an0malous

I’m not an AI researcher, but it seems like there’s a ton of waste having a universal model that knows everything when any individuals use case requires like generously 10% of what’s stored in the model. Does it even need to have memorized knowledge stored in the model or could it just look up info and docs like humans do? If all you need is the language and intelligence, I think Opus5.5 equivalent intelligence will run on an iPhone within 5 years.

This is the kind of creative work that I think will always need to be guided by human taste and vision but can be vastly enabled by AI. I really hope to see more people using AI to improve quality instead of velocity.

This is the standard VC enshittification playbook, people predicted this years ago. You subsidize prices with funding until you establish a monopoly, then you raise them as high as your customers can afford. It’ll continue to get worse from here.

I use the latest models every day for app development, it does still make many mistakes and poor decisions. Just today I dealt with an issue where it allowed a silent failure that would delete all of the users data without anyone noticing. The code it writes for one task is usually pretty good, but it still doesn’t always follow conventions well even if you’ve documented them. It also has no sense for long-term architecture or organization, or when to make tradeoffs for less complexity because it’s a temporary or prototype feature that needs to stay easy to change. I’ve found that if I do even 3 months of pure agentic coding with no code reviews, it’s aged 10x faster than a human coded codebase so it’s like a 30 month legacy codebase now. I know people who are obsessed with AI coding and they’re throwing away projects they’ve built over a year and starting over because it’s become too slow to make changes.

What kind of coding are you using these models for? Most of the people I know who share your perspective never go beyond the prototyping stage. I’d be curious to hear from anyone who’s been AI coding for more than six months, shipping it to real users, and isn’t looking at their code at all.

I don’t think it would have been released as a blog post with no details if it was done the old fashioned way. My point is that there isn’t even enough information for an intelligent person to glean anything from this besides for “AI did physics”. If there was a paper I could at least answer the question I was wondering about, “what form is the solution in?” It’s about transparency and the lack of it in all these AI company posts.

It’s interesting that you won’t say it’s wrong.

> Another approach is to follow that word, heresy. In every period of history, there seem to have been labels that got applied to statements to shoot them down before anyone had a chance to ask if they were true or not. "Blasphemy", "sacrilege", and "heresy" were such labels for a good part of western history, as in more recent times "indecent", "improper", and "unamerican" have been. By now these labels have lost their sting. They always do. By now they're mostly used ironically. But in their time, they had real force.

>

> The word "defeatist", for example, has no particular political connotations now. But in Germany in 1917 it was a weapon, used by Ludendorff in a purge of those who favored a negotiated peace. At the start of World War II it was used extensively by Churchill and his supporters to silence their opponents. In 1940, any argument against Churchill's aggressive policy was "defeatist". Was it right or wrong? Ideally, no one got far enough to ask that.

What You Can’t Say https://paulgraham.com/say.html

I don’t understand what the model actually did here or how it was verified, and I feel like that’s a general problem with all these “AI did breakthrough science” posts. I understand that I’m not an expert in these fields so there’s only so much I can pick up from a blog post about it, but these posts are basically always just “trust us, it did the thing.” I mean surely you could easily share the conversation trace and any artifacts/code it produced so peers in the field can verify the work.

> These scattering amplitude formulas are hard to compute, so hard that physicists almost always use approximations. They do partial calculations, cut off at a specific number of “loops,” a measure of how complicated interactions between particles are allowed to get. The more “loops” they include in their calculations, the closer they get to the real answer, and the harder, computationally, the calculation is to do.

I don’t even understand what type of solution we’re describing here, is it a formula? A program? A Lean proof?

I hadn't connected the dots until you mentioned it, but yeah this keynote indirectly says it's the end of Ruby and Rails. Ruby was designed to be coded by humans, and DHH says it's the end of coding by hand.

Interesting. Is your 4 year old raising capital?

What are the alternative AI solutions?

The enthusiasts are founders, executives and investors who don’t actually have to use AI to build products and therefore have a somewhat distant opinion about its real capabilities but will capture all gains from AI. They only build prototypes and rough drafts and don’t understand that AI makes the first 80% easier and the last 20% harder, and makes maintainability a nightmare where a codebase that’s gone through vibe coding for 3 months is like the equivalent of a 10 year old legacy codebase.

The burnouts are employees who are being asked to produce 10x the output, do their own project management, design, QA, devops, customer development, work long hours to make up for the gap in how much executives think AI boosts productivity and how much it actually does, and for 0.05% equity at best.

These are the pre rug pull prices. They'll increase prices 10x and nerf the models after they IPO.

Oh interesting, thanks

Why does Cloudflare Warp block archive.today? Did they just get strong-armed by publishers or ISPs or some other reason?

If you ignore what people say and instead observe what they do, the world makes a lot more sense.

I meant the modern community, I think they adopt the same “Rationalism” name. Basically the LessWrong community but HN is also pretty adjacent.

I used to read a lot of LessWrong, Yudkowsky, SlateStarCodex. Rationalism is really just another religion, with the faith parts carefully tucked away underneath concepts like priors or how one chooses to construct a model or assign/“update” probabilities. Their god is ASI. Their beliefs have no more fundamental basis in truth than any other faith.

There’s also a fundamental problem with the idea of using models and probabilities for decision making in that one has to choose a model before deducing any facts, and this leads to all sorts of post hoc rationalization. It’s basically just a culture of people with weird preferences coming up with mathy rationalizations for them.

The one RubyGems hack post alone is another 96) points

I don’t know who you’re talking about, even the most bearish people like Gary Marcus and Ed Zitron acknowledge that LLMs are useful in these same cases the OP admits. Gary Marcus is even still a long term AI advocate, he just doesn’t think LLMs are enough and we need more foundational breakthroughs. Zitron says it’s valuable technology but not worth the trillion dollar valuations the frontier labs are claiming.

The lack of temperament is very skewed towards the bulls who have been saying AGI is here, software engineering is solved, mathematics is solved, it’s going to destroy the white collar job market, and it’s going to kill us all for like 5 years now.

Which part of the headline is misleading? It is factually correct that Irregular was involved in incidents with all three AI labs.

It must have been reinstated because it was off the front page for a full day and suddenly back up in the last hour.

Why was this post flagged? This site has become ridiculous, people are routinely abusing the flagging system to take down posts they don’t like even if they’re obviously on topic and relevant to HN. And it seems like some users have substantially more flagging weight because these posts, likely this one, are often top 5 on HN.

Altman said it first: https://news.ycombinator.com/item?id=49652270

and there’s diminishing returns to larger models and more training data, and there isn’t more training data anyway

Related post from yesterday:

OpenAI considers slowing advanced AI development, Sam Altman tells employees

Link: https://www.bloomberg.com/news/articles/2026-09-11/openai-is...

HN Post: https://news.ycombinator.com/item?id=49652270

How many major companies are pushing unreviewed code to prod? Seems like mostly an early stage startup thing and their bugs and outages aren’t going to make the news. I’m skeptical that even a majority of Anthropic and GitHub’s code new code this year was unreviewed.

> the capability gap is underestimated.

Based off what? There’s pretty precise measurements of the capability gap where open weight models like Kimi K3 score higher than the latest flagship models just 6 months ago.

The one where it’s infeasible that a company that needs to live up to a valuation of like 5% of US GDP but is still burning cash and has no moat would willingly decide to slow down the pace of development of their core technology

“I don’t have anything to deceive you on, of course I’m not deceiving you. I’m engaging with you as a human being. Why would I deceive you? There’s no… no reason for me to do that.”

https://xcancel.com/cocktailpeanut/status/209733229184439951...

Which serious production projects have AI agents coding on their own? And I’m assuming that means they are routinely taking tasks and deploying them to production autonomously

Was that someone Mark Zuckerberg?