow intel.”

Enterprise & finance people who live in dashboards and daily reports.

Folks who don’t blink at $200/month because they already burn way more on data feeds, research subscriptions, or Bloomberg terminals.

In other words, it feels less like a consumer feature and more like a “bubble luxury” signal — “If you need Pulse, you’re in the club that can afford Pro.”

The irony is, Pulse itself isn’t really a $200/mo product — it’s basically automated research cards. But bundling it at that tier lets OpenAI:

Frame it as exclusive (“you’re not missing out, you’re just not at the level yet”).

Keep the Plus plan sticky for the masses.

Extract max revenue from people in finance/AI hype cycles who will pay.

It’s like how Bloomberg charges $2k/month for terminals when most of the raw data is public — you’re paying for the packaging, speed, and exclusivity.

I think you’re right: Pulse at $200 screams “this is a bubble feature” — it’s monetizing the hype while the hype lasts.

In the beginning, I was really curious about ChatGPT—a tool that could save me from useless blogs, pushy products, and research roadblocks. Then it started asking follow-up questions, and I got a bit uneasy… where is this trying to take me? Now it feels like the goal is to pull me into a discussion, ultimately consulting me on what? Buy something? Think something? It’s sad to see something so promising turn into an annoying, social-network-like experience, just like so many technologies before. As with Facebook or Google products, maybe we’re not the happy users of free tech—we’re the product. Or am I completely off here? For me, there’s a clear boundary between me controlling a tool and a tool broadcasting at me.

Don’t forget that Sam Altman is also the cryptocurrency scammer who wants your biometric information. The goal was and will always be personal wealth and power, not helping others.

https://www.technologyreview.com/2022/04/06/1048981/worldcoi...

https://www.buzzfeednews.com/article/richardnieva/worldcoin-...

I lol'd at this. And he eats babies!

Yea, humans tend to optimize for their own individual success. Do you not?

Some actually promote ideas and defend ideals, individual "success" doesn't always correlates with fulfilment. For instance, working for a non-profit, providing help and food to people who need it, helping associations live or creating a company that maximize doing good instead of profit.

> Do you not?

At the expense of and with complete disregard for others, while telling them I’m doing it altruistically for their own good? I don’t, no. Do you?

It's important to remember that hackernews (and ycombinator) is largely populated by wannabe VC sociopaths.

Hrmmm.

Engaging in wild character assassination (ie. spewing hate on the internet) in return for emotional upvote validation...I would argue is a great example of what you just described.

I don't doubt you've convinced yourself you're commenting these things altruistically.

But remember, when high school girls spread rumors that the good-looking popular girl has loose morals, they aren't doing it out of concern for public good.

They're hoping to elevate their own status by tearing down the competition, and avoiding the pain of comparison by placing themselves on a higher moral pedestal.

You’re comparing transitory conversations on a forum to a project with proven negative impact which was banned in multiple countries, and serious investigative journalism to teenage gossip, while ascribing twisted motivations to strangers on the internet.

That you would start from the assumption that someone’s motivation for a comment is not their opinion but a desire to farm meaningless internet points is bizarre. I have no idea if that’s how you operate—I don’t know you, I wouldn’t presume—but I sincerely hope that is not the case if we ever hope to have even a semblance of a productive conversation.

You are right, we all do at some point. Further I think it is OpenAIs right to do so as they invented the product. But in this case I feel even more gaslighted. It's as with the internet as a whole. A great invention for society but pretty much unusable or misused as products are designed for those people success.

Homo economicus is a fallacy

I'm not sure I understand your point. Is there any conversation where this behaviour occurred that you feel comfortable sharing?

I haven't observed it, and I'd be curious to understand if I've missed it, or if me and you evaluate things differently.

Their point is that ChatGPT is clearly maximizing for engagement with the product, rather than working as a straightforward tool that aims to solve your problem at hand as fast and efficiently as possible.

The decision to begin every response with “that’s a fascinating observation” and end every response with “want me to do X?” is a clear decision by PMs at OpenAI.

Poster is questioning what might motivate those decisions.

Exactly this. Thanks for clarifying. It is the Want me to do X? thing that makes me think why would you ask? Further on the Pulse feature: Why would you start a conversation?

You tell it to stop doing that.

This only works as long as they let you

There totally is! https://www.nytimes.com/2025/08/08/technology/ai-chatbots-de...

It's as if OpenAI saw that as an instruction manual, I really don't like the direction they're taking it.

Likewise Ken Liu (the English translator for the Three Body Problem) has a really good short story "The Perfect Match" about the same concept, which you can read here: https://www.lightspeedmagazine.com/fiction/the-perfect-match... It was the first thing that came to mind when I read this announcement.

Scott Alexander has a similar 1-page horror story from 2012 where an earing that gives advice consumes the user: https://gwern.net/doc/fiction/science-fiction/2012-10-03-yva...

try clicking "Listen to article"

Hola

Product managers live in a bubble of their own.

It seems not useful for 95% of users today, but later can be baked into the hardware Ive designed. so, good luck, I guess?

Absolutely not. No. Hard pass.

Why would I want yet another thing to tell me what I should be paying attention to?

AI doesn't have a pulse. Am I the only one creeped out by personification of tech?

"Pulse" here comes from the newspaper/radio lineage of the word, where it means something along the lines of timely, rhythmic news delivery. Maybe there is reason to be creeped out by journalists from centuries ago personifying their work, but that has little to do with tech.

is not this just privacy nightmare???

we have google that has search history,location etc now we give OpenAI our character,personalization etc

how much dystopian that in the future there will be a product based on people feeling from 5 years ago????

I need this bubble to last until 2026 and this is scaring me.

Vesting window?

I'm just stubborn and don't want to go defensive yet with rate cuts here. I think the rich want one last good peak to get out on.

Welcome to cognitive capitalism https://defragzone.substack.com/p/welcome-to-cognitive-capit...

so GPT tiktok in nutshell

Technology service technology, rather than technology as a tool with a purpose. What is the purpose of this feature?

This reads like the first step to "infinite scroll" AI echo chambers and next level surveillance capitalism.

On one hand this can be exciting. Following up with information from my recent deep dive would be cool.

On the other hand, I don't want to it to keep engaging with my most recent conspiracy theory/fringe deep dives.

Thanks, I hate it.

As with everything in the AI space currently:

A strange mix of dystopian yet blissfull...

Yet another category of startups killed by an incumbent

lol wtf, do they know it's not April 1st

ChatGPT IV

Episodes from Liberty City?

Why?

Wow, this is a new level of desperation..

OpenAI is a trillion dollar company. No doubt.

Edit: Downvote all you want, as usual. Then wait 6 months to be proven wrong. Every. Single. Time.

I downvoted because this isn’t an interesting comment. It makes a common, unsubstantiated claim and leaves it at that.

> Downvote all you want

“Please don't comment about the voting on comments. It never does any good, and it makes boring reading.”

https://news.ycombinator.com/newsguidelines.html

Welcome to HN. 98% of it is unsubstantiated claims.

Here's a free product enhancement for OpenAI if they're not already doing this:

A todo app that reminds you of stuff. say "here's the stuff I need to do, dishes, clean cat litter fold laundry and put it away, move stuff to dryer then fold that when it's done etc." then it asks about how long these things take or gives you estimates. Then (here's the feature) it checks in with you at intervals: "hey it's been 30 minutes, how's it going with the dishes?"

This is basically "executive function coach." Or you could call it NagBot. Either way this would be extremely useful, and it's mostly just timers & push notifications.

This will drive the opposite of user engagement.

Humbly I suggest vibecoding this just for yourself. Not building a product - just a simple tool to meet your own needs.

That’s AI: permissionless tool building. It means never needing someone to like your idea enough or build it how they think you’ll use it. You just build it yourself and iterate it.

Can this be interpreted as anything other than a scheme to charge you for hidden token fees? It sounds like they're asking users to just hand over a blank check to OpenAI to let it use as many tokens as it sees fit?

"ChatGPT can now do asynchronous research on your behalf. Each night, it synthesizes information from your memory, chat history, and direct feedback to learn what’s most relevant to you, then delivers personalized, focused updates the next day."

In what world is this not a huge cry for help from OpenAI? It sounds like they haven't found a monetization strategy that actually covers their costs and now they're just basically asking for the keys to your bank account.

We don't charge per token in chatgpt

No, it isn’t. It makes no sense and I can’t believe you would think this is a strategy they’re pursuing. This is a Pro/Plus account feature, so the users don’t pay anything extra, and they’re planning to make this free for everyone. I very much doubt this feature would generate a lot of traffic anyway - it’s basically one more message to process per day.

OpenAI clearly recently focuses on model cost effectiveness, with the intention of making inference nearly free.

What do you think the weekly limit is on GPT-5-Thinking usage on the $20 plan? Write down a number before looking it up.

If you think that inference at OpenAI is nearly free, then I got a bridge to sell you. Seriously though this is not speculation, if you look at the recent interview with Altman he pretty explicitly states that they underestimated that inference costs would dwarf training costs - and he also stated that the one thing that could bring this house of cards down is if users decide they don’t actually want to pay for these services, and so far, they certainly have not covered costs.

I admit that I didn’t understand the Pro plan feature (I mostly use the API and assumed a similar model) but I think if you assume that this feature will remain free or that its costs won’t be incurred elsewhere, you’re likely ignoring the massive buildouts of data centers to support inference that is happening across the US right now.

Reply on news.ycombinator.com