Hacker News Reader

andy99on Ask HN: Are people generally interested using LLMs for learning purposes?

You don’t learn anything from LLMs. They are good as a reference, not teaching.

They are likely actually worse than other electronic learning (which is already bad) because of how human preference optimized they are, same as eating ultra processed food is easy but not very good for you.

andy99on VibeThinker: 3B param model that beats Opus 4.5 on reasoning with novel SFT+GRPO

Could you please share which coding agent you are using with it?

andy99on FDA advisors unanimously vote to approve Moderna's mRNA after agency drama

For context, it’s a seasonal flu vaccine, the title is a bit unclear.

andy99on Pre-2022 Books

I’m pretty conservative with books and usually only read things based on recommendations anyway. I would rarely read a book published in the last few years just because news of it hasn’t travelled to me yet. I think worrying about AI generated books would really only matter if you’re at the bleeding edge of reading and looking for brand new stuff.

andy99on Big Banana Car

Meh, you might technically be right but the world is better if everyone can have a bit of fun, even police. I’m not super sympathetic to the “I was minding my own business driving a giant banana when the police pulled me over to ask me about it” argument - this guy seems fine with it too, but it’s not like there’s any reason to drive it other than for the attention.

And the “I pulled you over because you peeled out” - I mean it’s fun. Anyway, if it’s harmless I don’t really see the problem.

andy99on How Alberta Eradicated Rats

Maybe I’m just lucky but I’ve lived in Ontario (in cities and the country) and have never seen a rat here. I know they exist but it’s not like we’re inundated by them. Maybe some specific businesses have problems with them, but it’s not something you see every day.

I also lived in Montreal and did see them there sometimes. This was always early morning when I was out running - did not see them in Toronto, Ottawa, etc under similar conditions.

Edit based on reading some other comments: I have seen lots of coyotes and foxes so maybe that explains fewer rats. I know Montreal has coyotes but I’ve never seen them there and where I was there were also squirrels everywhere suggesting fewer predators.

andy99on Midjourney Medical

On the “we don’t know what to do”, I think it’s cool that they are trying something medical with it. Success obviously isn’t assured, probably it isn’t even probable, but I’m happy to see companies try this rather than launch yet another whatever or start a consulting business. I hope as the field (AI generally) matures, more people decide to try life changing stuff with it.

andy99on French physicist and media star loses doctorate after plagiarism investigation

Personally I think writing with an LLM is at least as bad as stitching together phrases from others.

The article doesn’t really expand upon what having fragments copied from others means. Even if it fits the letter of the definition, on a phd thesis that may or may not be a big deal. If he’s passing off the ideas of others as his, or faking his research by using the results of others or making them up, then that’s really bad. If he’s just using phrases / wording from others to get his original points across, it looks bad but I don’t see it as a huge deal, especially 30 years out from the phd.

A PhD is supposed to be original research, if the originality or integrity is in question that’s one thing, the rest is much more pedantic, even if technically wrong.

andy99on A backdoor in a LinkedIn job offer

I once saw an ad on LinkedIn made up to look like the CBC (Canadian news) linking to a fake video of the Canadian prime minister announcing a crypto investment plan for all Canadians, with a link to sign up. I reported the ad to LinkedIn and shortly after got a reply telling me they investigated and didn’t find any violation of their policies.

andy99on New U.S. college grads now have higher unemployment than the average worker

Are there people who think college education is a shortcut to generic employment? This seems like a very misleading statistic. Average earnings (including those unemployed), etc might be better. Telling me that it’s harder to get a professional job that I’m qualified for than it is to walk up to a McDonald’s or whatever and get a job is not shocking.

andy99on Benchmarks in Leipzig

“In the training data” isn’t really relevant for a modern LLM. The better question would be are they solvable using known techniques that have been fine-tuned in.

A simple example, as a non-mathematician: I’d expect a well trained LLM to be able to solve any integral that can be solved with integration by parts. I would be much more interested to see it solve one with no know solution using some novel technique.

Obviously this doesn’t really lend itself to making a benchmark, but if something is solveable by a known technique, and the LLM has has some kind of RL training re using that technique, seeing a solution isn’t too surprising.

andy99on Ultra-processed foods in the global food system: The role of tobacco companies

Don’t buy it then?

andy99on Mathematicians issue warning as AI rapidly gains ground

I’m curious, do writers and authors still really care about AI? I think by now most people are completely put off by AI slop, the value of AI writing or image generation is basically zero

So I suspect that the cloud will pass on math too, initial demos get extrapolated and people get worried but in the end slop is slop and serious people aren’t getting replaced or even threatened.

andy99on Meta workers can opt out of being tracked at work up to 30 min

Not replying to the above comment specifically as I obviously don’t know individual circumstances. But I find it ironic that people working in basically surveillance tech, who would gladly get paid to strip mine users’ privacy in order to market to them - you might say having open contempt for their users, suddenly get put off when the same is applied to them.

andy99on It's Not Just X. It's Y

It’s unlikely this is true. LLMs are way more mad-libs / templates than we like to admit, that’s (ironically) not a judgement about their capability, it’s primarily just an observation. But it’s also what plain old SFT, which I believe is the primary culprit, ends up imparting.

andy99on What if remote working, not AI, is to blame for weak junior hiring?

The best remote jobs I’ve had included many hours a week of no-agenda calls with colleagues, just catching up and talking about what we’re up to. This is very hard to make happen. Most people don’t want to, don’t see it as work, or more likely just don’t know anyone well enough to call and shoot the shit. But imo this is the only real way. Just doing transactional interactions, it’s very tough to stay well connected.

andy99on CAPTCHAs can still detect AI agents

I’ve never encountered a captcha on HN, do you guys use less aggressive settings?

The reason captchas bother me so much is they always seem to happen in the course of legitimate activities. Like I had one when trying to make a charity donation, or ordering something - I have no idea why it would be hard to distinguish such traffic as legitimate, I’m convinced it’s because I’m using a nonstandard browser, not allowing cookies, etc.

If I was trying an automation or to bulk download something or whatever, I’d take the captcha as an interesting professional challenge. When I’m trying to use someone’s services or pay them money, it’s just ridiculous friction and I generally abandon any transaction that makes me do a captcha.

Incidentally I have scraped HN and never encountered any problems, since you have an api for it

andy99on CAPTCHAs can still detect AI agents

Captchas are primarily to punish users for not allowing tracking, or using the “right” services, they may prevent some bots as a side effect (or a pretence from the provider) but it’s mostly for google and cloudflare to abuse their monopolies.

andy99on Sam Altman and Dario Amodei are both walking back AI jobs apocalypse predictions

  I've never seen a company punished for doing the right thing, caring for humans and providing a good work environment.

We’ll see how this goes over but I disagree. You don’t have to look hard in tech, especially a few years ago, to find groups of coddled “workers”, doing very little or at least doing what they want instead of what a business and customers want. This paradoxically ended up creating toxic work environments, and making it impossible to actually get work out of people. We’re seeing a correction now.

andy99on The AI bubble isn't like the internet bubble

  labor-led automation produces improvements in quality, while capital-driven automation increases throughput

I don’t know if this is true, but I do think that LLMs mainly get used where their proponents don’t care (whether intentionally or through ignorance) about the quality of the output, and want to minimize work / maximize throughout. Basically whoever is pushing them is playing the hypothetical role of capitalist in his assertion.

This explains the management push (ignorance) but also the user push (automating BS tasks). The common thread is that the user doesn’t have to take any responsibility for the output. This is why people don’t like having LLMs pushed on them, because for cases where they are responsible for or have to consume the output, they don’t work very well, but when it’s just something that needs to look ok at a glance and be handed off, everyone is rushing to use them.

andy99on Jensen–Shannon Divergence

Iirc (and I could be wrong, this is from memory) JS divergence is what is minimized in GANs (where we simultaneously train a generator and real/synthetic classifier with the goal of each trying to beat the other to converge on real looking synthetic data), at least for some training methods.

I don’t think GANs are used much now in comparison to diffusion models, but as recently as a few years ago they were the standard way to make fake data, a la “this face does not exist”

andy99on Artificial egg hatched 26 healthy chickens

  requires real hen for fertilization and laying

andy99on Not alive, but not dead: disembodied human brains used for drug testing

I had a dentist explain to me the same for getting my wisdom teeth out, as if it was a selling feature. At least for me, having my memory wiped is far more scary than just being put unconscious (or having some pain and a local anaesthetic).

andy99on We let AIs run radio stations

I’m definitely not in the “ai is sentient” camp, but it obviously has personality and emergent behaviours including when left to its own devices. There have been various experiments on this e.g. https://timkellogg.me/blog/2025/09/27/boredom

andy99on We let AIs run radio stations

It’s amazing how many people have completely misunderstood the article

andy99on We let AIs run radio stations

This is a research lab that looks at things like how LLMs perform on long running tasks. The research has nothing to do with replacing radio stations.

That said, the ship sailed a long time ago and has nothing to do with AI (except maybe recommender system). Spotify and competitors are actual automated “radio” stations. IMO, the second worst part (after ads) is the DJ banter, and I like to just listen to music. Before Spotify my digital “radio” was a lot of mp3s and shuffle play. People older than me (or more interested) had multi-disc CD changers.

TLDR, it’s really funny seeing people get up in arms about this experiment stripping the humanity away from radio, when automatic song playing has been a thing since I believe probably before radio was invented. This is about seeing what LLMs do with autonomy on a long time scale.

andy99on OpenAI and Government of Malta partner to roll out ChatGPT Plus to all citizens

Talked to someone at a large company who had admin access to survey results (require to do some analytics). The survey was “anonymous” but results were geo-located, and had some information about the team they came from, which in many cases was enough to clearly identify people. There is a difference between “doesn’t have a persons name on it” anonymous and actually anonymized in a way hardened against figuring out who is who. I don’t think anyone really does the latter.

andy99on The sigmoids won't save you

AI has scaled well according to convenient measures. It (neural networks) have the property that whatever you define, they can rapidly be trained master it. We’re able to show that various tasks of increasing complication do not require intelligence and can be framed as autoregressive RL problems. I personally don’t think AI is any closer to sentient intelligence than LeNet; it’s almost trivially clear, we know how it works. So we’re measuring something orthogonal, basically how well a universal function approximator can fit to a function we define, given arbitrary computing power, and calling that progress. What will be really interesting is if we’re able to find a way to properly measure what they can’t do and what’s different about real intelligence.

Edit: in particular I don’t agree with

  But if someone claims that the trend toward increasing AI capabilities will never reach some particular scary level...

One has to agree that the benchmark results are getting “scarier”, which is not automatically implied by finding more goals to optimize for

andy99on Hardware Attestation as Monopoly Enabler

I hope you contacted them to explain why. People usually think I’m a nut when I do it, or are too stupid to understand and think it’s a tech support issue, but it’s worth at least trying to make it clear that you are choosing not to use/do/pay something because of their choice to use recaptcha

andy99on Hardware Attestation as Monopoly Enabler

Came here with roughly the same thought. Given the stated importance to many of sovereignty and not being dependent on the US, why isn’t there more opposition? I assume it’s just ignorance?