The video isn’t about how this is all BS, conclusion from section 1:

They reviewed roughly 1,250 papers on AI self-improvement and found that 74% of them were published in 2026 alone. This field is moving extremely fast. Anyone telling you this is all nonsense has just not read the literature. Is this recursive self-improvement RSI in the bounded sense? Absolutely yes. Al systems are improving themselves within well-defined tasks and the improved versions are producing further improvements. This is a real thing. It is demonstrated that is in the papers and none of this existed 5 years ago.

The caveat in the second section is:

We have much weaker evidence that it can decide what the research problem should be. And that gap, I cannot stress this enough, between optimization and judgment, between execution and scientific taste is the gap that we need to close before anything resembling an autonomous AI scientist could exist. And this is not a small gap

That’s still scary, yeah it probably won’t lead to human extinction, but it has serious implications for both economic and social life moving forward

The actual interview is him talking about how AI is revolutionizing the company, including a slack chat bot that he says is generating 50% of there PRs, and he’s very excited to see where it goes.

He mentions one issue they have had with it and the article takes that and extrapolates that he thinks there whole AI push was a mistake, he doesn’t

Millennials: Boomers had it so easy, they could buy a house for $20k when they were 25

Gen Z: Millennials had it so easy, they could get a job paying more then $20k when they were 25

Anyone got any meth stocks I can invest in?

Gotta get your dopamine from somewhere

Her names actually sailor peg, and she lost her leg.

Yeah, if I have kids and my dad starts talking to them about what “really happened during the pandemic” I’m yanking them out of that conversation as quickly as possible to prevent them going down some idiotic rabbit hole.

Anyone know any studies comparing the energy usage of an AI image generator vs manually photo shopping. If a person spends say a half hour on a computer finding a photo and editing it in photoshop to this quality it could cost as much electricity as an image generator.

I think people are starting to overestimate the energy usage of AI while underestimating everything else we do. Like an image like this would take my computer ~30 seconds to generate, meanwhile I’ll use my computer for hours at a time to play games and don’t think much of that.

if you need sperm source locally

But the quality of sperm in my area sucks, I know , I’ve tried it.

Explain to me how the analogy is wrong then.

I have accepted that llms have hard limits, like you said with regards to information theory and godels incompleteness theorem. I don’t have enough knowledge to articulate those limits, just like I don’t know the physics that prevents you from shrinking the dye further but I accept that they exist.

So we agree there are hard limits, what I am saying is:

• within those limits is a huge problem space
• millions of people are putting labor in that problem space today
• LLMs have not saturated that problem space yet
• LLMs are making gains in that problem space, like with math and reasoning.

I’m not arguing where that “finish line” is because I know I don’t know enough about that. I do understand that 99.999% of problems aren’t passed that finish line and there’s a lot to go before LLMs reach that line.

Almost all use cases for llms don’t require creating novel math techniques, so why are you focusing on them? If “can’t create genuinely new techniques” disqualifies something from being economically transformative or intellectually significant, that standard would disqualify most human knowledge labor too.

Bringing up information theory and godels incompleteness theorem in regards to ai advancement is like saying there are hard physical limits on shrinking the dye in the 1980s. Yes hard axiomatic limits exist, but within those limits is a huge problem space. Even if we limit it to solving problems on a screen using existing techniques, that problem space is huge and occupying a large amount of time for the knowledge workers who make up ~25% of the labor force.

Unlike shrinking the dye though we are not near that limit, we still have a lot of improvement to make before AI can solve all the problems in that space. I think we can both agree the current LLMs aren’t able to replace all screen work. But it is making big strides in that problem space, such as going from failing basic arithmetic to solving the hardest math problems in a couple years.

Like with shrinking the dye there’s not going to be a singular breakthrough, like recursive self improvement, that’ll throw us up against those limits almost immediately. Progress is done incrementally over years with a bunch of different smaller improvements, like opaque recurrence. This does have some acceleration as the technology helps with the design of the technology, computers help chip architects design better chips, which make better computers, but a human is still required. Same with AI, every lab is probably heavily relying on coding agents to assist them with developing and deploying new techniques. The AI may not be creating these techniques, but it is accelerating the creation of these techniques.

How are you defining “new information”? If I ask Claude "what is 235 x 567 + 57,899? " There is a high likelihood that that question was not in its training data, but it still gets it right and generates new information. It will be using existing methods from its training data to “reason” to get the answer, and it hasn’t shown the ability to devise new methods. But 99.99% of the problems in the world can be solved by using existing methods.

Open AI isn’t going to go bust because it can’t make novel math techniques, almost all of the knowledge work requiring math doesn’t require new methods. If it’s able to master every existing technique used by knowledge workers without creating new methods, ie. your finish line, then they’ll have automated a huge chunk of labor and probably become a hugely profitable company.

Recursive self improvement may not be panning out currently so we won’t see any singularity style exponential take off, but we’re still seeing steady linear improvement in tasks such as math and reasoning through other techniques, like the opaque recurrence that astra is using.

Most harnesses that people use to interact with llms have memory and skills.

You can “teach” them a skill which is sort of like a macro so you can say “deploy to dev” and it’ll look up that skill and execute it without any extra input needed.

For memory it’ll keep files on your local of what it’s seen before and can pull from the information in those files if necessary. So if you always ask it to label reports in a certain directory “Midwest expansion project” the harness will add that to its memory files and start to label it automatically.

fact that solving pre-made problems where the solutions are known or knowable isn’t anywhere close to the same thing as asking truly novel research questions independent of a human prompt.

I understand that, but the original statement was about the models stalling out in progress.

Would you say a student stalled out in progress if they could barely do 2 + 2 a couple years ago and is now able to consistently solve some of the most complex math problems known just because that student isn’t creating novel research?

You might want to check up on the current controversy surrounding OpenAI’s math “achievements”.

I’m aware of the controversies, this benchmark isn’t about making new proofs on previously unsolved problems, it’s whether it can answer complex math problems, which it’s getting better at.

By my reckoning, the difference between Mythos and Opus is smaller than the difference between Opus and Sonnet. Same with the difference between GPT 5.5 to 5.6 is smaller than the difference between GPT 4 to GPT 5.

Do you have any benchmarks or data to back this “reckoning”

the technology has plateaued and is beyond the point of diminishing returns.

The technology is still advancing at a good pace and there’s no sign they’ve plateaued, especially in math and “reasoning”. For example the frontier math benchmark is showing big strides for the frontier models. A year ago they were only getting <10% of questions, now sol from openai is scoring 89% and they’re reporting the unreleased astra is at 98% .

They even had to create a new frontier math benchmark called erdos that all previous models scored 0% on and astra got 3% , which may not look like much but is a huge relative jump.

Yeah, so they’re the same. Heard the term from my Indian coworkers, apparently they’re very big in india

Is there a difference between a ghost kitchen and a cloud kitchen?

The pace of advancement is still going strong, especially in math and “reasoning”. For example the frontier math benchmark is showing big strides for the frontier models. A year ago they were only getting <10% of questions, now sol from openai is scoring 89% and they’re reporting the unreleased astra is at 98% .

They even had to create a new frontier math benchmark called erdos that all previous models scored 0% on and astra got 3% , which may not look like much but is a huge relative jump.

The models are still advancing, especially in math and “reasoning”. For example the frontier math benchmark is showing big strides for the frontier models. A year ago they were only getting <10% of questions, now sol from openai is scoring 89% and they’re reporting the unreleased astra is at 98% .