spenvo

So, ChatGPT-like tools are accelerating(/or locking in) the decline of community sites like Stack Overflow, which used to be an indispensable tool for both junior developers to troubleshoot issues and also technology builders to see where users were hitting pain points with your tech (so they could improve it).

A couple of major ramifications from its decline:

#1 (the bigger one): The decline of Stack Overflow-like sites will (imo) degrade or cap the quality of ChatGPT tools themselves on questions pertaining to code post-2022. I doubt that advances like "reasoning" or other AI breakthroughs are going to fully make up for the oncoming draught of quality training data. Sites like SO were a crutch that companies who underinvested in documentation leaned on (their attitude essentially being: "we've done enough, let the coders figure it out amongst themselves, b/c it works"). I doubt companies are going to suddenly realize they need to invest more in solid docs (for both the developers and AI companies). -- While many initially saw this coming (AI killing the web it trained on), now we have pretty dramatic data show that has happened.

#2: questions about new technologies and their shortcomings will be asked in the dark, giving AI companies valuable data that used to otherwise exist in public forums. Among other things, this will make it harder for tech-builders to know what to improve, therefore preventing it from improving as quickly, and keeping people more reliant on AI tools for troubleshooting. This seems to be another example of AI companies _creating_ problems that they are best positioned to "solve".

That's a fair point, and it's your call - however, if the new (major) development is covered in this way then 1) users on the front page won't see mention of it at headline level and 2) the discussion of that development on HN will be affected by/limited to the time-decay of a post that is 12 hours older. I understand that there are tradeoffs at play, it really comes down to if the development at hand is big-enough to justify another post, and, again, that's your call.

Ah good, so one was finally not marked Dupe or Dead. I believe even that thread was marked 'dead' for the first 50 minutes or so

This is new reporting that a hacker has breached TeleMessage, the parent company of TM SGNL and other modified chat apps. The breach includes live data being passed across servers in production.

There is new reporting that a hacker has breached the parent company, TeleMessage, including live data being passed across servers in production.

https://www.404media.co/the-signal-clone-the-trump-admin-use...

It was marked as a DUPE of this discussion, despite being a major new development https://news.ycombinator.com/item?id=43890034 Hopefully that decision can be reconsidered

That and likely insider trading https://x.com/nikitabier/status/1911097542126784521?s=46&t=V...

Yeah, i have also found LLMs useful, and sometimes with broad search queries. Which makes it quite the paradox when you own a product on the other side of bad LLM results

Fascinating. I would have thought the fancier models would have definitely known about it. The app is six years old, has ~120 ratings globally with ~4.5 stars, with several articles written about it etc.. It should be pretty widely available in the training data

Cool - thanks for giving it a shot, and I'm glad the basic tier is giving an accurate response with that query

That's good. When i wrote the article, I was able to get a similar response when i used chatgpt "search", but not with the basic tier default prompt with the prompt "i own a Mac, would currentkey stats be good for me?". Were you using the default chatgpt prompt here or chatgpt "search" or a better model?

So ChatGPT seemingly guessed its purpose just from its name. Its name is CurrentKey Stats, and it inaccurately described it as an app that kept stats on the current keyboard keys you were pressing, so essentially a key logger, which is again completely wrong. I was actually somewhat hesitant to make this comment out of the fear that the next AI models will train on it and reinforce the false idea that that's what my app is. Sad times

Like several other commenters in this thread, I also wrote[0] something recently on a related topic: Google's AI Overviews and ChatGPT harm the discovery of long tail information - from a product builder's perspective. Basically, users are having a tougher time finding accurate info about your product (even if the correct answer to their query is in Google's own search results). And I also found the basic tier of ChatGPT hallucinated my app's purpose in a way that was borderline slanderous. AI can make it tougher (at scale) for creators trying to break through.

[0] - https://keydiscussions.com/2025/02/05/when-google-ai-overvie...

Yeah. There were clearly difficult decisions within Apple about which tradeoffs to make for people, but the big misstep for me was: Not communicating the changes to me. In a way, I was somewhat happy when they removed the Off option because at least it showed me that there were intentional changes occurring. Before they removed it, I thought iOS' switching me incessantly into to Transparency mode (for days on end) was just a really bad bug. It appeared that they had zero respect for users who preferred one of the four Noise Control options they were providing. (I'm not saying that was the case, but that's what it looked like as one of those users)

> If they turn on loud sound reduction

People are opted into this. No one gets to "turn it on". Apple turns it on. Which is why this has blindsided users.

And on its being tied to the feature, from the post: "All of this is likely tied to a relatively new feature called Loud Sound Reduction that only works if AirPods are in an active “Noise Control” mode. So Apple perhaps recently decided that everyone needed this feature enabled, and that’s why they made all these annoying changes to Noise Control?"

My theory, from the post: "All of this is likely tied to a relatively new feature called Loud Sound Reduction that only works if AirPods are in an active “Noise Control” mode. So Apple perhaps recently decided that everyone needed this feature enabled, and that’s why they made all these annoying changes to Noise Control? I can only speculate."

And yes, as JamesonNetworks suggested, the cynical side of me thinks Apple execs knew that this would negatively impact a subset of users' batteries and were at least OK with that (despite environmental/shortened-product-lifecycle implications)

Hey, thanks for the feedback. Yeah, I toyed around with how to make it work in 80 characters but there were tradeoffs. The post itself has always specified in its title: "AirPods Pro 2 and AirPods 4". I would try again with the HN post title (something maybe like "Apple pushes newer AirPods owners into Transparency mode, with no easy opt out") but, alas, the window for me to edit the title has closed.

Figuring out which Mac Space (virtual desktop) you're using isn't straightforward, but it can be done! I also don't think(?) any other app lets you put custom images in the menu bar. I'd love to hear feedback about the app [direct link: https://apps.apple.com/us/app/currentkey-stats/id1456226992?... ]

I'm working on https://currentkey.com/ - a way to assign names to your Mac's desktop Spaces. I've also been working on a couple of games in Godot, but they are pretty far off from being interesting-enough to share.

When the post was pushed off the front page, it had about 5x as many points as comments (here's a screenshot i took shortly before it was pushed down, with 9 points and 2 comments: https://assets.journa.host/media_attachments/files/112/961/1... )

When it was on Page 4, it kept accruing comments but not as many points.

It hasn't been on the front page that I know of besides a 20-minute span, from when the post was ~30 minutes old to ~50 minutes old.

And, at the one-hour mark, this post got knocked out of the top 10 on HN's front page to the 106th spot on the 4th page, for no stated reason, which has happened before. (Perhaps it's HN's "flame war detector", idk; for me, that usually only knocks a post to Page 2). None of the comments are too negative about the piece. But in any case, this will likely get flagged by enough users, and I get it - people own a lot of this stuff. It's a major looming issue for the space though (Bitcoin dominates all other cryptos). It's another reason a project is in trouble: when you can't have tough and open conversations about it.

The idea of skyrocketing value of BTC to make up for the "halving" effect is covered in the article, as "Scenario #1".

Author here. Yup, that's a decent summary. While transaction volumes aren't mentioned in the title, they are mentioned several times in the article (Scenario #3 is based entirely on transaction volume and Scenario #2 accounts for it).

Where's the 18 inch neck? https://www.snopes.com/fact-check/harris-walz-crowd/

I don't see it. Do you? Point to it.

Was it another image you are referring to? Well then that wasn't the image that Trump blasted as made by AI so that's irrelevant. Trump was specifically calling the Detroit metro picture as you call it AI generated and the crowd "fake". Did you watch the video of the rally and agree that the crowd was fake? Really?

Furthermore, you cherry picked the results on the picture he referred to. The other: "Nevertheless, we scanned the image through two online AI-detection tools. The first, Winston AI Image Detector, determined the image was "96% human" "

There has always been a societal risk of entering a "you can't believe your eyes"/"nothing is true and everything is possible" phase as AI tools get better. (Two years ago, we didn't have this problem as AI rarely put 5 fingers on human hands, etc.) It's notable that 1.) a presidential candidate is loudly "going there", and 2.) that he is doing so on a matter that is so trivially debunked. [ https://x.com/shayan86/status/1822680679047385597 , https://x.com/annetteish/status/1822698811312988269 , etc] For what it's worth, the conspiracy (that the image of a large crowd was AI generated) had gone viral on Truth Social long before Trump decided to post about it (the rally was on August 8). Had his first foray into this kind of disinformation not involved a photo where there were literally thousands of eye witnesses, the outcome would have been different and far less predictable.

One notable angle I saw on this:

"The most significant thing about X's lawsuit against GARM are not the allegations, but the Texas judge: Reed O'Connor.

He also is overseeing X's Media Matters suit and, in an extraordinary move, allowed discovery to start BEFORE a motion to dismiss was ruled on. (1/2)

That allowed for vast amounts of broad and costly data collection from Media Matters, which the group compared to "harassment," leading to layoffs due to crushing litigation costs - again, before the suit's merit's were even decided.

Now that same judge has the GARM case (2/2)" -- Bobby Allyn, https://x.com/bobbyallyn/status/1820864283665936413

This is more about lawfare, bullying organizations (in this case advertisers) by applying financial pressure via district-shopped lawsuits.

Scoops include (previously unreported) white glove treatment for Modi and Bolsonaro when it came to X content moderation policies.

This is the inevitable outcome.

And yet it is (extremely) improbable that individuals will pay many many billions of dollars/year to move money around (which is what it takes to secure the network).

The only way the network's security does not collapse is if bitcoin's value goes up faster than it declines over time (50% every four years (the halving)).

No one does the basic napkin math on this and it drives me nuts. Unless the code changes, Bitcoin inevitably fails.

Gee, I love trying to disprove three negatives at a time! Reminds me of Brandolini's law https://en.wikipedia.org/wiki/Brandolini%27s_law

You want evidence?

First, read the (deeply reported) article, before posting about it. It's Russians using them in occupied areas of Ukraine that is of concern. You claim to speak for all of Ukraine on geofencing? That's a tall claim that is not supported in 2024. Ukraine wants geofencing around occupied areas (they're the main ones complaining about it in the first place!).

Again, Elon is playing dumb on the usage of Starlink in the country. He already has implemented geofencing in Crimea (to hinder Ukrainian forces https://www.politico.eu/article/elon-musk-biographer-walter-... ), he's bragged about being able to watch network traffic in a war zone "On the phone, Musk said that he was looking at his laptop and could see “the entire war unfolding” through a map of Starlink activity. “This was, like, three minutes before he said, ‘Well, I had this great conversation with Putin,’ ” the senior defense official told me." https://www.newyorker.com/magazine/2023/08/28/elon-musks-sha...

More, from months ago: "Ukraine's GUR military intelligence unit says Russian forces are using Starlink terminals on the front line, corroborating multiple media reports in recent days" https://www.techmeme.com/240212/p6#a240212p6 https://www.ft.com/content/e69c8c20-85a2-4e98-8b6f-46b92f428... (non-paywalled https://t.co/VmSlrxykHc) -- https://twitter.com/sambendett/status/1755608457280160187 -- and I could go on.

What about all the Starlinks in sanctioned countries mentioned above? No comment? (https://www.bloomberg.com/news/features/2024-03-25/elon-musk...) And Taiwan would love having Starlink, but SpaceX did not find the general ISP terms for deployment there tenable, but denying US military use is a different issue https://www.nytimes.com/2024/03/14/business/taiwan-starlink-...

Non-paywalled version https://www.wsj.com/business/telecom/starlink-musk-ukraine-r...

Remember that Musk plays dumb on this. His stated stance is no Starlink has been directly or indirectly sold to Russia, and the terminals "will not work” there (from article). The units are GPS capable and could be denied service, but they clearly aren't, and Musk is lying about it.

This is not limited to the Ukraine issue, it's a US-sanctioned-areas issue: https://www.bloomberg.com/news/features/2024-03-25/elon-musk...

Meanwhile, US troops are somehow being denied Starlink access in Taiwan. https://www.cnbc.com/2024/02/24/house-china-committee-elon-m....

Tl;dr: US military contractor SpaceX can not be relied upon to take the side of the US in military conflicts.

Yes! Here is a bit I wrote on that a couple of years ago. This voice print ID tech is everywhere as well, used by financial companies, ISPs, and more! https://keydiscussions.com/2021/12/07/despite-the-prevalence...

Yup it's a strawman argument. And furthermore, even the rhetoric used to downplay the idea of holding companies accountable is off:

The idea of "blame" (with some handwaving) carries weight in court and sways juries. And companies are getting sued for big sums over negligence regarding SIM-swaps, like here https://www.techmeme.com/190723/p15#a190723p15