Legend2440

Steam engines are pretty cool.

Oh yes there were. The Jacobian conjecture is "notorious for the large number of published and unpublished false proofs which turned out to contain subtle errors."

It's not quite the Reimann hypothesis, but many prominent mathematicians have spent years working on this problem. Yitang Zhang wrote his PhD thesis on it.

That's what makes it an interesting challenge.

Well, your takeaway is bad. This isn't image generation, it's tool use with a digital paintbrush.

This is a very well studied problem. Well-known mathematicians have spent a lot of effort on it - Yitang Zhang wrote his entire PhD thesis on it back in 1991.

It is deeply unlikely that a random guy at Anthropic just happened to solve it so they could pass it off as the LLM's work.

You do indeed have your head in the sand, that's for sure.

You are desperately searching for ways to excuse it, to explain how it can't be what it obviously is.

It's just a computer program dude. It doesn't have feelings or morals or consciousness.

It's a really complex machine, but it's still a machine.

Just this week an LLM found a counterexample to math problem that's been widely studied for over a century: https://en.wikipedia.org/wiki/Jacobian_conjecture

The conjecture was first stated for two variables by Ludwig Kraus in 1884[1] and then stated in full generality in 1939 by Ott-Heinrich Keller.[2] It was subsequently widely publicized by Shreeram Abhyankar,[3] as an example of a difficult question in algebraic geometry that can be understood using little beyond a knowledge of calculus.

The Jacobian conjecture was notorious for the large number of published and unpublished proofs that turned out to contain subtle errors.[4][5]

On July 19, 2026, Anthropic employee and mathematician Levent Alpöge presented an explicit counterexample in three-dimensional space, discovered by Anthropic's large language model Claude Fable 5, which disproves the conjecture for n > 2

If that won't convince you that LLMs do more than parrot existing ideas, you've got your head in the sand.

Oh, that's bs. You could say that about the tractor, or the mechanical loom, or any of the other incredible labor-saving inventions over the last two hundred years.

We want things to do our work for us so we can do other things. That's not bad, and it's certainly not the same as literally enslaving another human.

I think this is an error in the leaderboard. Looking at the test logs, they had reasoning effort set to 'high'. So it should be in the CoT category instead of the base LLM category.

https://huggingface.co/datasets/arcprize/arc_agi_v1_public_e...

  "kwargs": {
     "max_tokens": 100000,
     "stream": true,
     "reasoning_effort": "high",
     "rate_limit": {
        "rate": 2,
        "period": 60
     }
  }

Everyone? Speak for yourself. I'm not weirded out by it.

but you're not going to make a tiny model perform like one ten times larger just by that.

Small reasoning models do indeed outperform base models that are 10x larger, at these logic/reasoning tasks that require serial computation.

They do not outperform at tasks that rely more on world knowledge or memorization.

In most cases the base model cannot complete logic tasks at all, or only for very small instances; it's reasoning or nothing.

but that doesn't mean they learn any kind of algorithm

They do indeed learn algorithms and can step through them with CoT. This is what allows reasoning models to, e.g. reliably multiply large numbers by applying the grade-school multiplication algorithm.

I don't agree with your definition.

The point of reasoning models is that some tasks fundamentally require a certain number of serial steps. Base models are limited to learning parallel algorithms because of their parallel training, and so struggle on inherently-serial tasks like solving logic puzzles.

The advancement from reasoning is that it allows LLMs to learn a broader class of algorithms.

No, you are misunderstanding the paper.

https://arxiv.org/abs/2607.06764

The base model got 15%. They built an elaborate looping harness that allows it to burn 100k tokens "thinking" about the problem, which got the 57%. This is just an alternative approach to reasoning.

It took a while, but the original ARC ultimately fell to exactly the approach it was supposed to be protected from, i.e. big data memorisation

No it didn't. People tried big data memorization, and it didn't work. Base LLMs (even with millions of synthetic examples) never solved ARC-AGI-1.

It took a real algorithmic advancement - reasoning models - to solve it.

I think you will find that worldwide, family support is more common, not less.

In many cultures multigenerational households are the norm. You may never move out; you raise your kids in the same house your parents raised you.

That's not what your linked article says.

37% of survey respondants say they would need to use a credit card to cover a $400 expense. They wouldn't be destitute, they just don't immediately have the cash.

It also cites another study that put the number much lower, at 8%.

The industrialization of rich nations did not make Africa poor.

Once upon a time, everyone was poor. Then, some countries industrialized and others didn't. The ones that didn't are still poor today.

Some countries like China have recently pulled themselves out of poverty through industrialization. Someday Africa will do the same.

Yeah, people who live in areas that are poorly industrialized (Africa, India, etc) and whose lives have not yet been blessed with automation.

No, I mean the wealth you and I have right now.

All of it - the car, the iPhone, the healthcare, the indoor plumbing, the air conditioner, the closet full of clothes - is only possible because of automation.

But also, it is important to remember that automating and mechanizing is what brought society as a whole out of abject poverty.

It would be very short-sighted to look only at the jobs lost by the few weavers, and not the wealth gained by everyone else.

Prior to industrialization everyone was broke and people regularly starved.

If the loom smashers had their way, we'd still be dirt farmers living in abject poverty.

I can't actually think of a single piece of tech made in the past ten years that has improved my life

This is going to be controversial, but I feel that ChatGPT has improved my life.

Also e-bikes, 3D printers, and wireless earbuds really took off in the last ten years. Some of these were invented more than 10 years ago but saw significant improvements that brought them into the mainstream.

We reading the same media here? All the major tech news publications like the Verge or ArsTechnica take a highly skeptical tone towards AI.

Nearly every major newspaper company is actively suing AI companies over copyright infringement. They have been quite outspoken that they see Google's AI overviews as an existential threat to their business.

That's cynical nonsense and I don't believe a word of it. Prohibition was repealed by popular vote, not by bought-out politicians.

And we're not 'oppressing' fent addicts by arresting them when they pass out on the sidewalk in front of my apartment. They are causing a problem for the rest of society that we now have to deal with. That's why we do the war on drugs.

A harness is just a piece of software that exposes APIs for reading files, connecting to the internet, etc.

It parses the LLM output for tool calls, executes the command, and puts the output back into the LLM input. That's all there is to it.

This is mostly incorrect. The DEA enforces the Controlled Substances Act and the list of scheduled drugs, not FDA regulations.

If you are out there using e.g some experimental cancer drug that hasn't been FDA approved yet, the DEA isn't going to stop you. Their mission is to stop recreational use of drugs that get you high.

The FDA and the DEA have no connection to each other.

The DEA is a law enforcement agency that aims to fight illegal drug trafficking. the FDA is regulatory agency that aims to ensure the food and medicine legally sold in the US is safe.

Governments have spent considerable effort on it, but it's tough to ban something as popular as tobacco or alcohol.

Voters tend to get what they want, and a sizeable fraction of voters smoke or drink.

I would’ve thought that following FDA’s announcement last year, kratom had already been outlawed.

The FDA can say you can't sell it as a supplement or food. But they can't stop you from possessing it or selling it as a chemical.

When the DEA schedules it, it is illegal to possess or sell in any capacity.