Everybody's Lost Their Minds | Hacker News Reader

Everybody's Lost Their Minds

netmeister.org

373 pointsibobev348 comments

I feel this post. I am tired of "directing" agents when in reality it feels more like trying to herd a group of toddlers.

Sure they can mostly write better code than a toddler but this constant nudging and reminding and reiterating and stopping them from using the token budget of the whole company for a one off script. It gets tiring and I feel like I am losing brain power while doing it. Maybe it's faster but explosive diarrhea is also a faster way to produce shit.

AI builds like 95% for me than i play around with it for a bit for the last 5%.

I'm still a lot faster for certain things with AI but its weird. Instead of just being annoyed about some weird bug, i'm annoyed about an LLM which can do everything and then falls over itself on weird issues.

It feels weird true but it still also relevant faster?

I never liked telling people what to do or watching people do something. Using AI is exclusively like telling someone to do something who is either super smart or dumb as a brick and you never know which persona you will. The only thing you can be sure of is that it will produce whatever it produces very fast with 100% confidence. Also I care for code style, consistency and overall maintainability because I don't want to be locked into using LLMs and I can only really do that when I proof read everything. I don't want a huge project that will take a full rewrite for a human to understand it.

Even worse is having to telling people to do stuff who then turn around and use AI to do a poor job without the human rationalizing the problem or result.

I think the main problem is a conflict of objectives. AI companies need you to use more tokens so they are not going to improve that. They are optimizing the models to get as close as possible to the point in which people would stop using them because they are useless but without crossing that line.

Proof of this is the amount of unnecessary tasks that Claude does just as an excuse for not doing a good job doing the tasks that we ask it to do

Conflict of objectives is fine, it's not like we have a single provider of models vying for us to spend our money on. Something like collusion and price fixing across the industry would be a bit different.

Every provider has a primary incentive for users to consume more tokens.

Welcome to management.

I feel like humans don't switch between being an idiot and a genius quite as fast as LLMs do. But yeah I have been thinking about my future and I am not sure I am cut out for the role of manager.

Managing subordinates whom 1) Dou didn't hire 2) Are only there for cost reasons 3) You can't fire no matter how much they screw up.

> Maybe it's faster but explosive diarrhea is also a faster way to produce shit.

i'm in the midst of a very large global roll out of a new system including about 24 separate erp integrations. My team's system is dependent on another system coming online first then we go then the integrations come online then a massive data migration happens and ends with all the associated testing. So like a huge number of dependencies and I'm second in line. Today we get word that the first system is having issues with a vendor and will not come online as scheduled (our dry run deployment is tomorrow and spans all next week as each dependency gets their turn. Week after next is PROD.

We're not heavily integrated into that first system but we are in spots and so there's a reason why they go first. Writing around that system is feasible but we're looking around 5k lines which isn't a lot but there's still testing and sign off. My offshore team is in bed, meanwhile they, and onshore, all have other shit to do anyway so pulling them just puts everything else behind.

I tasked an agents to make the changes as soon as i woke up to the news, they got it mostly right. It took two rounds of fixes with some offshore testers who stayed up late but the PR got in and merged. We just did (literally like 5min ago) a validation run against our target system and we're green for tomorrow. I told the testers to goto bed and then i pulled up HN and opened a coke.

There's no way in hell that would have happened without the coding agents we had on tap. So, yeah, i find a lot of value in the rate at which AI can work on code even if it's not perfect every time.

I don't want to deny that. I mean I am using it. And if works well it's great. If it doesn't though it's exhausting. And if it produces bad code, it can do that at a speed that is overwhelming.

It's great it can do that. But it in no way justifies the social and environmental costs of AI, which is the point of the article.

True, it would not have happened back in the day. But oftentimes what would happen is that sane, sensible leadership would have a talk about delays on the critical path and adjust the schedule accordingly.

Maybe not as satisfying to people who enjoy being heroes. But it would not have been an automatic disaster.

“Look at how much more busywork I can do for the same salary!”

Naw, you've missed a critical detail

> My offshore team

This is partially driven by the good old enterprise game of not being at fault when things go wrong and being able to communicate that to the stakeholders

It's not fancy but it's necessary in a lot of enterprises

> “…talk about delays on the critical path and adjust the schedule accordingly.”

And self-imposed deadlines are the worst—not saying that’s the case here. And when your leadership sets the goals, you do your best. But fast, good, cheap; pick two.

And if the deadlines really are hard, then you should be actively managing the critical path from the get-go. The very existence of upstream critical path dependencies that, as the parent poster’s story demonstrates, weren’t even necessary in the first place. Which suggests that this heroic agentic quick fix was the proverbial pound of cure.

I’d even go so far as to speculate that perhaps agentic coding also steered people away from the ounce of prevention in the first place. Humans aren’t perfect either, but over the past couple years I’ve had to spend increasingly much time asking, “Why again are we creating all this unnecessary complexity?”

Ah yes complex global rollout being forced to production on an arbitrary rapid deadline is exactly the time when more code of unknown quality should be added to the system at the last minute.

I review all the agentic code and commit it myself. I also write some features manually, especially when I want to really make sure that I don't screw up. E.g. for authentication or for observability.

I also found that slowing down your development (just by a bit!) on purpose helps. Do not rush into implementing your new idea and give yourself at least a couple of days to think it through.

> herd a group of toddlers

It's called a fancy word "steering". Most of engineering is now steering or providing "taste" to AIs so they don't produce slop

Surely there's some point at which you can take over, and they'll have saved you time while not causing too much agony?

Maybe it’s a personal thing but I found that at some size it would be like having someone chuck a big foreign codebase in your lap and saying, “Ok, I did most of it, now you make it work.”

I kinda have to stay engaged the whole time the best I can.

What that boils down to is, “go ahead and yolo small things I can fully review” or “really wrangle it close through a larger effort” or “go full vibecode and prayer.”

Exactly this!

This doesn't seem like a well-thought-out post. I mean, I probably could have told you that from the fact that it's either pro-AI or anti-AI (in this case the latter), but this is a particularly poorly thought out post. Others here have pointed out the water-wasting red herring, but there are several other pieces of nonsense here — the general environmental doom-mongering, and labeling recursive self-improvement as "mystical", for example.

Me, I pointed Claude Opus 5 at a C codebase I've been working on for years and it immediately found five serious bugs and told me how to fix them, as well as how to use Linux system calls I wasn't familiar with to solve some other problems. And I gave it a new language for PEGs I'd written up a couple of years ago, and it wrote me a working implementation in OCaml that afternoon, fixing several bugs in my example grammars along the way and resolving a conceptual problem I'd had a roadblock on. I asked Fable 5.1 to implement a minimal proof assistant, and it wrote a clone of J-Bob in Python, which I'm studying now. Although there is a certain firehose quality to this stuff, I sure don't intend to use AI to actively de-skill my brain.

That doesn't guarantee that AI will be a beneficial innovation overall, of course! As with most innovations, it probably depends on the balance between people being able to use the innovation to gain control over their own lives, and people being able to use it to gain control over others' lives. (Barring a FOOM scenario, of course, where our prediction ability is nonexistent.)

So, I don't think a pro-AI post can be well thought out, either. It's too early and chaotic to understand what's going to happen, and it may actually be uncertain. Pro and anti are both far too simplistic.

Not criticizing or calling into question the contents of the posts or your rebuttal. Just wanted to point out this is not a Normal HN post or blog. Most blogs have the self removed to some degree or ultimately focus out of the self. This is clearly a journalistic blog, a shared diary entry sort of prose. You can see it in the first paragraph. It is basically intended to be emotive or get across the idea of getting emotions out about a topic, and ironically this is what makes it a not-bad-post.

Basically, I don’t know how well thought out it is, but the prose argues that the point isn’t for it to be an objective take, and that if it were well thought out (ie highly iterated on and more objective or factual than subjective) some personal meaning that was intended to be expressed is lost along the way.

In other words It may have not appealed to you but that’s fine - you might just be outside the target demographic. In any case I don’t think it necessarily should be more well thought out. I prefer that these pieces exist or have their place. I definitely prefer their existence to a world where every personal Blog post has to arise from fiduciary or technical concerns, instead of from the soul. This is a soulful blog. I like it.

It just seems like a lot of hate resulting from dysfunctional responses to life circumstances — admittedly, probably pretty bad life circumstances, from the sound of it. Spreading hate online, especially accompanied by lies as in this case, doesn't seem like a socially positive action from my point of view. The author should have kept it to themself and maybe talked to someone who could give them a better perspective, rather than letting their hate out into the world to multiply.

Its on the front page, there is some value.

Being on the front page means it's engaging. It doesn't mean it's valuable. The actual value of many front-page posts is negative.

Amen. I read it because it was shorter than the Effective Altruism community stuff I'd been trying to wade through. It did seem a bit entertaining, but I found the ending a disappointment.

You ought to apply your own standards to your own comments.

The OP is an article that:

1. Raises a number of broad social and ethical concerns about AI. 2. Provides links to back up some of these claims.

Yes, it also vents some frustration and includes personal anecdote, but there's nothing wrong with that. If you think something is a net negative for society, you should be frustrated. That doesn't make an argument "unreasonable" or "not well thought out". In fact, your entire response to the article is just anecdote and broad claims with no evidence or citations to back them up.

If the article author has "a lot of hate resulting from dysfunctional responses to life circumstances" you must have a lot of naivety and kindness as a dysfunctional response to life's circumstances.

If we do not critique things we perceive to be negative developments in society, things won't get better, and we'll let bad decisions slide. A somewhat emotive critique is not "hate". The author at least backed up their points with a few links, which is much more than we can say for your comments.

Hacker News makes me so sad sometimes.

What you perceive as naïveté is a kind of world-weariness that goes deeper than you can possibly imagine. It turns out that kindness is not in fact naïve. It is the only reasonable response to the human condition, which is far more desperate than you know.

Emotive critiques can easily be hate. Hate is an emotion, so hate-filled critiques like this one are quite emotive.

The historical record of people improving society by critiquing things they perceive to be negative developments in it is not really that great. Yes, that's how we got, for example, the abolition of slavery — but the anti-abolitionists justified their own positions by critiquing the genocide in Haïti, which was itself justified by people critiquing things they perceived to be negative developments in it. For every Civil Rights Movement we have a Great Leap Forward; for every Sherman Antitrust Act we have a Prohibition.

You may not be able to avoid activism, but never forget that other activists spent their lives digging the ditches you're filling up, and others will spend their lives digging them again.

The way out of this is to spend as much of your energy as possible creating positive-sum things rather than in zero-sum struggle.

Yes, everyone who disagrees with you is unhealthy, okay.

What do you mean by "firehose quality to this stuff"?

Quick to produce, emerges in vast quantities, and getting hit with it can consume all your attention immediately

Well, for example, in one afternoon, I have a parsing engine to understand, which is 9 pages of code. I thought I knew OCaml, having written a few interpreters in it, but I didn't know the {foo| |foo} multiline string syntax, I didn't know about Printf.ksprintf, I didn't know about Buffer.create, I didn't know about String.escaped, I must have once known about List.exists and List.rev_append and forgotten, I'd never thought of pattern-matching Array.to_list Sys.argv as a way to parse a command line, I didn't know about Sys.file_exists, and now I also have to think about what the parse result should look like and how to use it. And Claude correctly pointed out that the way I had thought about handling things like heredocs (or {foo| |foo} strings) would violate the preconditions for using the Packrat parsing algorithm, and suggested that I use the alternative approach of parameterized productions as in OMeta, which I need to think over.

I think that was the same day I prototyped an approach to using Service Workers from localhost as a platform for persistent mobile apps that don't show up in the app list but also don't rely on the TLS CA system for their security. Which involved learning a lot of stuff about the Service Worker lifecycle and localStorage and IndexedDB that I didn't know before.

It's just a lot.

I love how well this comment captures the experience of engaging with LLMs as a way to deliberately challenge and expand knowledge.

When I see people worry about LLMs atrophying their mental abilities I assume they haven't thought very hard about ways to lean into using them to expand their mental capabilities.

I'm guessing you've written a well-thought-out, clearly expressed blog post about this that I would absolutely love to read, if only I knew where it was.

I've been formulating this idea for a while now, but I don't think I've written it up properly anyway. I should do that!

I hope you do, because I'd love to read it!

Yes! Please do.

I assume that they don’t have time to do that, and are lamenting the fact that exercising those capabilities is no longer on the critical path to getting paid work done.

Although I could be wrong about this, I suspect you still need them, for the reasons I've described in this thread.

I have been prototyping some network proxies in a largely "vibe coded" fashion and it has been a mixed experience.

It has definitely accelerated my learning: through the process of writing rigorous plans, I have learned so much about network programming... the intricacies of HTTP2, how to design a fast MITM proxy, performance pitfalls in SSL...

These are things I studiously avoided learning for a decade, now demystified, mostly through Claude patiently rendering many figures and diagrams.

And with Fable, I was able to ship a working Envoy Rust extension in a one-shot.

On the other hand, my grasp on the details is tenuous. Every so often the agent can't progress and I am stuck for ideas, hitting different prompts like they were slot machines. I forget how to solve problems by myself.

I can feel myself becoming supine and dependent on something I know is going to become more expensive, a dependency AI vendors will seek to exploit. It has happened quickly. I do not really consent; it is forced on me by my employer.

One thing I may try this week is typing in all the code myself. With the level of back and forth that I've been happy with so far, Claude Pro seems to only be able to produce about 1000 lines of code a day in two or three five-hour sessions. I can totally type 1000 lines of code in a couple of hours if no problem-solving is needed. So maybe I'll look at Claude's code, switch screens to my editor, and type until I don't know what to type, and repeat.

Obviously there are higher-level design issues that I might not notice while engaging with the code at the byte-by-byte level, but I won't have the embarrassing problem I had a couple of weeks ago where duck.ai wrote me a CPU in Verilog, and after I tested it, I tried to write something in Verilog myself and couldn't even figure out how to write a syntactically correct XOR-gate module. (Verilator's parse errors are not designed as a Verilog tutorial, it turns out.)

An interesting idea, like crib notes. Forces me to pay attention, and exercises the mechanical memory of programming. I'll try it.

Agreed. And typing it in takes much less time than almost anything done during coding/dev.

This hasn't worked out that well for me. I have to try to do it myself, do something wrong, and ask for an explanation of what i broke. Just typing doesn't leave me with any retention.

Yes! The best part is, at any point they are right there to answer any questions, or to challenge or validate and experiment with any ideas! You can go down fairly deep rabbit holes in either case to understand a topic, and -- unlike any other medium other than a live human expert -- you can hone in only on the areas that matter to you most. Or you can just ask it to do something (as long as it's in the digital realm) and see for yourself what happens.

I always say that since one of the best ways of learning is by doing and given that with LLMs we can do so much so quickly, we are in a golden age of learning for those who really want to learn.

Water-wasting is not a red herring if you live in an area where droughts are an issue. It just isn't a problem everywhere universally.

You’d have to live there, in the presence of a data centre for AI which actually wastes water. The numbers people throw around on social media are absurd. I’m pretty sure they do not take into account the circular reuse of water for cooling, and just look at flow rates passing over chips.

Like golf courses in the desert.

(... Palm Springs has around 100 golf courses)

While we're talking about Americans using anything but the metric system (https://news.ycombinator.com/item?id=49746745), how many golf courses' worth of water is Google's 10.9 billion "gallons" a year?

ChatGPT tells me

> As a historical U.S. benchmark, an average 18-hole facility had about 80.7 irrigated turf acres and used approximately 152.5 acre-feet/year—about 49.7 million gallons (188 million litres) annually.

So about 220.

But it's still, ahem, a drop in the bucket compared to the California almond industry.

A while back I satirically suggested https://simonwillison.net/2026/Jul/17/spot-birds-not-golf/ that if hyperscalers wanted to relieve the pressure over data center water use they could buy up a few dozen country clubs and turn them into public parks.

(I was only half joking)

I am sure many smokers in the 1950s also saw the health benefits of tobacco, that they were very happy to experience the increased energy after a smoke and could get more done as a result. I‘m pretty sure none of them felt their lungs and throats deteriorating either.

Likewise, you may not feel like AI is de-skilling your brain. But it probably is. We don‘t have the research (yet) that AI is addictive and causes brain atrophy, but all the evidence is certainly pointing that way, the same way the evidence were pointing that way about tobacco in the 1950s. My only hope is that it doesn’t take us another 2 decades to prove the obvious.

Oh, I definitely feel a strong pull toward the low-effort path of just letting the AI do everything. It's so much better than I am at writing code, after all! This code it wrote for me is beautiful! But it makes terrible mistakes constantly, and when I'm reduced to pleading with it to check whether it's made mistakes, basically there is no hope. And, of course, it would put me in the Homer Simpson position of declining to get up and walk just because there's a motorized wheelchair at hand.

>Likewise, you may not feel like AI is de-skilling your brain. But it probably is.

It is, but why is that a problem? Moving from physical labor on the farm, factory or construction site to sedentary work and lifestyles would also atrophy our muscles so now we pursue physical activities we actually enjoy like gym and sports instead of hanging on to those dangerous and strenuous physical labor jobs just to stay in shape.

Similarly, you will use the AI to churn through the repetitive mundane tasks the AI can do, and then work your brain on the activities you actually enjoy.

Now if you choose to not exercise your brain and dumb yourself down the same way sedentary people don't exercise their bodies and get obese, then I guess that's on you, but we can't force the world to technologically regress just because some people are lazy and don't exercise, the same way we haven't forced construction workers to use pickaxes instead of machinery to tear up a road just to make sure they stay in shape and don't get too fat from sitting in their air conditioned excavators.

> It is, but why is that a problem?

I used to have several good answers, but now I use Claude for everything, I can't remember any

>then work your brain on the activities you actually enjoy.

1. As seen in this coountry, most people did not in fact "pursue physical activities they enjoyed". They got much less healthy and then the white collar machines started to churn their brains too. I do not see this ending any differently.

2. As many have noted (and has been known for decades pre-AI), getting more work done in a modern workplace does not reward you with more of your time back. It gives you more work and higher velocity plans. It may even end up compensating you less. There's only more benefit to this if you are your own boss, and so far a lot of those kinds of people are simply selling grifts, not truly disrupting the industry with newfound efficiencies.

>but we can't force the world to technologically regress just because some people are lazy and don't exercise

I didn't want to go too off topic, but since this is such a sticking point I need to make a big note about the US: most of the world is more sedentary than 100 years ago. But the obesity epidemic is mostly US-cerntric.

Which shows that, while we can't force a diet, we can definitely regulate incentives. But the US chose not to. We had all kinds of initatives to try and regulate what we put in our food and what to subsidize. We instead said corporations can do whatever we want and subsidized chemicals that are heavily taxes or banned in many other western countries.

I know that's a scary word for this community, but this analog of regulations can work here in AI as well. We don't need to ban AI (but I'm not opposed to it in primary schools), but we can shape policy to prevent "corporations can do whatever they want" yet again.

>we can definitely regulate incentives.

We can't regulate incentives. Unless we can somehow go back to how the world was in the 1960s, the incentive today in a global competitive world economy, is to make more money than your economic rivals, otherwise if you only regulate yourself but if they don't, then they will economically dominate you in the long run and turn your government into their puppet, your country into their playground and your people in their sweatshop workers. It's economic neo-colonilism 2.0 electric boogaloo.

Regulations only work if all players in the world adhere to them, but history has proven they don't. They expect everyone else to adhere to the rules and slow down so that they can get ahead of you.

> But the US chose not to

Smart. Then US can be the survivor in this global race. As someone in the EU I am aware of more top EU-born tech companies relocating their offices/R&D to the US than vice versa.

> but we can shape policy to prevent "corporations can do whatever they want" yet again.

I fully agree. But again, this will work if all major nations regulate AI equally, otherwise, AI corporations will move where they have more freedom.

>We can't regulate incentives.

We call tax subsidies "tax incentives" for a reason. You subsidize high-fructose corn syrup, you incentivize society to include cheap oils in their food.

>the incentive today in a global competitive world economy, is to make more money than your economic rivals

On a domestic level this is irrelevant. It doesn't matter how cheaply China makes corn syrup if there's still a cost of importing it into your country. And if it's still a problem, you add tariffs for more (dis)incentive. That was the original way they were supposed to be implemented, not as this massive general tax on doing any business with a country.

And if it's still a problem despite that, direct regulations and bans from usage in market consumption products is a last option. We can , do, and have done all of these for hundreds of ingredients.

>Then US can be the survivor in this global race

I'm talking specifically about food here, you know? You were so focused on the obesity epidemic that I wanted to highlight the actual underpinnings of it instead of leave it at "the world got lazier and fatter, that's progress".

Clearly tariffs don't so much to software. AI needs different solutions.

>AI corporations will move where they have more freedom.

If AI is a societally damaging tool, then I see this as a win. Let them move, let the US economy correct back to rationalism again, and let's improve society instead of CEO's bottom lines.

I also do believe in the invisible hand. If big tech doesn't like potential US regulations, smaller startups will happily compete for that role in the country.

> I didn't want to go too off topic, but since this is such a sticking point I need to make a big note about the US: most of the world is more sedentary than 100 years ago. But the obesity epidemic is mostly US-cerntric.

Note: nearly all of the western hemisphere has obesity rates of >25%. Obesity is also high in the Middle East and Australia/NZ.

The US, with an obesity rate of >40%, is exceptionally obese, but it's not correct to say that obesity is mainly a US problem. It's quickly becoming a worldwide problem.

https://en.wikipedia.org/wiki/Epidemiology_of_obesity#/media...

Agreed. I haven't written a line of code since last November when AI had the big shift into something good.

The average HNer would imagine me scrolling TikTok while I periodically swivel to Claude to type "continue" while my brain shrivels away.

That I can delegate lower level work like implementation means I spend my time at increasingly high leverage positions like deciding what gets built. And any time I want, I can get as granular and inquisitive as I'd like about the lower level work. After all, I probably chose it from a multiple choice AskUserQuestion tool call, each with a blurb and shortlist of trade-offs.

All this talk of de-skilling doesn't make a lot of sense to me because what are these skills people covet so much that defined most of the day to day of a software developer? What, the implementation details of some framework or the intimate details of your company's crappy legacy software stack?

Dunno, I have bigger plans for myself than that.

That may very well be true, but as you're describing it, I think of all my fellow engineers, all of them very talented, who moved up into management, arguing exactly the same point of thinking at higher abstraction levels. And after one or two years go by, they will come to me with requests or opinions that are literally the dumbest shit I've ever heard. Like arguing that we (holding most of our data on MariaDB and Postgres) should move everything into Kubernetes custom resources because "the Kubernetes resource model is the best database ever". That's a real thing that I've had an engineering lead say to me. When I asked how you're gonna do an efficient table join on Kubernetes resources, I just got a blank stare.

De-skilling is too broad though. Or maybe not broad enough. I've been using GPS sat nav since it came out, in the Tomtom days. I don't know how to get around to places without a gps to spoon-feed me instructions. My ability to multiply two digit numbers in my head has totally atrophied since my younger years. I also don't know how to ride a horse or grow corn though. Or how to milk a cow or goat. Yet somehow I'm still (for now) able to put food on the table. I can tell you subtle differences between the output of Fable vs Opus, or Sol and Astra. I'm real good at Linux, and it made my career, but honestly Claude is better at it than I am, and it's relentless. I get tired and distracted easily. When people lose one of their senses, the other ones sharpen. Maybe. I have a couple of practices I do to sharpen my mind on memory and recall and cognition. You're right on the brain atrophy, but like the gym replaced manual labor, we should do the same for our minds.

The de-skilling of AI includes your ability to express your thoughts, to communicate your ideas, to come up with new ideas, and creative solutions, to converse with other people. This is an entirely different category then the ability to navigate on highways and city streets.

Also worth noting is that your GPS has probably not de-skilled your navigation abilities when you walk or bike in your local area.

One thing I've learned in life is to accept that I'm in many ways an outlier, which means that even if the research is done, it frequently doesn't apply to me anyway.

Like I wouldn't doubt that research could show that more time spent on youtube is inversely correlated with learning. But if you just use it to watch lectures or tutorials for some skill, obviously you will be an outlier.

Or when "research shows that dieting doesn't work," obviously the result there is "research shows that most people don't maintain a healthy diet," and it shouldn't dissuade you from doing so. Choosing to do so anyway is what makes you the outlier.

And so it will likely go with AI's effect on the brain. Most people will likely use it to avoid thinking, and the research will show it to be a disaster. A small group will use (or already is using) it to challenge them and expose them to new ideas. You can choose to be in the latter group if you'd like.

>You can choose to be in the latter group if you'd like.

And now you inherit the burden of proving you are "one of the good ones" after a mass amount of disappointment from the normal use case. I don't think the "it's okay, I'm different" methodology ever works out unless you can prove you are so utterly beyond the competition that you can't be ignored.

But an outlier is just that. And as we see here, outliers cannot change society single-handedly. You at least need to be an outlier good enough to convince a lot of "normal" people that you are to their benefit and be willing to lead those "normal" people.

You don't need to prove anything or lead anyone or change society. You have a tool in front of you that can help you achieve your goals or distract you from them. You decide how you want to use it. Just do your thing.

> It is easy in the world to live after the world's opinion; it is easy in solitude to live after our own; but the great man is he who in the midst of the crowd keeps with perfect sweetness the independence of solitude.

Ralph Waldo Emerson

So you're just saying "it's okay I got mine" in that case, and don't care if the rest of society rots because "I can use it well". Fair enough.

And I assume you never need to worry about applying to a W2 job with that mentality

I mean sure I care. I'd love for my kids to grow into a world where everyone around them is a super jacked highly competent and motivated ubermensch, but what are you supposed to do? How are you supposed to get people to not let themselves rot other than to remind them that it's an option, and all they need to do is to decide to take it. If anyone doesn't care, it's the person who's resigned to the inevitability of rot even in the face of someone telling them they're looking at a tool for self-improvement.

>I'd love for my kids to grow into a world where everyone around them is a super jacked highly competent and motivated ubermensch, but what are you supposed to do?

I just gave an example. Use your power to lead and improve your community, and society as a whole. Use your leadership to add guardrails so the tools assist the user, as opposed to the current state of unchecked capitalism that simply extracts funds with no care for results.

You dismissed it and said to focus in yourself. Which is exactly how we get to such a situation.

This contradictory feeling of powerlessness while simultaneously striving to use such tools to influence the community simply shows why the wider population is gaining mistrust for the tech. Telling people to be better people because ot works for you has never ended well.

I influence where I can, and I have encouraged my team at work to use models as learning tools. Probe into their claims. Ask them to prove things with a quick demo/test/PoC. etc. They have tons of embedded knowledge that you can dig out. But running for president of the world so I can demand guardrails is a little out of my scope.

Reminder that this reply chain was about someone saying how they use LLMs for personal growth, and someone replying that they're deceiving themselves. You don't have to get hung up on how to fix society (who were already happy to rot with TV, instagram, tiktok, etc.) and keep that from seeing a useful tool in front of you.

> One thing I've learned in life is to accept that I'm in many ways an outlier, which means that even if the research is done, it frequently doesn't apply to me anyway.

This is a complete logical fallacy. It is quite literally mathematically impossible for you to not be "typical" for at least some aspects of your life. If you weren't, you'd probably be dead, or immortal.

Sure, I expect that if you gave me the LD99 of some poison, I will die. But there are still many things (and my point is this is one of them) where all it takes to be an outlier is to decide to be one.

I am pretty sure that likewise in the 1950s many smokers did indeed decided to be the outlier who smokes occasionally, and for the right reasons, and using the “correct” method/brand of cigarettes. And that every one of them decided that they were not gonna get addicted to nicotine. I am also sure that every single one of them did indeed get addicted to nicotine, and many of them died from cancer.

It sounds like you aren't very familiar with the research on tobacco addiction.

https://journals.sagepub.com/doi/10.1177/002204260203200220

https://www.tobaccoinaustralia.org.au/chapter-3-health-effec...

The majority of smokers did get addicted to nicotine, but significant percentages did not. However, it is true that even many "non-daily" smokers died from cancer as a result of their smoking.

I know that research on tobacco is full of bad science and bad actors, and experts in different fields (such as social psychology) continued to publish bad science on the topic well into the 2000s. This includes your 2002 German paper (which I can’t read because of a paywall).

Your second source says nothing about addiction in the chapter you posted, however they do talk about addiction in chapter 6, and they claim that most people agree that smoking is addictive, but heavy smoking is more addictive.

Still, (chiming in), your claim that " I am also sure that every single one of them did indeed get addicted to nicotine, and many of them died from cancer" could use some evidence to back it up.

My understanding is that only about 35% of those who ever try nicotine become addicted.

If you are going to instead say that nearly everyone who smokes occasionally will become addicted, you need to back it up with evidence, because it's not apparent.

I am speaking general. I don‘t have historical sources either that people spoke like that in the 1950s, and that is a much more glaring omission of citation. If I am wrong about that my whole argument falls apart. But if say 65% of regular tobacco users who use some strategy or otherwise believe the addictive properties of nicotine won‘t apply to them for some reason, then my argument still stands.

I don‘t know the figure here, but if it is true that 35% of those who try tobacco get addicted, and given that we know nicotine is highly addictive, I would guess that most regular users do indeed get addicted (and so would most people).

A better analogy is probably TV. You could watch the Flintstones (along with smoking ads!), or e.g. Mr. Wizard.

I am specifically talking about this cognitive bias of believing that the mean behavior does not apply to you. Very few people believe this about cigarettes anymore because the evidence is simply overwhelming and public health experts along with anti-smoking activists were able to (despite incredible odds) shift the common rhetoric closer to what is true.

There is a similar phenomena going on with ads. Most people believe they are unaffected by ads, but in reality they absolutely are. And the advertising industry has vested interest in the general public believing this illusion (otherwise they risk being regulated).

We are seeing the exact same phenomena with AI usage as we did with smoking in the 1950s. All the evidence is pointing towards these products being bad for you and being addictive. Yet most users of AI believe they are not addicted, and that they have found a way to use AI in a way which is less harmful for them. This is bias, these users are wrong, so are you, and so is OP. This is addictive to you, and it is bad for you.

I know research is lacking, but eventually we will have the evidence to show the case as clearly as we do today with tobacco.

On the other hand there's the consensus bias where you believe that everyone else is like you, when in fact they are not (which is why I said this is something I've had to learn to accept).

Sure, for most distributions, most people will be statistically normal, by definition. On the other hand, I know that e.g. state standard test scores when I was a kid said I was 99th percentile academically (and there was some kind of IQ test for the gifted program? I remember I qualified in all 3 areas they tested me), so I have some evidence to hypothesize that I might be an outlier along other psychometrics too. And I'll go ahead and guess that many people here are also not the median in many ways.

I don't think that this makes me somehow immune to lung cancer from smoking, but I might suspect that when I read about others' use of AI and how it rots their minds, they're perhaps not using it to learn Lean (the process of learning has mostly been the AI spits out crap proofs that seem way more complicated than they ought to be, and then I spend some time trying to find ways to simplify it. We have a back-and-forth about different encodings, or ideas for syntax macros, etc. Then I can compare with e.g. Mathlib's solution) or checking their understanding as they work through some notes. I don't know what most people use it for or what they get addicted to, but I believe it might be different from what I do.

With regards to ads, I live somewhere where outdoor advertising is banned, I block ads on my computer, I don't watch TV or movies or "news," and I don't install whatever apps, so there aren't a whole lot of channels for ads to reach me. I suspect again that this is not the typical experience. That said I do view ads as fraud-adjacent and damaging to markets, so fully support strongly regulating or outright banning paid advertising.

This comes across as an athlete excusing their smoking behavior because their superior physical abilities makes them immune to the bad effects of smoking.

I assure you a 99th percentile in 100 m dash may have believed that smoking did not affect them the same way it does an average person, but they would be wrong. And so are you about your AI usage.

You keep assuming that LLM usage is automatically harmful with literally zero reason to believe that, and several anecdotes from people on how they use it positively.

There's simply no reason to even draw a weak analogy to smoking. You'd do much better to draw an analogy to e.g. methylphenidate or amphetamine, which certainly you can abuse and get addicted to, and might even be harmful for the average person (even carrying the possibility of psychosis!), but which carry cognitive benefits with minimal side effects for others.

I'll give another concrete use: the other day I scanned a storybook with my phone and had codex OCR it and compare with a list I've started of words and sound patterns my kid knows how to read so that it could identify irregular words or new sounds for us to go over before we start the story.

Please, concretely, let me know what the adverse effects were there.

I keep saying it because all the evidence (while still not conclusive) is pointing in that direction, just like they were with tobacco in the 1950s (It took another two decades for the evidence of tobacco to become conclusive; and two decades still for regulators to catch up).

We do know that LLMs can cause psychosis, we do know it can cause people to behave in an unhealthy manner, etc. We are seeing more and more evidence that AI usages in education is at least correlated with bad grades. We are seeing people exhibit addictive behavior around AI. There is plenty of reasons to suspect that as we gather more evidence that they will become conclusive about the addictiveness and the unhealthyness of these products. For me this is obvious.

Evidence doesn't point in that direction though, even if we establish correlations.

Like the education effect is pretty obvious: most people will use it to cheat on their homework and avoid learning. Some people will use it to learn new topics or with more depth, and give them additional targeted probing questions to answer (basically, give them more homework). Average achievement will probably drop. The high achievers will probably do even better.

Boiling it down to "it harms your brain" when people are already telling you how to use it to learn more effectively is just silly. Most people are happy to rot their brains. They'll take any opportunity to do it. If you're an outlier, you're probably aware of that fact on some level, and it doesn't make sense for you to make decisions looking at the median who behave completely differently. You have to look at the mechanisms behind the data and interpret under what situations they could plausibly apply to you.

Like I said, the parallels to the 1950s tobacco usage is glaring. People used to say the exact same about cigarettes, that the way they specifically use cigarettes, the way their body is build, the way they control the habit, etc. etc. all leads to them being unique in some way or another making the habit healthy actually.

The evidence were all pointing towards tobacco being both addictive and unhealthy, but it still took two decades for the evidence to become clear and public health experts to be proven right. Maybe in two decades I will be proven wrong and you will be proven right, but I doubt it. When the evidence are all pointing in one direction, it is usually because they are right.

> We do know that LLMs can cause psychosis, we do know it can cause people to behave in an unhealthy manner, etc. We are seeing more and more evidence that AI usages in education is at least correlated with bad grades. We are seeing people exhibit addictive behavior around AI.

Aren't most of these true of amphetamines/methylphenidate also, though? Except the "bad grades" part. Certainly Adderall is habit-forming for some people and has risks for abuse, just like it can contribute to mental breakdowns and psychosis.

Likewise, the psychosis-LLM link is incredibly rare.

It kinda looks like you're overstating the evidence. This is not a good comparison to 1950s tobacco -- by the 1950s, the scientific evidence for tobacco being unhealthy was already quite substantial. The problem was getting that information to the public in the midst of a disinformation campaign. Here, for contrast, the evidence that moderate LLM use in professional settings has overall negative impacts is still pretty light.

I only use LLMs rarely, and I don't have any kind of emotional ties to the issue. I just think your claims extend decently far beyond the evidence.

My impression was that the frontier AI coding models are kind of resetting the barriers to entry/reward ratios to before the dot com boom, when people didn't go into this career for money, and so most folks that ended up in it were the 1% kind of talents who really wanted it. Perhaps that's an elitist thing to say? I don't know. That's not to say that AI is pushing people away from computing, it's just that it's creating such massive low resistance paths to getting to the goal without the "productive friction."

p.s. and at the same time, it's such a fantastic learning tool. It distills the intuition of a whole world and is able to transmit it on demand, like in your examples. It's an interesting dichotomy.

Yeah thats the silver lining in the whole mess: For those of us who enjoy software, hardware and building things, this will be a new golden age and we can filter out all humans and elves.

What about the dwarfs?

I am a dwarf and I'm digging a hole.

I don't know, maybe? I think it's opened the doors to a lot of people who don't really want to program, but do want, for example, a studio appointment booking management system. Which is good; people's needs matter just as much if they don't enjoy programming. You can use it to create a pretty wide range of things, which seems like it's going to change the things that get created in hard-to-predict ways.

Structurally, though, I'm sort of pessimistic.

Throughout my career as a programmer, I've earned much more money than electronic engineers whose skills are mostly very similar, and probably much of this happy situation is due to the fact that my productivity depends only on me having access to a computer with GCC and Perl or Python on it, and the internet. If I were living in rural Malawi, I probably couldn't practice my trade, because the US$200 it takes to buy a working laptop is a year's wages there, and people lack the social permission to innovate. (William Kamkwamba's astounding achievements under those circumstances deserve more recognition. Read his autobiography!) But, if I work in any developed economy, that's somewhere between 1 hour and 20 hours of my earnings, so I'm not dependent on investors to create jobs for me.

By contrast, electrical engineers mostly can't ship their designs unless an investor is willing to put up the capital to produce a batch of PCBs or chips — probably three batches or so before you get a fully working product. If you're working on a cutting-edge chip, that's several million dollars, so you're probably not going to save that up while working as an employee. This puts electrical engineers in a much worse bargaining position relative to investors. To a smaller extent, the same thing applies to machinists, auto mechanics, and many other skilled workers: they only get jobs if investors invest to create jobs for them.

The capital-intensive nature of current AI seems like it might push the negotiation equilibrium for programmers two or three orders of magnitude in favor of investors.

But it might not work out that way, and, at least at the moment, I'm finally bringing about numerous projects I'd been procrastinating on for years.

> It distills the intuition of a whole world and is able to transmit it on demand, like in your examples. It's an interesting dichotomy.

The only problem I have at the moment is it's very prone to hallucinations, even now, yes frontier models astra fable with all the bells and whistles. This makes it hard to confidently use while learning because I have to be on the lookout for lies while I'm learning, which is precisely the moment I am least able to distinguish them. So instead theres just a constant low lying dread.

Nonetheless, I am able to get some value out of them. Just not all, everything requires manual effort to duplicate and check which you should probably be doing anyway as part of learning

I'm not doubting you, but what's the most recent hallucination you've come across? I haven't had one in ages.

I mentioned the minimal proof assistant I elicited last night. When Fable 5.1 thought it was done, I asked how we knew the prover was sound (which means, in the jargon, that the theorems that it proves are actually true). It checked and immediately found five different ways it could "prove" false theorems with it (and fixed them).

Then it suggested writing a fuzzer to try to flush out more soundness problems. It loves fuzzers! And usually they are an excellent cost/benefit tradeoff. However, in this case, I questioned whether a fuzzer would ever actually succeed at finding proofs of a theorem it was set to prove, even unsound proofs, and after doing some tests it admitted that the fuzzer it had proposed would have been completely useless.

Then the conversation was incorrectly flagged as me working on a malicious security attack, so I was downgraded to Opus 4.8. I switched back to Fable, renamed the file in its scratchpad, and asked it to please use the term "generative testing" instead of "fuzzing". Thanks to not using the computer-security name for generative testing, there were no further false flags.

Then, this morning, I was reading the spec it proved its program fulfilled, and asked whether a certain trivially incorrect alternative program would also fulfill the spec (a misformalization problem rather than a soundness problem). It churned away for a while, discovering that while, actually, no, that program would be rejected, a different trivially incorrect program would pass, and credited me in the docs with pointing out the issue. I pointed out that in fact the issue it had found was completely different from my stupid misreading of the spec. It fixed the doc.

Fable 5.1 isn't Mythos but it's generally considered to be a "frontier model".

So, from my point of view, the whole experience has been a constant fractal of hallucination, in which I have to constantly struggle to keep my grip (and Claude's grip) on actual reality, because it's so willing to make up surface-plausible nonsense.

______

P.S. Also, in another task today, it thought pip wasn't installed and was trying to figure out how to work around it. But that's not so much a hallucination as a failure to recheck assumptions — I hadn't installed pip on the machine before the first Claude work on it, and it just assumed that was still the case. Also I think that might have been Opus rather than Fable, so it's not as strong a case.

I feel like this pattern is a great way to practice and sharpen critical thinking. It's like a whole new skill to deal with computing systems in this way.

Not the GP, and this is just the free ChatGPT, not a frontier model, but just a few days ago it happily confabulated an entirely incorrect version of the plot of Iain M. Banks’s "Matter" when I asked it to analyze the novel in a certain context. Claude fared better, though it also made some mistakes. I don’t expect the models to know or recall the specifics of the plot of every novel out there, but it would be nice if they didn’t make things up.

Probably over twenty today in my work day alone. Its a normal part of working with agents. They do not always have the context they need, but fail to be aware of this.

Misinterpreting the results of performance testing it had just run to argue for the exact opposite of what the data it generated showed.

Stating something was current guidance (a quick read shows the document it found was from 2016, it was served the last edit date in it's API call, along with newer documents that contradict it).

Assuming what a Jira ticket said without ever reading it with it's tools and then doubling down on the contents it has assumed.

On learning specifically, it constantly gets grammar in foreign languages wrong it can write fluently when prompted correctly but when asked the sort of wrongheaded questions from flawed premises learners commonly make it is prone to making stuff up. I experienced this yesterday when asking about how danish comparisons work.

Heck i mean see https://alignment.openai.com/misalignment-reports/self-gener...

They are prone to make up restrictions for themselves you never asked for. Ive seen this behavior on occasion also.

lol i love playing with food labels. the things it spits out for anything de papa in an otherwise English paragraph is terrifying. ai must solve the spanglish potato problem. also the spanglish con issue.

Did you see the Jev model demo where played wiki racing against GPT, Claude and gpt hallucinated links in Wikipedia pages?

Are any of the projects you mentioned team-maintained and deployed live to users other than yourself?

No, thank God.

it's a lot more thought out than a lot of the code getting pushed to production right now

> I pointed Claude Opus 5 at a C codebase I've been working on for years and it immediately found five serious bugs

These 'serious' bugs aren't important if you've been working on this code base for years and only now they pop up. I suspect - although I'm guessing - that probably a hobby project.

Frankly speaking, although this comment is very dismissive about the opinion blog article - assertively stating that this isn't a well-thought-out post - it seems to me that the personal anekdote is irrelevant and that the blog post in question still fully stands.

It's definitely been a hobby project so far, but suddenly its quality has improved a great deal, which I think greatly improves its chances of graduating to being a daily driver.

> general environmental doom-mongering

I don't think this criticism is very well thought out. There's ample evidence at this point that we are sliding precipitously toward environmental catastrophe and, no, the massive data center build out to support people generating slop illustrations and build toy theorem provers is not helping.

We're not just sliding towards environmental catastrophe — we're deep into it! Ocean acidification is killing corals, we just had the hottest month in recorded history, mass refugee movements from climate disasters are fueling xenophobia and warfare in many parts of the world, and the rate of species extinction keeps increasing, as it has for decades. Extinct species are irreparable environmental damage; they're gone forever.

However, this is not because of massive data center buildouts. Goldman Sachs's projection in https://www.goldmansachs.com/insights/articles/us-data-cente... is that by the end of next year US data center power demand will have increased from 31GW to 66GW, roughly 17GW per year. The EIA says https://www.eia.gov/tools/faqs/faq.php?id=87&t=1 that the US uses about 95 quadrillion Btu per year (Americans will do anything to avoid using metric units). That's 3200 gigawatts. So the US's data centers are currently using 1% of the US's energy and adding ½% per year, so that in 02027 that number increases to 2%. 99% of the US's energy use is due to non-data-center causes.

And direct energy use is a way in which data centers have an especially large environmental impact. Their other potential impacts, such as cement production, sewage, parking lots, commuters driving to them, and freshwater consumption, are much less than that 1%.

Energy use, by the way, doesn't damage the environment. Burning fossil fuels does. All of China's net new energy production is coming from renewable energy — they're shutting down coal plants faster than they're building new ones, and they're decommissioning gasoline-powered vehicles faster than they're selling new ones. A friend of mine from the US told me that one of the most surprising things about visiting Shenzhen was how silent the downtown was, because almost all the cars are electric there — a much higher rate than for China overall, but even for the whole country, it's over 50% of new cars.

AI might turn out to be a wonderful thing or a terrible thing, but at least for the next decade or two, that won't be because of datacenter buildouts, which are an utterly insignificant fraction of its effect.

> So the US's data centers are currently using 1% of the US's energy and adding ½% per year, so that in 02027 that number increases to 2%.

1% of an entire nation's energy production for a single, non-essential use-case is a lot. For comparison, circa 2016, US farming, which we generally need to live, takes up 1.9% of energy production (https://www.ers.usda.gov/data-products/charts-of-note/87964). Hopefully this makes it more obvious that consuming 2% of all energy production, even though the raw magnitude is small, is a lot for a single, non-essential service currently controlled by a handful of companies.

> Energy use, by the way, doesn't damage the environment. Burning fossil fuels does.

I'm aware. The current reality we occupy in the US involves mostly burning fossil fuels to power these tools, and some plants deliberately skirt rulings around environmental safety. Management of resources and the economy is about trade-offs. Yes, in principle a certain activity may not depend on a bad trade, but if your only option right now is a bad tradeoff, it's "in theory" independence from this doesn't mean you should go ahead and do it.

> However, this is not because of massive data center buildouts.

It's because of a number of factors, obviously, but the data center buildouts do not help stem the bleeding, they (currently) exacerbate it. This idea of "well, it contributes to <bad thing> but less than <other bad things> therefore we should do it" is myopic and hopefully it's obvious to you why such a line of reasoning is not going to lead to any mitigation of any kind of crisis.

It's about trade-offs. If the overall ROI is worth the absurd costs, then, yes, it's all worth it. But "hey I (N=1) personally like this tool and get value out of it" does not justify global costs that contribute further to, as you said yourself, irreparable damage.

This isn't even touching on a host of other problems involved in the buildouts, namely making districts decidedly less attractive places to live and displacing current residents by allowing a handful of companies with massive amounts of power to directly disrupt the lives of citizens, for no direct recompense. (https://www.texastribune.org/2026/07/09/texas-data-centers-a...) (https://www.tpr.org/2026-09-17/residents-file-lawsuit-challe...)

> AI might turn out to be a wonderful thing or a terrible thing, but at least for the next decade or two, that won't be because of datacenter buildouts, which are an utterly insignificant fraction of its effect.

If the 2% energy consumption ends up being the final tipping point for total climate collapse, you might feel differently about this statement.

Sure, it's possible a certain 2% could be a tipping point, but seriously, look at US energy use as a whole, and it sounds unlikely.

https://en.wikipedia.org/wiki/Energy_in_the_United_States#Fi...

As far as sources go, reading further shows wind and solar still expanding. I don't think we can generalize anything from energy consumption of data centers.

Yes, that kind of indeterminate conclusion I can accept. What I cannot accept are people like the parent who want to ignore the very possibility of an issue or the role of this buildout in contributing entirely or who want to claim others are telling "lies" when this is patently not the case.

The amount of sophistry and absolute dogshit argumentation this forum invites now thanks to general boosterism is pathetic.

> A friend of mine from the US told me that one of the most surprising things about visiting Shenzhen was how silent the downtown was, because almost all the cars are electric there

Can that be true?

I've read at some point at which speed on average the sound of tires on tarmac will be louder than the sound of an engine, and it was a surprisingly low speed. I don't recall the exact number, but I remember that it was below the average in-city traffic speed (save stop-and-go averages).

This definitely matches my real-world experience. I can't usually make out engine sounds in passing traffic, because the tire noise is so freaking loud. I certainly couldn't blindly tell electric from ICE by sound of a passing vehicle at, say, 40 or 50 km/h.

Totally agree. It is more of a rant than reasoning. Anyone who claims to know how this will end (good or bad) is extrapolating beyond their own knowledge and is vastly oversimplifying what is likely to be one of the most transformative technologies of the last 100 years.

In the last 50 years, only the creation of the Internet compares. Like with the Internet, nuclear, television, etc. it will probably be a mix of good and bad.

I don't like the general dismissal of environmental concerns around LLM use as "doom-mongering". There was some miscalculation or misunderstanding around data center water usage once, and now this keeps being used as a supposed knock-out argument. Water consumption, however big of an issue it may or may not be, is just one of maybe a hundred ways in which LLM proliferation is amoral and anti-environmental.

The same bugs me about cars. Suddenly, driving cars is cool and modern and unproblematic again, because the market is switching to electric engines. As if fossil fuel consumption was the only thing environmentally problematic about building and driving cars! It's not, not by a long shot.

The environmental impact of LLM use also has to be considered in light of the wasteful ways in which they're being employed in general. Sharing and reusing has gone out of fashion. Millions and millions of LLM-assisted programmers, for instance, or now re-inventing and re-implementing the same things, in the most inefficient way, because that's more profitable to the oligarchs than actually collaborating and building on each other's work.

I’ve had a pet theory for a long time that society loses it’s collective mind every 10ish years or so, usually starting with a good reason, but the outcome is mass stupidity followed by mass forgetting leading to a new cycle. War on terror, GFC, crypto, covid, now AI. Although the cycle does seem to be shortening.

Also we’re just so damn stupid to learn from history, or even just people 10 years older than us, that we repeat the cycle.

And there I guess is where saying a new sucker is born every minute is from.

I have a more sympathetic view, I thought covid could have been much worse. Of course there were silly moments like the existential panic from a toilerpaper shortage but it seems like the vast majority of people acted responsibly and kept their heads over the long haul. In the meantime, we developed revolutionary vaccines that solved the problem.

There's the lingering societal PTSD to consider, but I think most of the challenge in the aftermath are latent issues that preexisted covid and it just brought them to the surface. I think our world a couple decades from now will be stronger than it would have been in a universe where covid never happened, and actually has a better chance at dealing with the next covid-level threat. Perhaps that is this, in a way.

Covid proved that a healthy looking society is not always so healthy, and I think it works in reverse too. The outcome of the upcoming election will tell us something about how far we have to go.

This might be Australia specific, but I think a lot of the movement restrictions were unnecessary. And long term damaging to not just adults but there’s a whole cohort of kids/teenagers who just missed out on certain things and we’re patting ourselves on the back we survived?

The places that needed restrictions such as hospitals & nursing homes still got covid by & large. We didn’t necessarily need to close down the whole show.

I’ve had my jabs, also had covid multiple times. I’m not a disbeliever, I just think the collective reaction was completely OTT.

Kids mostly forgot and adapted back. It is adult, mostly men, who just cant get over fear of vaccines, injury of having to do something a little uncomfortable for someone else.

Trivialising this to fear of vaccines misses the point. The lack of human communication being the new normal, cut friendships and shattered communities are the real killer here. Especially for ones that didn’t have a lot of social life or family to begin with.

Almost every single meetup group in Northern Virginia was completely destroyed and did not grow back. I watched as dozens of local meetups, several running for many years, disappeared over night. I managed one as well and my crew held on to ours for as long as we could. We managed to keep it running online for several more years. But the social damage was done. After running the thing for 12 years, not being able to find anywhere that was willing to host us in-person again, and all of us original organizers having moved to different cities, there was nobody willing to take over.

And the massive damage this did to children's education. I think people who didn't have a child in the school system at the time might not understand how poorly the schools functioned during that time.

My child had two intelligent and supportive parents, including a stay-at-home mother, and it was still rough. Many of her classmates just flat-out lost a year of education. There's still obvious effects on her cohort years later.

The lack of human connection has been a trend for many decades now, covid just accelerated it all at once. Not like there was much choice, lots of people died, we did what we had to.

The communities that did exist were always running on fumes, basically only existing due to inertia. When the inertia was cut, even if only for a couple weeks in some places, they were done. That’s it.

The social damage has been here, and we’ve been zombie-ing our way through it. Covid just let the undead stay dead, and turns out we can’t revive it.

> Trivialising this to fear of vaccines misses the point.

We had concerted anti-vaccine campaign. Literally trying to make you feel afraid of vaccines or feel sorry for kids that might get them. And yes, it was fear spread by men who fancy themselves manly men.

> The lack of human communication being the new normal, cut friendships and shattered communities are the real killer here.

I am particularly salty about this one. Because the exact same ideological and political groups who are throwing complete temper tamptrum over lockdown are the ones who want women to be stay at home moms. I did not, lockdowns were easier, you had a lot more communication and everybody was in the same position so you could realate.

Lockdowns were an easy mode compare to the thing they want force women into.

I live in a relatively wealthy town and most people here had a pretty safe series of lockdowns (bit hair-raising for me as a freelancer but ehh). We are collectively massively pro-vaccine; I know only two people who loudly resisted (and one of them nearly died in hospital from COVID).

But I can tell you that our collective sense of (unusually strong) community has taken six years since the first lockdown (four and a half years since the end of mask restrictions) to truly return to what it was before.

So even in this quite cushioned place I think we didn't really return to a sense of proper ordinary community positivity for quite a long time (it helps that we had a pretty great summer). Last summer things still felt very disjointed.

Kids, young adults, I can assure you, resent what happened and how poorly we managed it.

In other towns where community was weak, a lot of damage has been done, and you can see that in how little collective spirit there is in dealing with new challenges. Even around where I live, things like reduced footfall and contribution shortages in community services have been used as a pretext to cut them entirely. We have done damage to young people by simply deleting community things we benefited from that they no longer will.

Of course kids and young adults resent it: a part of their childhood and formative years sucked ass.

But the resentment isn’t towards the handling, it’s towards the reality. It sucks covid happened and it shouldn’t have. But there’s nothing that could have been done to save their youth; it was already ruined for those years.

I mean, if the lockdown didn’t happen and instead a few million more people died, would they have preferred that? I doubt it, especially considering such a situation would inevitably result in even stricter restrictions going forward. After all, lockdown was a reaction. If you make the thing it’s reacting to even bigger, even more of a problem, then… well… the reaction will be more extreme.

> But the resentment isn’t towards the handling,

Oooooh yes it is. They are so mad at how badly we handled it. And they are right to be.

> I mean, if the lockdown didn’t happen

That isn't their complaint. Their complaint is not the first one (except that in our case in the UK it was idiotically weeks late and major sporting events happened; the older ones understand how stupid that was). It's the fact that we subsequently undermined it, and then did it again. And again. And again.

I kind of touched on this. Lockdown was a reaction. If it didn’t happen, then the thing it’s reacting to would have been worse, and then lockdown would have been worse.

No, but the point is not that lockdown happened.

Kids are not angry that there was a lockdown; they understand the science of it (in many cases better than adults) because it was a pretty rich topic of conversation in their tedious remote lessons.

They are angry that we fucked up the first lockdown (rolling it out too late after demonstrably obviously bad decisions like letting major sporting events continue, followed by utterly stupid measures that undermined it after it was lifted). They understand there had to be another winter lockdown; they thought we undermined that by having stupid, arbitrary "zones" when another national lockdown would have been safer.

The older ones also IMO pretty correctly feel resentful being stuck individually in their expensive rooms at universities when they are not a high risk population and could at least have been quarantined collectively or just provided distance learning without wasting massive amounts of money on accommodation.

Believe me when I say, young people have absolutely got the receipts. They know what happened, why it happened, and who fucked up.

Millions of people die because of covid in Australia? Is that what you’re suggesting the outcome would’ve been if we didn’t lockdown?

That’s basically all the elderly or an entire cohort, I’m really struggling to believe that.

Sorry, I’m saying this from a US perspective, where we had 1 million or so covid deaths. We also had lockdowns, but it was state dependent.

The point is: if the lockdown didn’t happen, would the outcome be better? Remember, it’s not like Covid would go away. It’s not like we get teleported to an alternate dimension in which everything is okay.

Lockdown was a reaction.

I think people would like for Covid to have never happened, and they think that if lockdown didn’t happen then it would be just like before Covid. No, it wouldn’t be, it would still be a new world. That much was inevitable.

That the path we took was the only path is where I draw issue. It was a choice to lockdown, I think it was a bad one. Covid killed even with lockdowns & vaccines. Birdflu is apparently ravaging Australian bird populations, just because it hasn't jumped species yet doesn't mean it won't... maybe we should lockdown just in case right?

I'm not in a position of power to make the calls as if something like this was to happen again, I'd only caution we look the prior experience to be a guide for what comes instead of pretending outbreaks haven't happened before.

We locked down AFTER people started dying, not before!

I'm saying that if did not lock down, then there would have been millions more deaths, and then an even more severe lockdown would occur. It's an inevitability, that's just how life and the world works sometimes. Sometimes bad things happen and they're just bad, and there's no way to make them good. Covid was one of those things. Covid was bad, a lot of people died, and that's that. There's no alternate reality in which Covid exists and everything is fine. That's a dream. A nice dream, but a dream.

In the U.S. we hardly had what you’d call “lockdowns”, more like government suggestions that you shouldn’t expose yourself or your loved ones to a potentially deadly virus. Restaurant goers, teachers, commuters, and others wanted to avoid sickness and death, so many organically avoided activities that would leave them exposed.

The US response to covid was much less locking down than Australia.

If you look at the "deaths by covid" graphs for 2020 and 2021, Australia barely rates a blip. Meanwhile, the US has 850 000 deaths in those first two years.

Kids here don't resent covid happened. If they resent it at your place, it is on adults. But of course it is not exactly the same as before, people adapted and carried some adaptation with them. I work a lot from home and I am not giving it up unless I really have to for example.

For years now, kids go to school as before, they go to do sports as before and those whose parents dont allow unlimited gaming meet each other in person.

> In other towns where community was weak, [...] things like reduced footfall and contribution shortages in community services have been used as a pretext to cut them entirely

That has nothing to do with covid or covid handling. This is about your community not wanting to pay community services regardless of covid.

> If they resent it at your place, it is on adults.

Isn’t that exactly what I said? I mean, who else would it be on?

> That has nothing to do with covid or covid handling. This is about your community not wanting to pay community services regardless of covid.

No, it’s the result of cynical regional government exploiting how long communities took to bounce back from a series of mismanaged lockdowns to claw back some money. But that is not the point. The point is it is another reason young people resent how adults handled COVID and direct consequences. These systems needed time to recover footfall. Instead adults fucked it up gor kids.

Ah yes, lets throw in casual sexism in with your victim blaming.

After being burned by the first wave, Australia locked down hard and early for subsequent waves. It was the correct thing to do.

As a result, they had relatively short periods of lockdown compared to other countries, such as the UK, which had comprehensive vaccination but dithered about going into lockdown.

IIRC Melbourne had the longest lockdown in the world.

The lockdown was not spread equally. There were enough people just ignoring it.

Some having midnight parties because it was extra exciting to subvert the curfew, with incidental spreading of COVID.

Some were outright trying to have "herd immunity" parties to deliberately spread it.

And hence the official lockdown kept having to be extended.

Fuck I hated those people.

For a city, yes - seven or eight on again / off again hard lockdowns with a total of 262 days(?) or so of stay in place isolation.

At the state level Western Australia continuously self isolated for something like 18 months with a 14 day quarantine period for anybody wanting to enter for longer than it takes to turn a truck around.

Very different types of lockdown for sure.

It did. That’s why I said Australia was burned at first and changed their policy.

(Actually the longest lockdown in Melbourne was the second one, but it was only a couple of weeks after the first, I misremembered it as being one continuous lockdown.)

Subsequent lockdowns were shorter than elsewhere in the world because they reacted faster.

That’s the thing about large scale public health interventions like that, if it works it will look like you overreacted and you still won’t know if it’s because it worked or if it was never as bad as was feared.

It's my understanding that there's some room to be skeptical about the value of the lockdowns after a certain point and costs are undeniable. But for a period of time hospitals were basically overrun and medical staff were pushed way past the limit. To do any less would have been utter disregard for human life and the people who were trying to keep us alive. There were huge outbreaks in specific areas because people decided to ignore it and have a wedding or a party. So many videos of rebellious people on social media only to follow up with another video a few weeks later gasping for breath in a hospital room.

Most people didn't do that, and the people on the front line held it together despite the insanity. With vaccines brought the opportunity to open up more safely, and it seems like many places found a responsible balance in doing that based on available hospital capacity. And it seems like others took a more cautious approach and it is a fair question whether it went too far.

At the time:

1. no one knew that much about COVID. Even the authorities and experts!

2. It was a respiratory thing kinda like a cold or flu that killed some people. And then more people. And of course the news hyped that. So many theories and crackpots and experts and armchair-experts spouting all kinds of shit. It was tiring to open a web browser.

3. No vaccine. No vaccine for at least a year. You just have to wait for this tidal wave of sickness to hit you and everyone around you.

4. I recall seeing on the news what was happening in Italy. March 2020, something like 600 people a day dying of covid. Italy's hospitals were being overrun with covid patients. It wasn't obvious who needed treatment and help to survive the night and who could just go home and be (quite) miserable for a bit. There were so many unknowns! Hospital overrun resulted in not just covid patients dying but also anyone else who needed to go to the emergency department was affected to some degree. Partly because the system didn't have the capacity and partly because you'd catch covid on top of the reason you were there. The systems were choking.

I was in Melbourne the whole time, through every lockdown. Everyone has an opinion on whether it was the right thing to do, but for me, it was the society agreed upon (via authority and literal epidemiologist expert guidance) way to avoid killing a single digit % of the population.

Don't get me wrong, I was so fucking sick of being couped up... but I couldn't look myself in the mirror if I had've put some of my neighbours through that first wave harsh covid. I couldn't say as to whether it would have killed some of them.

Yes, many people had it rough in so many different ways. Bureaucratic shit was being made up almost day to day and people were also trying to scam it day to day. So it isn't surprising shit happened. Everyone has a story.

However, to say "We should have.." belies the fact that hindsight is easy and recollection is rose tinted.

Thank you.

1. They still don’t, you just lived through an episode though so you feel good because you survived and it feels like you’re prepared.

2. Cool story.

3. Another cool story.

4. Yeah. People lost their minds because people got sick and gate crashed the ED for what’s to mostly normal person a bad flu.

You’re proving my point that people lost their minds. Our hospitals are more up the creek presently. I can’t emphasise that enough. All the billions upon billions of dollars spent during & since on R&D, staff, resources, beds. PRESENTLY - no covid outbreak - thank the lord.

> You’re proving my point that people lost their minds.

Nah, minds were not lost. Italy simply had the bad luck to show us what "zero response" looked like.

> mostly normal person a bad flu

Nope.

Italy now: flu season for the year = ~3000 deaths

2020 Italy COVID for the year = ~74000 deaths

Bit of a difference. And that was with Italy also trying lockdowns and everything else once they realised how bad it was.

These people always insist on literally ignoring the body count. Drives me insane. It was so far removed from a normal flu and everyone in the medical field will tell you that emphatically yet here people are acting like it was nothing.

I never said there was less deaths, don’t put words in my mouth. I said for most regular people it’s likely a bad flu.

Did you know Italy’s death rate is still elevated, even 5/6 years after? Best go into lockdown since we don’t know why.

Millions and millions of people died, you know.

The biggest damage from COVID will be that it serves as a dry-run for future techno-faschist governance. The way in which the masses were not only willing to give up their own liberties but also pressure their peers into following the outright unconstitutional "emergency" laws was truly shocking.

I appreciate your optimistic view of it. What I got out of it was that Covid was a sort of dress rehearsal for a far worse outbreak that could eventually happen, and we (Americans) failed miserably.

It was a dress rehearsal alright, but for techono-fascist control of the poulation.

Journalists fear mongering and making sure the hoi polloi know their place as powerless pawns in a game they couldn't possibly understand is a constant.

They know what they're doing, the people who fund them know it and the people with half a working brain cell sooner or later hear someone like me and go 'oh, that explains it'.

It's much simpler than some mythical cycles - manufactured 'mass stupidity' is a conscious, deliberate, ongoing choice.

It used to be religious organizations, then it was newspapers, now its social media platforms. The playbook has been the same for thousands of years - nothing has changed.

To save everyone some time - the playbook is you're cattle, I'm shepherd. I'm master, you're slave. It's not anymore complicated.

The bit 'educated' people miss is that most people would love nothing more than to be the master (freedom they call it? American doublespeak). They dislike being slaves, not the concept of slavery.

Anyhoo.

I agree with you but I think it happens every 4 years, not every 10.

I first started noticing this in the 90’s .. comp-sci grads were entering the industry with no clue whatsoever how to do software. (Of course its gotten worse since then, but I started noticing it in the 90’s..)

Everything old was thrown away as being archaic and obsolete - only the New Cool Shit™ mattered .. except nobody had ever done anything, properly, with the New Cool Shit™ - everything was still running on the old.

(Unix was dying, everyone was learning only Microsoft tech because “that was the future” ... until it wasn’t. Now, a majority of us are running 21st century Unix workstations given us .. by Apple, of all companies, which didn’t die so easily, after all..)

People entered the market not because they were truly interested in improving the world through computerization, but because they wanted the lifestyle afforded to them by a career in “Information Technology”, a nascent term back then which simply meant “career path for the budding materialists”. Trouble is, you can’t be materialistic with computers, they are almost entirely a spiritual device, which value is determined strictly and only on the decision of value made by humans. It is humans which decide the true value of all that physical hardware after all.

I think this cyclical amnesia is driven by materialist consumerism. People who take a less material approach to technology, tend to build stuff that lasts a long time.

The reason I think it happens in cycles of 4 years, is that is the average time it takes someone to enter the academic churn, and get spit out on the other end, ready to join the work force and do big things.

Or .. not.

  > amnesia is driven by materialist consumerism

anecdotally, but i see this everywhere not just in tech...

> human resources are still a zero-sum game, and every individual feeling super busy in their agentic silo doing a thousand things at once does not, in the end, help solve the kinds of projects that require cross-functional collaboration and team work.

This rings true.

It's interesting how there's an increasing split of software engineers into those who're very happy with AI and leaning more into it, and those who're increasingly skeptical or dejected.

I don't know how this will play out, but I haven't seen anything like this before.

Can I be both?

Absolutely!

I have leaned heavily into it… and became good at “agentic thinking”, but now I’m going to reverse. I have become dumber, sloppier, lazier, more distracted. I’m less happy than before. It’s hard to escape the productivity addiction and expectations though.

I've seen this happen to multiple co-workers, in the span of months. Two were laid off so I don't know how it's going with them, and another has backed off but not fully abandoned AI code. I'm resisting it entirely because of these problems.

I'm very happy with the performance and ability of AI. I think the best software ever will come out of this. It's very clear to me that AI empowers one to make "just ok" software effortlessly, and "incredibly excellent" software with the same level of effort we previously used to make "just ok" software. So there will be a lot of slop, but there will also be some real diamonds created by those who put the extra work in. The extra work might not involve manually writing or even reading much code, but it is work nonetheless.

I still count myself as "dejected" overall. I'm dejected by the economic shape of AI which seems destined to keep concentrating outsized rewards to the top few who own it or have enough capital to shackle it to their will. I'm fearful that intelligence, once it becomes a commodity you can rent any time you may need it, will be socially devalued in our already anti-intellectual society. I worry that in a couple decades we may resemble the world of "The Machine Stops" (1909), where nobody really knows how things work.

Of course we already live in such a complex world, no one person has complete knowledge of the "stack" they rely on. I can write software in C or assembler but I can't design a CPU and don't really understand how one is manufactured. (I may someday study that field, but I haven't yet!) But I take solace in the fact that somebody out there, has put the time in to study each of those things and is an expert in them. At every layer and in every niche of our engineering world there are masters of their craft, and there are students learning the fundamentals to keep that knowledge alive and growing. What if in 30 or 40 years that is not the case, and there are whole corners of knowledge the modern world depends on where everybody is reduced to "I dunno how it works, but Claude said..."?

I think it's a useful tool, but peak workflow is basically the rough existing development paradigm, if that can be defined, but let AI do the grunt work (typing, some researching) and help you understand things in your own brain, from which you make well thought out decisions and are able to validate output, later define feature requirements, debug, etc....a maintained understanding of a complex system.

The approach currently being foisted upon me is to offload almost everything to LLMs (have it write a complex tech spec -> one shot into tickets -> one shot implementation of said tickets), with humans reviewing throughout, and repeating this iteration until it 'looks good'...throwing sh*t at the wall until something sticks, basically. However, we're being expected to review at a far faster pace than one reasonably can while maintaining an understanding of the system.

Meaning, with this paradigm, humans will largely be rubber stampers and will atrophy their abilities as a result, which will seemingly result in compounding adverse effects.

TLDR in my experience the AI can help with some 'busy work' eg translating a well defined solution in my head into code. The time it takes to understand an implementation and ensure correctness will not change. LLMs cannot reliably do this. This traditionally has occurred during implementation by the implementer, and during code review by code reviewers. I don't think simply shifting the implementer's comprehension 'duties' to code reviewers provides any benefit.

To me, doesn't feel much different than when infra-as-code became popular in the 2010s

It was a huge shift in the world of system administration, gave birth to devops, and, along with the rise of containers & cloud computing, it radically altered how companies ran their infrastructure

Some people leaned into it, others didn't and a huge rift opened up between them

I used to say "automate all the things" - I still do, but I used to too

AI tools are just an acceleration of that mindset and in my world of cloud/devops/sre I'm heavily leaning into AI-driven development and daily workflow

But that’s such a niche part of the industry. AI is in everything.

I find AI autocomplete and chat (asking specific questions, getting general code I can adapt, etc.) to be useful. I'm not sold on agents and agentic workflow.

When I tried using agents for something as a test it immediately started doing things instead of being part of a conversation. I want to be actively involved in the process, not sit back and let AI agents write/generate stuff that I'd have to/end up rewriting/modifying anyway when it goes against the design I have in mind.

The other thing I don't like about agents is their ability to run any command [1]. That seems like a nightmare w.r.t. the potential for leaking secrets (signing/access keys, etc.) or doing damage (deleting files, database tables, etc.).

[1] You can set the option to review every command it runs, but you're then just hand-holding the agent.

I'm in a similar position. I can easily imagine the kinds of tasks where "agentic workflow" seems useful. I just don't have any of them personally right at the moment, and think the world of software has more of them than the needs of software actually justify (due to previous bad engineering and planning).

> The other thing I don't like about agents is their ability to run any command. You can set the option to review every command it runs, but you're then just hand-holding the agent.

The thing about this is that you're treating "agent" as if it refers exclusively to the products offered for that purpose by the major LLM companies. Nothing stops you from designing and implementing your own, with a security model that you think is more sensible. The models won't have been trained against your specific tools, but you can still give them descriptions of what those tools do and then prompt them to make appropriate use of them. All that's really happening is that the LLM outputs some JSON, and your client detects that the output was a tool call (instead of a direct response), parses JSON, invokes whatever other code with whatever parameters and perhaps writes back a response.

There have been several previous HN posts on this, e.g. https://news.ycombinator.com/item?id=45840088 .

This is kinda where I’ve landed as well. I find the chat super helpful as a solo dev almost just as a sounding board for ideas and it is good at pointing out things I’m missing or not thinking through completely. I write almost all the code myself, I like the actual physical activity of writing code and find it keeps me engaged mentally for longer stretches.

I’ve also tried the agentic thing, but I find it so draining laying out in such explicit detail exactly what I want and then cleaning up what it’s messed up, like it’s easier to clean it up than just keep prompting and pulling the magic “Do This” lever.

I probably will never code without AI again. I’ve been doing this for 30 years since I was a teenager, was very fortunate in my career and don’t have to work anymore, but I still code and build because I love it, maybe even more so now with AI, but I would absolute hate my life if I was forced to just run agents all day long and never actually write or be the primary thinking driver for code.

It's a bit like the reality distortion field and how some engineers are not susceptible to it.

> but I haven't seen anything like this before

Just a bunch of controversial ideas: structured programming, null, intellisense, microservices, cloud first, serverless, OOP, "Clean Code", exceptions, SaaS.

For my social circle, the distinction falls almost perfectly along the line of people who see software engineering as a craft / passion, versus those who see it as something that needs to be done to get paid.

The first group is generally disillusioned, for two reasons. First, it's an attack on their hobby. Second, they sort of can't let go: they want to own the product, they want to understand if the code is good or bad, and they spend a lot of time cleaning up after LLMs. Which, frankly, is not fun. You're no longer doing anything clever, you're just a janitor.

The second group is over the moon about it because they feel it gives them the elite powers previously only reserved for greybeards.

I don't think this is unique to coding. It's the same thing with writing. If you like writing, you probably hate LLMs. If you see writing as a boring chore that needs to be done, you are delighted to have a button you can press. Haha, take that all the influential bloggers, tech writers, etc. I can now do your job at 10x the speed.

The one thing that poisons the entire online discourse regarding LLMs to me is just simply not knowing who you're dealing with and what their problems are. This, like all things, is complex and nuanced.

For all my efforts to get LLMs to work for me, they've failed. I figure it is because the tech space I work in (Industrial Automation) simply doesn't have the training data that someone building webapps does. The few times I've put LLM to the test its failed horribly, and since my job has me existing in safety critical scenarios I don't humor what AI gives me at all.

So when I bitch about LLMs and their useless nature, its because of how I've experienced them in my craft. They get 95% of the information right, but that 5% it gets wrong can have very serious repercussions.

But when I'm talking with individuals online who see them as revolutionary? I don't know if their job is editing templates or formatting data or any other host of problems that AI can be revolutionary for. I don't know if they're full of themselves, having drank the AI koolaid. I don't know if they're just really deep into the LLM world and figured out things I have yet to learn (or am unwilling to pay for). But from a HN or Reddit discussion, they're all one in the same and without trying to figure out who they are, their opinions all carry the same weight.

> For all my efforts to get LLMs to work for me, they've failed. I figure it is because the tech space I work in (Industrial Automation) simply doesn't have the training data that someone building webapps does.

Putting safety critical aside (given that is a giant aside) for the moment, I'm a little surprised at this.

One of the most useful things I've found for AI Agent stuff is having an agent live on my home network and finally get the hundred or so random sensors and other IoT devices finally configured and talking together.

It's definitely not perfect though, and not anything I would trust my life to. This year however, I was able to set up enough sensors in my garden to have by far the largest hot pepper harvest I've ever had by min/maxing watering and fertilizing via way too many sensors and data fusing. This was mostly describing the sensors involved, the outcome and dashboarding desired, and pointing it to the various vendor API docs.

I could have done all that stuff manually, but never would have gotten around to it since it was just a tedious boring task. But with AI help I now can tell you precisely the soil moisture level, EC reading, ambient temp, etc. for each of my planters. And with some combining of my rain gauge data and sunlight sensing, tie that into automated irrigation runs. Then the bot just pings me when it thinks the next fertilizing cycle is due.

I don't know if that's revolutionary, but it took a project that would take me weeks to accomplish myself due to around a hundred sensors to configure/setup/etc. to a couple nights of messing around learning what all this talk was over these AI agent things.

I think there are plenty of use-cases where if AI gets things wrong 5% of the time it really doesn't matter a whole lot. I have plenty of these areas at my job.

The biggest part of that is how there's a lot more community involvement for IoT and things like Home Assistant than there are for the proprietary PLC systems.

In the PLC/SCADA space a lot of the development and resources are closed source. There's been a big push in the last few years for Open standards but that's slow to adopt and I'm sure in time we'll see the Controls space catch up (it is notoriously dated compared to the rest of the software world).

But in the meantime its like the perfect storm of terrible-ness for AI. The vast majority of my SCADA work is in AVEVA products, mainly Wonderware and Edge. Wonderware's scripting language is something called QuickScript.NET which is like 80% of VB.NET. The issue with AI is that being a statistical engine it very quickly starts on the right path, then veers into what is very obvious VB.NET training data and starts doing things that are impossible in Wonderware. Edge uses VBScript, but a limited subset of it. So querying AI for help mostly just spits out useless code in the same way; confidently declaring what it gave is possible when it only results in runtime exceptions.

In the PLC space you can get AI to do pretty mundane things with ease. Writing ladder logic or structured text is pretty okay, but like most things that was never the bottleneck in project development: you're not saving a lot of time for the cost of an AI subscription license. As mentioned, Open standards are starting to be adopted pretty quickly and a lot of PLC systems allow import/export of the proprietary formats into XML now which is going to be beneficial in the long run, but we're not quite there yet.

The problem is always that time it messes up, and it can be very subtle. Off by one errors which would be perfect fine in one environment can be catastrophic totally depending on how your I/O Comms were configured. I was trying to interface with some hardware that I asked AI to see if I could solve the pain points of remote enable/disable via network comms. It very eagerly told me I could use OPC-UA Methods to achieve what I'm after, but I was working a platform that didn't support Methods. It even fabricated a path to fake call a method which obviously did not work when applied to a real world scenario. So I asked it how to use Modbus which requires use of FC16 to act on the hardware. Again AI very eager to tell me how to do that, but hallucinated what could be done. Lastly I went for the TCP route which again had AI giving incorrect information - it was far closer this time, having instructed me to send a command in a format which doesn't work out of box, but could be manually configured in the hardware to work. Fortunately I was testing to see if it could save me time, not trying to see if it could do something I was unable to do, but I try to imagine what would happen if I allowed it to do that type of work on a VFD, or PID loop. That tiny little bit of misunderstanding by the agent could be catastrophic.

Each time I share this type of experience (and I have others) with the AI community I'm told I'm doing it wrong. That if I was using better models or different configurations that wouldn't be an issue. Telling me that I need to pay for models - as though I want to fork over cash when their demos are wildly underwhelming. It gives me a lot of overall distrust in the AI community because of this. It feels a lot like interacting with Linux fanatics who are unable to accept that it has faults that aren't tolerable by the layperson. That these people cannot see the faults because it works great in their narrow needs.

But at the same time, using AI to give little tips. How can I speed up this SQL Stored Proc? Why is my DataBinding failing to update? What is causing this OS upgrade to fail? Its done pretty good work. I see it as a tool that works along with me and an IDE, not as a human replacement like so many feel it is ready to be.

I'm surprised to hear that. LLMs make it practical to prove safety-critical C code correct without the cost that has traditionally made formal verification prohibitive. If you haven't explored that area, it's worth a look.

My experience with C and C++ has also been excellent. I work on very large, high-performance C++ systems used to run airlines and airports, and LLMs have dramatically improved both the quality and the performance of our code.

That said, LLMs are tools, and like any tool they take skill to use well. It took me months of experimenting and learning from others to get where I am now.

I made mention on a reply to the other commentator but I feel the issue is due to lack of training data in the PLC/SCADA space. If an AI engine with the appropriate amount of training data was available to me I'm sure it could do the work just fine, but at the current state we have partial knowledge which is too similar to other languages. The result is agents which quickly hallucinate capabilities of the SCADA or PLC hardware and write out code that fundamentally won't work. The result of this is that I distrust AI to get it right enough that I won't kill an operator or blow up a production line. The automation world is too closed source and until the push for open standards that have begun to gain traction really takes off we're sort of stuck.

But then at the same time the code of an industrial system typically isn't the bottleneck in development. Selecting the hardware and waiting for the physical installation is a far bigger bottleneck on a project than what I do. As a result, in the end it isn't the "Correctness" of code in the sense of it is doing what it was instructed to do properly, but the correctness of matching software to hardware and that's where I've seen several failures in AI. If I tell it I have a Schneider Altavar VFD and I get code to interface with an Allen Bradley PowerFlex that's one issue. But if I get code to interface with a Altavar that expects specific parameter adjustments on the drive that haven't been set that's a separate issue and more typical for the faults I've seen within AI. If 99% of drives are configured to operate pumps and that's what AI learns for training, the moment I need it for a mill, aeration fan, or conveyor belt (which have distinct physical requirements) then we have the issues I've been seeing and it takes a trained eye to spot those deficiencies in my experience.

Thanks for the detailed reply. It’s always interesting to hear about real experiences from a different software domain.

So it sounds like the main problem is the under-specified hardware. I can definitely see how that would limit the usefulness of LLMs.

I used to work on consoles, including the N64, which are fairly low-level compared with a typical Windows, Linux, or Mac application. But you’re obviously working much closer to the hardware, which is a completely different world altogether.

Not my experience at all. I’ve coded for fun for 40+ years and professionally for 30, and I absolutely love Claude Code. I write far more high-quality, high-performance code, and I learn much faster. I no longer have to agonize over which fun project to pick. I can do them all.

There is a trick, though: using these models well is a skill. It’s a tool, and you won’t get good results without experimenting, learning from others, and settling on a workflow that fits what you care about.

I haven't really seen this, personally. Basically everyone I know is both at once. We acknowledge that it's really good at what it does and use it, but it's not enjoyable to use, and we end up dejected about the future of our craft and our jobs.

Right now I'm leaning towards using it at work because I'm being paid to ship stuff, but it really forces me to NOT use it after work in order just to keep my personal skills sharp. That last step used to be optional, now it's not imo.

I hated AI from the start, but then gradually warmed up to it. It does help me to write code. And it gets better every day, sometimes literally.

Today I think it is a tool. Like a framer's nail gun. It helps us to be faster and potentially better at what we do. Oh, and got to ride Waymo recently and it got me places, alive and well. So it does some things right.

But the true answer is not so simple. A friend of mine recently asked me a question: "My child wants to go into software dev, should I let him?" Well, crap, maybe consider something else?

> Spending upwards of 75% of my time directly or indirectly dealing with AI every day has absolutely robbed me of most of my enjoyment of my work. Most days feel like that Twilight Zone where you wake up and you're the same, but everyone else is different.

Oh my god this nails my experience to a T.

Completely off topic, but I'm absolutely fascinated by the fact that this blog's central column is built as a... <table> Good old times. Definitely not vibe coded, though.

Table layout, my old friend!

And of course this is inferior to the most advanced CSS technique, which is... literally ASCII-art tables. (As words instead of lines.)

You might also be fascinated if you inspect the layout of the website you're commenting on.

If it works....

I think there were always two camps when it came to writing code, a strong bimodal distribution. People that care about the perceived code quality (aka the ‘craft’) and people who treat the code as a means to an end. At least I always felt this tension. And there is always an exception, and a person can shift between two camps during the same day. But would generally lean towards one. I know I do. And one is not better than the other, they could both be very smart and well meaning. And you need both for success in a for-profit business. It’s been discussed ad nauseam but one particular metaphor I like is Jesus Christ vs James Bond. James Bond would win most of the time. Most people tend to think like James Bond, and if they don’t, the organization strongly nudges them in this direction, promotions get decided and it becomes self reinforcing.

Now a new powerful tool shows up and I think it is time for the Jesus Christ camp to throw in the towel.

Depends what you’re doing. Constraints are different between projects, and advice given for how to utilize AI rarely takes that into account.

Don’t sling when you need to craft, and don’t craft when you need to sling. Knowing the difference matters more, and you can spend your entire career on one side or another.

Craft was premised on the idea that a good writing and composition matters because it would be easier to read and comprehend (sometimes accepted by orgs because it was sold as cheaper).

Now AI can read the code so the craft-leaning people would need to somehow sell it as 'fewer tokens will be spent on a well factored code' or 'well factored code reduces AI hallucinations'?

Now a Slinger (using your terminology) can just ask AI to make the code look 'craftier' - and that is the best case scenario.

In other words, they will not buy what we sell and if they will they can now produce the results for cheaper. Overlooking the devil in the details, as usual.

I enjoyed this lovely rant by Timnit Gebru today. It's titled something like "AI companies are stoking fears of extinction to avoid discussing actual harms,"

https://www.youtube.com/watch?v=aCB-NQtsKrs

How cynical. What if the fears of extinction are _rational_? Given the stakes, we're obligated to treat the threat as a real possibility until proven otherwise.

People will go to great lengths to avoid believing truths that dramatically challenge their worldviews. Consider the case of Rudolf Vrba and Alfréd Wetzler, who escaped from Auschwitz and were initially met with disbelief when they described the atrocities they'd witnessed.

I agree. AI skeptics are widely met with disbelief when describing the atrocities that LLM operators are inflicting on society right now.

> Given the stakes, we're obligated to treat the threat as a real possibility until proven otherwise.

No we're not. The seventh day adventists also believe that the world is about to end, why are we not obligated to treat that threat as real?

Nobody ever gives a real reason why extinction is a plausible threat from current model development, and Chinese researchers don't seem to agree that it's plausible either.

>Now imagine that we had spent all these resources on doing the basics: ensuring your organization has an up-to-date and comprehensive asset inventory with fine-grained package listings; building infrastructure that supports regular, frequent, and automated OS and applіcation updates; automatically rebooting systems when they hit, say, 30 days of uptime to ensure these updates are picked up;

That would still leave the root cause of most of our problem un-addressed, ambient authority based Operating Systems are fundamentally insecure. Unfortunately, all of our systems share that architecture.

Do you remember when people were proud to say they were computer illiterate and refused to learn how to use them for as long as they could? Do you know the tales of John Henry or Paul Bunyan? I am sure there are plenty of other analogies that can be drawn, but what cannot be denied is that if most people believe that this is progress, then this is where we are heading.

I am using safari on an older ipad and initially thought this website was an art project. The screen was black with an unlabeled toggle button. Tapping the toggle made everything visible. A dark mode bug, I guess. The intrigue was mild yet odd.

I saw this post in my RSS feeds early this morning. The post was so poorly thought out, pessimistic, and full of anti-AI myths that I unsubscribed from this blog's RSS feed. I'm surprised to see it make the front page of Hacker News.

And I’m now inclined to subscribe to this person’s blog. So I guess everything remains in balance.

.... enjoy the articles and fight against the inevitable future?

I am not concerned about being a meat proxy (fine).I worry about the possible "unfounded oracle." A model that although constrained by humans, (like delphi) acts without any human criticism. Like multivac from Isaac Asimov's short story:

https://www.commonlit.org/texts/the-machine-that-won-the-war

If you're surrounded by fools, or maniacs, or any other adjective you don't like, then that's your clue to change environment and the people you surround yourself with.

Most of the times it's really just that. Entire nations of people can go nuts or went nuts before you were born. People will never change, so go where better people are instead of banging your head against the wall.

> half of the people you interact with have turned into meat proxies

Samuel Butler presaged this in Erewhon (1872) : "May not man himself become a sort of parasite upon the machines? An affectionate machine-tickling aphid?"

were we building fabulous software beforehand? It was clumsy, buggy, expensive, demanding, oppressive, cluttered.

AI , trained on that garbage , is producing what now? Supposedly clumsy , buggy and so on.

The lack of self awareness is just hilarious.

> these AI models not only lend themselves to generation of Child Sexual Abuse Material

Err, how exactly does the generation of questionnable material abuses a child? Assuming there's no actual CSAM in the training corpus, the causal link between that and children being actually exploited seems tenuous at best, similar to drawings of Japanese middle schoolers, fictions about underage people...

Those are nothing like purchasing or seeking out pictures, movies, drawings, or depictions of actual abuse of actual children. As disgusting as it may be, not all child porn is CSAM.

Of course, if there is CSAM in the training data set, that's a different story.

And let me be clear: I don't whether, nor how much, the consumption of purely fictional child porn by adults ends up hurting (or helping) actual children. For all I know it normalises associated behaviours and is responsible for a significant portion of incest cases around the world. Or it acts as a safety valve for paedophiles so they don't turn criminal. Or both, or none, I don't know. My guess though is that it's similar to how action flicks and FPS games influence gun violence.

---

That was about the worst I could say about this post. Most of the rest resonated with me. I'm especially scared of the dependence and deskilling.

A starting point: https://en.wikipedia.org/wiki/Relationship_between_child_por... There seems to be conflicting evidence, fraught with spurious correlations on both sides. Honestly, I'm not sure which policy I'd adopt in the face of such uncertainty. I'd probably fund more research.

I've been coding for 22 years, 14 years professionally, so I appreciate AI coding.

It was annoying before when I had a very clear idea of what I wanted and yet it still took a whole week to implement.

Now, with AI coding, I can implement any decent sized feature, exactly how I want, in a few hours and it's all polished and ready, all edge cases handled.

Everything it's generating looks almost exactly like what I would have written, just written faster.

Personally, I feel avenged by AI because I had been saying for years that you don't really need static typing once you reach a certain level, getting types right isn't the most difficult problem and isn't a significant source of bugs. Everyone was gaslighting me about this... Yet now everyone can verify themselves because AI has reached this level I was speaking of.

Now anyone who coded with a frontier model on both TypeScript and JavaScript can see that it makes no difference at all. I can't remember ever seeing Claude mistake the type of a variable with either language... And TBH, I find Vanilla JavaScript from AI to be more succinct, not over-engineered like Typescript. Which is a point I'd been trying to make for years about incentives and actual patterns of usage (as opposed to theoretical benefits).

If you have a blog post on the topic I'd be interested. My experience with AI coding isn't really the same, but my experience with typing discipline definitely is, and I've thought about writing about it myself but don't really know where to start.

hard to believe anyone who writes something this overwhelmingly positive.

I use a ton of AI, it writes a shit load of code for me, and I eventually get good results.

Its also goes off on ridiculous tangents, implements things i dont want, over engineers almost everything. Makes large code bases convoluted and unreadable. Yes humans also do that, LLMs do it a lot faster.

> Now, with AI coding, I can implement any decent sized feature, exactly how I want, in a few hours and it's all polished and ready, all edge cases handled.

No way a human wrote that, sounds like a late night TV advertisement.

You obviously haven't worked on a good codebase. I've worked on both so I know the difference.

Same feature on a good codebase vs bad codebase is like 1k lines of code vs 20k lines of code and is worse in the latter case. Not an exaggeration.

Clean code is real, it's just not what's been peddled to us.

If your code is full of bugs and you ask Claude or Codex to build more features, it will take 10x to 20x the effort because of workarounds and hacks which keep compounding.

If your code has very few to no bugs and no hacks/workarounds, it will also implement hacks/workarounds and no bugs. The difference is pretty insane actually.

You can try it yourself; create a really low quality codebase, full of hacks and try to build features with AI based on its recommedations.

> Clean code is real, it's just not what's been peddled to us.

My read is: the people who "peddled" it had mostly the right ideas, they were just bad at implementing them, and bad at writing code in general, and also the contemporaneous popularity of Java caused some serious issues.

> So no, I'm not going to set ethics aside and then actively offer myself up as tribute to self-amputate my brain. I'm sorry if everybody else can't get rid of the brain slugs, but at this point I'm just looking to get off this ride.

Same. I quit, aside from two contracts I'll serve until they run out, I'll be studying Electrical Engineering. Even if AI takes over that space... good luck constructing a robot with the dexterity required to pull cables.

To anyone in IT... go and study for some qualification in tasks not easily doable for a machine. The trades are going to be a goldmine especially with all the boomers retiring. Or buy some ranch in the midst of nowhere and farm artisanal chickens, cattle or goats.

(For legal reasons, the latter was a joke. Farming is extremely risky, requires a lot of upfront cash, a lot of what used to be generational knowledge, but it seems to attract a lot of IT people, especially for whatever reason the goat farming)

there must be a trope by now for article titles sounding interesting yet scratch the surface and its only about AI.

If you think vulnerability research simply doesn't matter, you have a lot of company in that opinion. It's something Bruce Schneier used to argue, and Marcus Ranum, and a bunch of other people that used be on closed secret vulnerability-sharing mailing lists before Bugtraq blew those cliques up. These are very old arguments.

But if you do think vulnerability research matters, and you're trying to argue that frontier models aren't a seismic change for that discipline, you have almost no company. Vulnerability researchers are overwhelmingly leaning on automation to find vulnerabilities and, just as importantly, generate the tooling required to test hypotheses.

You can feel about that however you want to feel about it. I mostly don't care, except: you can watch people like this being negatively polarized back into the bad old days of the mid-1990s, content-free CERT advisories, and vendor-controlled "responsible disclosure" by a use case that frontier models unimpeachably excel at.

It’s amazing how half the developers on earth live on a completely different planet now. There are plenty of new challenges, sure but we are far past the point where we can have a debate about “is it useful?”

And yet we continue to do so. I understand the feeling of loss some people may be facing. And there are definitely some really bad practices - like nakedly spewing claudspeak at your colleagues instead of communicating. Or raising a PR you don’t understand. There are asymmetries we haven’t learned to navigate. But we aren’t returning to a world where it doesn’t dominate our discipline so it’s best to find opportunities.

The question most of us are asking isn’t “is it useful”, but “is it worth it”. Engaging with these tools involves no small amount of self-debasement and long term degradation of skills; do I want to sacrifice myself on the altar of productivity?

For me the answer is still absolutely resolutely “no”.

You’re already using a computer to automate massive amounts of what used to be manual human effort.

Why is using AI tools self-debasing or degrading?

Because AI removes you from analytical thinking, and the brain is a muscle. The less you think, the less you are able to think.

Programming is inherently analytical and logic-driven. It’s a great brain exercise, even if the code has no value. When you use AI, you lose that exercise. What replaces it isn’t the same.

I mean, consider. I can solve a crossword, or I can ask AI to solve the crossword. AI will do it perfectly, and faster too. But which one is better for my brain? What was my goal with solving the crossword, anyway? On the surface it was to find a solution to the crossword. But, maybe, deeper down, the goal was to improve myself in some way.

People conflate the improvement of products with the improvement of self. Maybe it’s a result of our consumerist mindset. But the truth is the product can improve greatly, and you can be making it, while you yourself degrade.

We write better and better code, at a faster rate, but are we better programmers than before? Maybe, maybe not. I’m leaning maybe not.

It’s not new or unique. The assembly line is the same. Sure, I can build better furniture faster at the factory than by hand. But what is the goal here? Is it to make the best furniture, or to become the best carpenter?

If it’s to make the best furniture, the best product, then I’ve won. If it’s to become the best carpenter, then I’ve lost heavily.

The best carpenter obviously doesn't use any power tools. Are they allowed to use hand tools, or do they need to split the wood with their bare hands?

It's not clear to me why programming is inherently analytical and logic-driven but prompting and iterating with an AI is not. What about Applescript, where it's programming but in something closer to natural language?

The core of this seems to be that there are people who assume that users of AI turn their brains off the moment they open a Claude prompt. There are surely some people who are doing that, but there were also plenty of people who were shipping sloppy code before. There are far more people for whom this another tool in their toolbox.

> But, maybe, deeper down, the goal was to improve myself in some way.

Or, you know, to pass some time in an enjoyable way. To have fun.

I dont know how you can work in this discipline, use AI, and also ask that question.

almost every developer i know is fully aware, we are degrading every aspect of our skills.

The code writing, code reviewing, code understanding.

The more "time" passes and the more the code base grows, the more disconnected we become.

Our horse riding skills have degraded even more.

Jokes aside, I think this gets at the core of it.

If you really loved horseback riding, or you really loved breeding and training horses, or you were really good at it, I'm sure that cars fucked up your day, and that it sucked to see cars replace horses.

I'm sure there were plenty of people who said that cars were horrible and that we should stick with horses. I'm equally sure there have been C programmers looking derisively at Python programmers, Python programmers looking derisively at Node programmers, etc etc.

But there's a pretty big gap between the feeling of unhappiness that something you liked is no longer as in-demand / profitable / common / whatever, and saying that cars aren't useful and we're all worse off for letting our horse-riding skills degrade.

Lots more people die in car crashes than horse accidents. Good analogy!

Do they? Horses aren't particularly safe, there's just way fewer of them and they don't travel as many miles.

Normal people didn't ride horses everywhere. They weren't used like cars. I'm so sick of this analogy. People walked places. If it was too far to walk you used a horse drawn bus/tram. If you had to much to carry you used a horse drawn truck. Every single person wasn't riding an individual horse ten minutes each way to buy groceries every day like that do now with cars. Cars unquestionably kill drastically more people than horses ever did or even could. Tens of thousands a year in this country alone! The leading cause of death for children and young adults!

We're pretty off track given that ya'll are nitpicking the metaphor rather than the actual topic, but claiming that horses were never common doesn't really have any connection to the claim that per-mile-moved / per-human-carried, horses are also dangerous.

They are nowhere close to cars

https://www.biamo.org/wp-content/uploads/2023/01/HorsesOnly-...

Never use analogies when arguing online. It only ever results in exactly this, people nitpicking the metaphor. You've essentially given them a way to "be correct", about something irrelevant.

I'm not nitpicking, it's just a bad metaphor, entirely, on its face. No nits about it

Do you have something you want to discuss about the actual point I’m making, or are you acting strictly on behalf of the metaphor police?

This does not get to the core of it, LLMs are not equivalent to cars.

All of your analogies (in other posts) discount that: - LLMs are not deterministic - they do not produce high quality/expert level output. - They are not easy to control consistently.

Once again, and i mean this in the nicest way possible. You do not sound like you understand how things operate at an expert level of engineering. I do not know of any senior engineer who thinks LLMs can consistently produce quality output.

You are way too invested in your argument and refusing to see other perspectives

It seems really convenient for you to speculate about my level of expertise so that then you can make claims that no senior engineers think something.

I guess I'm an outlier then.

AI has radically improved my ability to parse and understand new codebases, to validate my hypotheses, and to work in complex systems.

I wish I knew what I was doing differently than almost every developer.

You're not an outlier. The anti-AI stance is likely to be vocal online. They're quite incentivised to talk about it. Most programmers have adopted AI assisted programming (my own assertion) and most will feel like you and I, that it is useful. I know of a few programmers personally who hate it, but even they use it. There is no going back.

It is pretty obvious that when you come into a large incomprehensible codebase (which most are), AI will be able to reason about it more quickly and guide you through it.

I switched away from my work for a few minutes, but at the very moment I'm using AI to summarise some spaghetti code in a particular area in our codebase, so I can refactor it.

And that is my style of working at the moment. AI coding is useful, sometimes amazing, but sometimes irritating, and from experience I think just telling it to refactor this area would result in something I'm not happy with. So instead I'm getting a handle on it, I'll form the design I want, then tell it to do it. I have a feeling a lot of complaints are from people who won't bother to do that.

You are not an outlier. 40+ years of hobby and pro coding experience here and I absolutely love Claude Code. It enables me to dramatically increase code quality, performance, and features while still doing the fun parts and letting the LLM do the boring parts.

Not getting distracted by water-cooler talk about how AI is going to take your job and that the end times are around the corner, presumably.

The bigger challenge is that the mechanics of the day to day work have entirely changed. Mechanically, using C is similar to using JS. You’re tying and compiling and checking results, whatever. With an Agent what you actually do is completely different. Even as technologies have changed in the past and people have learned more, your mechanical day to day work would look similar, just with different tools. Not so with AI.

So it should be utterly unsurprising that many people are going through a bit of an existential crisis around this. Because what you actually sat down and enjoyed doing, what motivated you, may be entirely gone now.

And on top of that, there’s an extreme amount of pressure to do more in less time, which is by definition stressful. And do much more context switching.

I’ve often spent more of my time reviewing code than writing code off and on for decades as my role and team size changes. So, none of this is new to me except that the loops are much tighter and I don’t have to debate the social cost of pointing out what is workable but not acceptable after the 4th review cycle when I suddenly notice something new in code that had been there all along.

The truth is a lot of developers are terrible at reviewing design specifications and code. They rarely - if ever - find any real correctness issues and more than half believe it’s mostly a status / dominance game. Many can’t force themselves to deeply think about the code if they aren’t writing it.

Maybe it’s worth actually listening to what these people are saying, or you may find yourself increasingly frustrated and confused as these debates intensify and AI becomes more politically and socially toxic.

I've heard what they are saying, and a lot of it is factually untrue and confused. I wasn't happy about all of this when it first began happening, yet wasn't working. But once it was actually working, I realized I wanted to build software more than I wanted to craft software. Some people want to continue practicing their craft, which is a fine thing to want. But don't pretend its anything else.

As I see it, "engineers" going all-in on AI are missing two fundamental truths of our profession:

* Writing code is a form of communication, as well as a process through which complex systems are thought through, understood, and formalized.

* Complexity is managed by building on top of robust, deterministic abstractions.

Vibe-coders deny the need to understand complex systems and pretend that AI is a new layer of abstraction. I think that both of these perspectives are dead wrong. This remains the case even if LLMs are marvelously good at generating code.

Put another way, software engineers who mostly deal with code through their agents have switched careers from engineering to some form of management, even if they deny it to themselves. In no way does this deprecate the field of software engineering.

And then there are the externalities which have been discussed ad nauseam at this point, and remain as true as they ever were. (Economic, environmental, political — take your pick.)

Not everyone using coding agents is vibe coding. I still review code until I understand it, judge the tests cover it, and that it is the right way to do it. They constantly make mistakes like over building, handling contingencies that can't exist, duplicating code etc. Its still MUCH faster than the alternative. I've been working this way for decades. It is not different from reviewing human code and designs except the loops are much tighter, and refactoring is so much cheaper. There may be a lot of people who can't do this work effectively, but the thing about skill issues is they can be improved if they are recognized for what they are.

Almost everyone who uses coding agents extensively describes offloading some degree of understanding to the agents: reviews, tests, documentation, filling in bits "that don't matter," whatever.

As someone with years of FAANG experience, reviewing code competently was actually more strenuous than writing it in the first place. In fact, I'd estimate that truly understanding a codebase required a similar time commitment to writing the amount of code in that codebase. So I don't see how unprecedented speedups are possible with a human in the loop unless the reviews are only being skimmed.

The key to productivity and increasing complexity is inventing solid, thoughtfully designed abstractions and building on top of them, not taking shortcuts by way of a code extruder. If code is constantly getting duplicated or falling into the same patterns, figure out how to abstract or encapsulate it.

Reviewing code is more strenuous but it does not take as long - unless you count all the procrastination that happens first. And it goes much faster if you make a comment and get immediate changes because you don't have to rebuild context the next time you look at it. And yes - not every detail needs to be fully understood; sometimes verifying tests cover what they claim to is enough. Knowing which details are important and which can be hand waved is a bit of an art and I can't claim to always do it perfectly and don't always do it well for human authored code review either. And yes, I have to ask for refactorings and suggest abstractions in some cases. But there are now many tasks that coding agents actually do better than me, and they cut fewer corners, handle more edge cases, test more edge cases, research suggested designs, reviewing library docs & code much faster, etc.

Here is a prediction for you.

AI coding is only going to increase. Eventually coding is going to join several other fields where it is much more about verification than the building of the code. I predict that eventually most programmers are not going to be looking deeply at the internals of the code, but are going to be spending most of their time on the verification.

I predict shit will just be in a constant state of brokenness, all the time, with no one understanding what the fuck is going on.

You can only superficially “verify” what you do not fully understand.

If that were to happen those companies, products and practices would fail. They would be out-competed by all the organic enterprise shovel-ware and SAAS shit-code that has set such a high bar of quality over the last 30 years.

It will be the same category of shit, but accelerated tenfold by the spam machine.

Unfortunately, shittiness in and of itself does not cause business failure. Otherwise, the industry would have moved on from Electron long ago.

Please, lets revisit in the next few years. I think one of us is blinded by bias, and I think it is you, and I think the only way to judge is evidence, since arguments obviously don't make any difference.

This was the exact line I heard during the brexit debate here in the UK. I am not from the UK. I'm not from the EU. I don't know whether than makes my view biased or not, but I hope that I was as unbiased as I could be during that referendum.

It seemed to consist of a bunch of unconstrained and disjointed lies, and whenever the other side tried to point out they were lies, they were told they just weren't listening.

You were right about one thing though, it was increasingly frustrating.

“You’re wrong because this thing somehow reminds me of another, completely unrelated thing.”

The author speaks for me.

> And yet, despite having found literally thousands of new vulnerabilities (only a fraction of which were reported to Open Source projects, by the way), I don't think that we're any safer than before.

I’m not sure I follow. There are dozens of zero day, full RCE/privilege escalation linux vulnerabilities that have been found and fixed due to the use of LLMs. Some of which existed for decades. Unclear how that doesn’t make us safer than before.

> If you actually thought your product will kill all humans, then you could, you know, like, just stop building the torment nexus. All by yourself, no government regulations required.

Wrong. This is a classical prisoner's dilemma. _Everyone_ has to stop, otherwise the undesirable outcome happens faster. It requires coordination, which can be achieved through regulation.

This is someone's vent post.

How perspicacious of you!

Thanks <3

Printing press was bad. Usenet was bad. Google was bad. AI is bad too. Robots? They will kill us.

I cannot say that I agree to all the issues in the post. But men I am sick of meat proxies too.

I cannot take this post seriously -- when it linked to the "Crimes" that Anthropic committed it showed that this author has not understood *any* of the papers or systems that he is critical of in that moment.

absolutely uninformed.

A good account of one's own exhaustion but a remarkably poor argument about AI. This person spends 75% of their working life dealing with something he hates, reverse-engineers a worldview from the resulting burnout and concludes that everyone else has "brain worms."

There is one useful observation buried in it, which is that when vulnerability discovery becomes cheap while triage and remediation remain expensive, you create a larger queue rather than a safer system. Fine. I think that point has been raised in many of the other discussions on this site. But then he inflates that into "finding vulnerabilities has never been the bottleneck" which is demonstrably false. Discovery may not be the bottleneck in fleet-wide CVE management but it is plainly the bottleneck for every serious vulnerability that nobody has discovered yet. His own examples of flaws surviving decades of expert scrutiny shows this.

The Glasswing figures show a triage-capacity problem. They do not establish the essay's much stronger claim that "we aren't any safer." You simply cannot infer that from the ratio of findings to fixes! You would need to know which findings were exploitable, which fixes were accepted and deployed, whether attackers would otherwise have found them, how much review work was consumed, and whether the incidents were prevented. The essay measures literally none of that.

Its alternative history claim is also unsupported. Supposedly every participating organization diverted dozens of senior engineers, spent millions, and would have produced more security by assigning those same people to asset inventory and patch automation. Perhaps. Where are the staffing data, budgets, opportunity-cost estimates or comparative outcomes? Promotional model credits are not engineering expenditure and vulnerability researchers, at least to my layperson's knowledge, are not interchangeable with infrastructure engineers.

The broader essay repeatedly collapses distinct anchors and problems into a single villain called "AI." Grok's sexual-image scandal says something damning about xAI. Military targeting failures say something damning about the military, its contractors, its data and its command structure. Datacenter expansion creates genuine environmental and infrastructure costs requiring policy. But none of that establishes that someone using an LLM to understand an unfamiliar API or to write a PRD is morally participating in every one of those failures. That's just guilt by association.

The irony is that the author condemns anthropomorphic language for diffusing corporate responsibility (which is fine), then spends the rest of the essay treating "AI" as a unitary agent responsible for all sorts of things, from corporate procurement and military decisions to power generation, bad software, ugly prose and programmers becoming lazy. He recreates exactly the accountability blur he claims to oppose.

Overall it's a poorly thought out vent piece.

AI does not use much water. If you believe it does, like this article does, you have fallen for propaganda and lies, ironic in a post calling everyone using AI stupid.

The main cause of water fear mongering is that one woman’s well was dirty during construction of a data center. First construction is not the center running, and having lived on a well they get dirty for all kinds of reasons.

datacenters are stinky though and ruin property values. (as well as can, and has, polluted water sources in some cases)

I don't disagree with your assertion; however you do have to realize that their grievances, while not completely informed, are not wrong.

Someone complaining about water usage is a very strong indicator that they aren't well-informed enough to contribute anything worth reading, though.

You still have to grapple with the merits of their argument to find out if the indicator is correct or not.

In a world where everybody seems to be quoting grok with no further thought process you would think people would be more welcoming of someone who is "slightly wrong" on occasion. It's the strongest indicator that a flawed human being got their information not directly from a machine.

No, I don't.

If someone is spreading deception, intentionally or not, then it's entirely within my rights to point out they are without engaging further. If they want to be taken seriously, they have to at least indicate that they are willing to self-correct in the face of direct evidence that they are wrong.

> If someone is spreading deception, intentionally or not, then it's entirely within my rights to point out they are without engaging further.

What is it with millennials and "rights"? It's a sincere question. People don't owe you anything and neither do you owe me anything, sure... so why do you demand another have a higher standard that you yourself hold to. I know you don't hold yourself to that standard because you would rather nitpick someone on a single inaccuracy then grapple with a very real issue (which you yourself would not appreciate if they did that to you). The respectable thing to do is to note the inaccuracy and approach the valid portion of the argument sincerely.

What you demonstrate is a lack of care for those around you, and a lack of care for what is actually true. You would rather "umm actually" then treat others with an ounce of respect. It's shameful and devoid of grace.

Maybe yours do, but all the CyrusOne date centers outside of Chicago smell like is grass and landscaping; they smell a great deal better than the car exhaust choked strip malls that dominate the exurban sprawl areas they tend to get sited in. I would much rather have one of those DCs as a neighbor than I would a Walmart.

As long as they can operate under capacity (for the grid) I don't disagree.

However I live in a much nicer area than Chicago and any Walmart, large mass of people, or DC would really destroy the PV as well as it would be an eyesore.

I don't really understand why so many just want to over industrialize so many beautiful places so that a corp. can sell the next gimmicky iteration of our "best model yet" which is just slightly slower and more expensive then the previous.

They're stinky? How? And how do they pollute water sources? A datacenter is a building filled with computers that uses water and air for cooling. What pollutants get "emitted" by a bunch of CPUs and video cards? No one seems to go into details.

There are certainly problems that can be crated by dumping heat into a body of water, but that's not the argument being made.

I'm willing to bet that a datacenter creates a fraction of the pollution of a similarly sized industrial facility such as a chemical plant, refinery, smelter, or steel mill.

It's mostly the emissions from diesel generators which many DCs use as it's far too expensive to ever shut them down. Often diesel generators are used to lighten the load on a power grid as well, so many DCs will run generators 24/7.

That source of power, since our grid is so outdated and struggles with the weight of these power-hogs, EVs as well as increased usage of AC during the summers create some dangerous conditions.

There is the risk of heavy metals getting into the water supply as the copper degrades but that will only be a problem down the line, that's prescient but not yet realized.

> I'm willing to bet that a datacenter creates a fraction of the pollution of a similarly sized industrial facility such as a chemical plant, refinery, smelter, or steel mill.

I agree; however I would not want a DC or a refinery, smelter or steel mill anywhere near my town. We have industrial districting for that reason, and because a lot of DC projects want to be put next to lakes and water sources, this causes issues for the residents.

> And how do they pollute water sources?

In the case of one amazon data center, they were pulling in water that already had nitrates in it due to agricultural run off, but it wasn't enough nitrates to cause a problem for anyone until the data center came. It pulled in a massive amount of water, a lot of which was evaporated and that left behind water with very high concentrations of nitrates.

Normally, the nitrates would be filtered naturally and would mostly remain in the top layers of soil. Amazon pumped out 10s of millions of gallons of the vastly more contaminated water. It was so much water that it was forced much deeper into the ground. It was pushed deep enough to reach the aquifer which contaminated the water source for an entire community of 45,000 people.

Suddenly the water coming out of people's taps was polluted at 10x the federal level for safe drinking water. People got sick. Cancer rates spiked. Babies died. Amazon was forced to settle and pay out 20 million for water treatment and to dig new deeper wells for the community.

> I'm willing to bet that a data center creates a fraction of the pollution of a similarly sized industrial facility such as a chemical plant, refinery, smelter, or steel mill.

This is whataboutism. The existence of other bad things, or even worse things, doesn't make data centers good. They are a problem for the environment regardless of whatever else is also a problem and that's something that needs to be considered every time someone wants to build one. The companies who build and operate data centers should be held 100% accountable for the harm they cause.

And what are you basing that understanding on?

https://www.eesi.org/articles/view/data-centers-and-water-co...

According to EESI (a bipartisan-founded non-profit), a large data center:

> can consume up to 5 million gallons per day, equivalent to the water use of a town populated by 10,000 to 50,000 people.

Also noting that AI datacenters can use more.

Consume it how, exactly? Are the staff drinking it all? Because the the kit sure as hell isn't. :-/

From the article:

> Approximately 80% of the water (typically freshwater) withdrawn by data centers evaporates, with the remaining water discharged to municipal wastewater facilities.

Is that supposed be a lot? That article states that all US data centers combined consume 450 million gallons a day, equivalent to 90 towns, which does not sound like very much.

I'm not the one here making assertions, I'm asking questions of the person making claims. I'll leave it to others to judge what is 'a lot' or 'reasonable'.

I don't think 'anyone who thinks X is Y' is a good basis for discussion, which is the real point.

This sounds like a lot, but it’s actually not very much. I mean, one singular almond requires 1,000 gallons of water. Beef requires shit ton, too.

Go vegan for a few weeks and you’ve just negated all your AI water use for the rest of your lifetime.

The major hyperscalers report water use: Google consumed 10.9 billion gallons in 2025, up 34%, versus Amazon's disclosed 2.5 billion.

Thank you, I wasn't aware they did this:

https://sustainability.google/reports/google-2026-environmen...

https://www.aboutamazon.com/news/sustainability/amazon-data-...

The Us uses 320 billion gallons a day. Google thus used 0.01% of total US water.

Or about 0.6% of the water used to grow almonds

Now compare to the planned datacenters.

US data centers are going to 3x this year. Still not a lot.

3x the resource usage? Or the raw count?

3x the compute

It's helpful to put this number in context, although if the people you're talking to will do anything to avoid using the metric system, that becomes much more difficult.

Fortunately, now we have computers, which are great at unit conversion, so we can cut right through these feeble attempts at obfuscation. In non-medieval units, 10.9 billion gallons per year is 41.3 million cubic meters per year or 1.3 cubic meters per second. The river a few kilometers from me has a discharge of 22000 cubic meters per second (of fresh water dumped into the sea), so Google is the equivalent using 0.006% of the Rio de la Plata. Lake Superior contains 12070 cubic kilometers of water, which is 1.207 × 10¹³ cubic meters, also of fresh water, so Google's current use would dry up Lake Superior in only 290,000 years.

Google is mostly not very close to Lake Superior or the Rio de la Plata; a more relevant river might be the Sacramento River, which drains into the San Francisco Bay where Google was born, and it's much smaller than the Rio de la Plata. Specifically, its average discharge is 797m³/s, so Google is using 0.16% of the discharge of the Sacramento River. (But some of that is at datacenters far from California.)

https://en.wikipedia.org/wiki/Water_in_California#Agricultur... tells us that agriculture in California uses 34.1 million "acre feet" of water per year (Americans will do anything to avoid using the metric system, especially when they're trying to be sneaky). In non-medieval units, that's 42.1 billion cubic meters, and per year, it's 1330 cubic meters per second, almost exactly 1000 times Google's use. Of that, 18% (240 cubic meters per second) is spent on alfalfa, or was in 02014, anyway.

One major difference is that you need fresh water, like the water of the Sacramento River, to irrigate crops, but you can cool data centers with salt water from the ocean. It's less convenient, because there are a variety of annoying things that happen with biofouling and corrosion, but that's what you'd do if hyperscalers scaled up by another factor of 100× so that their water usage became large enough to be environmentally significant.

But that doesn't happen until they have 100× the computing power they have today. So anybody who tells you AI datacenters are a major water user is lying.

You can always tell an AI apologist by how they latch onto the water thing and ignore the electricity, air pollution, noise pollution, etc.

The electricity, air pollution and noise pollution are all from natural gas. Data centers aren’t loud, on site temporary natural gas turbines are. We can instead install solar and batteries and none of these will be issues. We already grow corn for fuel at 1/10th the land efficiency of solar so it’s not like we don’t have the land for it

I feel like there's a huge political/economic opportunity in the US simply from getting government to shift subsidies from corn/ethanol to solar. Of course you also have to get buy-in from the farmers, but.

I'm not talking about hypothetical possible futures in which people do the right thing. The datacenters that are being built right now and which are in operation right now use single-stage gas turbines, last I checked.

HVAC equipment can also cause a lot of noise pollution all on its own.

You could, you know, stop spreading deception about water usage if it bothers you so much when people call out deception about water usage.

At this point, I almost consider it trolling to bring it up: purposely throw in something incredibly stupid, and then throw a hissy fit when predictably called out on it.

I, personally, do not bring up the water thing. So you're talking to the wrong person here.

"That's because finding vulnerabilities has never been the bottleneck in information security."

Thats just not true.

How will more updates fix the human component?

Sry not sure what you mean.

I only wanted to say that finding security issues was def bottlenecked by humans.

I agree that everyone's lost their minds, but so has this guy.

LLMs are actually producing a lot of pretty good stuff, can debug things that would have taken an entire team a week, and the people running AI companies are also generally like the most boring supervillains imaginable. All true.

It is also completely screwing up the global economy, and most people barely understand what it is and what's happening.

It's also not going to collapse anytime soon as a technology despite what the economy does, and it's going to get a lot worse before it gets better. Everyone who's extreme on it on either end is generally wrong.

I believe both extremes are right, because more than one thing can be true at once.

that we're creating god but that it's also a fad that will go away soon?

I think you just accidentally described what's already been happening for 7000+ years.

Though if any followers still exist, they probably wouldn't agree that 𒀭𒂗𒆠 was just a fad.

To save others the search query:

> Enki (Sumerian: 𒀭𒂗𒆠 dEN-KI), also known as Ea (Akkadian: 𒀭𒂍𒀀 dE₂-A), was the Mesopotamian god of wisdom, crafts, fresh subterranean waters, magic, and incantations.[1][11] He was believed to rule the Abzû. In Mesopotamian astronomy, he was associated with the stars of the southern band of the sky.[17] Enki's wife was Damgalnuna, and their children included Nanshe, Asalluhi, Marduk and Enbilulu. His sukkal (attendant deity) was Isimud. Servants of the god included lahmu, kulullû, and the Seven Sages.

I figured that with the crafting, piping in fresh water, and magic incantations, Enki would be the most data-center-adjacent of the deities. :P

yeah, exactly. soon and god are both subjective. think more on it “going away” is subjective too.

rotary telephones and pay phones went away but we still have phones.

> It is also completely screwing up the global economy, and most people barely understand what it is and what's happening

What is screwing up 'the global economy'? LLMs?

Probably the concentration of a too-big-to-fail percentage of capital into a single digit number of tech companies, many of whom are engaging in circular financing deals between one another while also ransacking the entire consumer electronics space.

something like 25% of the current American GDP growth is because we're shoveling money into AI companies that have no idea how they're going to pay it back beyond selling subscriptions to chat bots, much of the world will take a hit on that crash

Have you tried buying a computer lately

Reply on news.ycombinator.com