It is surprising how many exceptionally strong mathematicians have started working for OpenAI and Anthropic. These people would have easily become professors at top universities if they stayed in academia. I think many mathematicians, especially the competitive ones at the top, have a “progress at any cost” attitude (and I’m sure the paychecks helped). As for the results, you still need good mathematicians to sift through all the output to identify that the proofs are valid.

I would honestly be positive about universities developing their own specialized math AI (in an ethical manner) to help mathematicians get these kinds of results, but right now, AI is inseparable from these evil companies. Thankfully, I believe this is a likely outcome in the future because the AI companies will one day implode.

From what I’ve seen, most prompts are in plain English. I suppose the part where the AI parses the statement correctly is much easier than the part where it boils a couple lakes in the process of bashing its head against the wall trying millions of different combinations of random shit from the literature to slap together a proof. For one of the big results (cycle double cover), the prompt specified that the AI could use 64 subagents and was required to not give up for at least 8 hours. The tokenmaxxers would be proud, we didn’t need that forest anyway. Thank god math doesn’t have a CTO to look at the expense reports.

Among the three big results I’ve looked at (unit distance problem, cycle double cover, Jacobian), two were counterexamples and one of them had a short 3 page proof using ideas from the 1970s. The Jacobian conjecture is an extreme case because a single counterexample is enough (for unit distance, you technically need a family of counterexamples), and it is easy to check with very basic computations. It is telling that all of these announcements came from OpenAI or Anthropic employees, who presumably have unlimited access to their AI. Nobody really knows how many resources they spent on this, or what else they tried. Nobody really seems to care about this question, either.

I think there is a phenomenon where supposedly hard questions are much easier than expected, because by chance nobody found the right approach for a while, and eventually it becomes famous as a “hard problem” which makes nobody want to attempt it.

What I’m more worried about is many people starting to use AI to try and prove small lemmas for them in their projects. Of course, a $200/mo subscription is absolutely necessary to them. This honestly feels like a repeat of Claude Code back in February. The software engineers eventually realized that AI is absurdly expensive after the AI companies realized that spending $14000/mo to service a $200/mo subscription is a bad idea. If the AI vendors couldn’t squeeze money out of rich software companies, what exactly are they gonna get out of poor mathematicians and universities? Also, there is the cognitive decline caused by overuse of LLMs that has yet to set in.

I think a serious possibility is that AI generated papers flood the zone with uninteresting incremental results that are eventually meaningless and full of mistakes. Right now, math is full of smart, dedicated people, so at least major results are reviewed carefully. But as AI alarmism drives away many honest people from the field, the remaining mathematicians will be burdened with far more work to review, and their cognitive faculties will be eroded by LLM use. Despite 4 years of development, $3 trillion of debt, mountains of stolen data, all the agents and harnesses and loops and other expensive tricks, as well as the advantages of Lean in math research, LLMs still hallucinate.

I believe this is happening with software, but at least there are objective consequences for screwing up there (guy gets his home directory deleted, email is sent on a guy’s behalf without permission, small business gets every customer subscription cancelled). But nothing bad happens if there is a mathematical mistake in a paper and nobody catches it. One could say to just provide a Lean proof, but there is still the issue of making sure the Lean code actually matches the content of the paper. Exactly what force will correct things?

Still, I don’t think this is the most likely possibility. The AI companies are extremely unsustainable financially, and it’s not like they’re very popular. Once they collapse, I believe there will be a re-evaluation of how LLMs should be used in research. If they are used (let alone trained), someone is going to have to pay the bills.

In the end, we have to ask ourselves the question of why one does math. To me, math is not really a field where you memorize trivia. The real value comes from being able to think abstractly and rigorously from first principles, and from understanding why something is true rather than just knowing it is true. It is another aspect of your ability to reason as a free human. A few dedicated people go into math research, but your skills can easily go to many places. If you’re starting undergrad, you have plenty of time to see how this all pans out before making a decision.

Here’s A Taxonomy of Omnicidal Futures Involving Artificial Intelligence by Critch and Tsimerman, which approvingly cites our favorite AI 2027. Nobel disease is still a thing (I think the rationalist beliefs are quite pervasive among mathematicians).

I find it really funny how after he gets booed he says, “If you don’t care about science, that’s okay, because AI is going to touch everything else as well. Whatever path you choose, AI will become part of how work is done.” Yeah, if you’re worried that AI is only going to fuck up science, don’t worry, it’s going to fuck up everything else as well. Was he trying to stick to a (terrible) script, or is he genuinely this incapable of reading a room?

“When someone offers you a seat on the rocket ship, you do not ask which seat. You just get on.” No, my mom taught me about stranger danger. I know what to do when a sketchy old man named Eric Schmidt pulls up with a rocket ship that says FREE ICE CREAM.

“The rocket ship is here. Let me give you some advice. First, find a way to say yes. Listen.” Thanks for revealing how AI adoption is really about coercion. It doesn’t matter what you think, AI is inevitable and you ignorant Luddites are gonna have to find a way to like it.

Truly a masterclass in public speaking by Eric Schmidt. When the audience reacts negatively to what you said, just double down and shove it down their throats. You’re a billionaire, so you know better than them.

How many people, if they were given $1.3 million just once in their lifetime, would figure out far better uses for that money than this guy?

The last several years have been the monkey’s paw moment for rationalists, where they keep getting what they want and realizing it’s actually bad. As for why they keep getting what they want, just look at who’s funding them.

(Also featuring a “Chinese curse” that isn’t actually a phrase in Chinese. At least it’s not “may you live in interesting times”.)

I attended a town hall hosted by the department at my university supposedly for general discussion about department affairs. Considering the university had recently made moves such as adding “AI” into the very name of the department, I had suspicions that much of the discussion would be about AI. (I realize I’m doxxing myself but whatever.) I mostly came for the free food, but I was also interested in seeing what people thought about AI.

The event started with a talk by a prominent professor with major administrative power in the department, and indeed the talk was mostly about AI. His views were that he personally didn’t like AI, but he believed that it had changed the world (particularly in programming), and that it was going to stay. One of his justifications for pivoting the department to AI was ensuring universities had some say in AI and not letting all the control go to unaccountable corporations.

The reaction from the audience was a pleasant surprise to me. He asked everyone how much they were excited about AI (hardly anyone) and how much they were worried (most of the audience). By far the most amusing moment was when someone asked, “What if the assumption that AI is inevitable is wrong? What if AI does not live up to its promises?” (Sadly, I don’t remember the exact words that the person said.) The professor’s response was that by this point, there are so many trustworthy, smart, prominent people who definitely wouldn’t fall for scams, and they have adopted AI. He trusts those people, so he trusts that AI is genuine. I don’t know if the audience member accepted this explanation, but I hope not. Our modus operandi is FOMO.

The pizza was only ok, not really worth a 90 minute event.

This really goes to show how much they need to rely on the LLMentalist effect, despite the AI boosters insisting that the AI is totally different now, everything changed in the last few months. They do not care about creating a useful, reliable tool. That concept doesn’t even occur to them, since why do that when AI is magic?

In any case, they are incapable of creating a useful, reliable tool. Deep down, the only thing the AI companies have at their disposal is the ELIZA effect. OpenAI has every incentive not to truly eliminate AI psychosis, because they need engagement. They only want to mitigate the extreme cases where people go insane and cause bad PR for them. But mild AI psychosis is totally fine, it’s great when people are addicted to your product and make the numbers go up!

Somehow this is no worse than his usual fare, such as a thumbnail that is just a bunch of colored lines resembling a line chart but without representing any actual data, with some random marked points labeled “Dark Farms” and “Human Zoo”.

No, I’m not kidding.

Unfortunately, our problem right now is not Donna the below-average Democrat but Donald the fascist. And when it comes to fascists I do not ask if they are above or below average.

The fire code thing really is an excellent example of LessWrong Brain. Fire truck drivers insist on needlessly large trucks (no citation) which makes roads 30% wider than they would otherwise be (no citation) which has “probably” “non-trivially” contributed to larger cars (no citation) leading to enough additional road fatalities to cancel out the lives saved by stricter fire codes (no citation).

The LessWrong Brain argument starts with a deliberately contrarian conclusion and proves it with a Rube Goldberg chain of logical syllogisms. Of course, citations are strictly optional, and they are free to misinterpret them as they see fit. The only real standard of each claim is “looks good to me”, but you are supposed to be impressed that they managed to string a dozen of them together to reveal some shocking, deep truth of the world that nobody else knows about. The AI 2027 nonsense is an infamous example of this.

He uses the word “fermi” which is cult jargon based on Fermi estimation, a.k.a. guessing shit with back-of-the-envelope calculations. Not exactly what you want if you want to convince people to reform fire codes, especially if you have zero citations for anything.

I guess people just aren’t rational enough, and the only reason the fire codes are so irrational is because people are emotional about fire codes. Firefighters are apparently revered as heroes, when it is the LWers who should be the heroes. After all, firefighters merely save people from fires, while LWers buy multimillion dollar mansions to talk about saving quadrillions of hypothetical people from hypothetical basilisks!

It’s fine, spyware is only a risk when it’s bad people’s spyware. It’s totally fine when it’s Anthropic™-approved spyware!

As for Mythos catching things, maybe they should have used Mythos on their very own Claude Code considering that it has hilariously obvious security exploits, such as this one which inserts an arbitrary string into a shell command. Actually, never mind I don’t see anything wrong here, maybe we should burn another $20k in electricity running Mythos on it again to find out.

In basically every case in history where people decided to kill a bad king, there was a period of chaos and violence that followed it. The killing of Charles I happened during the English Civil War, and the killing of Louis XVI happened during the French Revolution. This has happened many times in Chinese history, with the fall of an imperial dynasty leading to several decades of civil war (most recently in the early 1900s). But I guess if you have a big clever brain with big clever thoughts, you don’t need to look at history.

If the only way to get rid of a bad king is to kill him, he will do anything he can to defend his power, including using as much violence as necessary. (People generally do not like being killed.) Even if you successfully get rid of him, good luck establishing a proper government afterwards with all the violence you’ve caused. And who knows if the new king is gonna be better or worse? A better system would instead have a mechanism that replaces officials on a regular basis, say every few years, and ensure that these replacements are peaceful. Oh wait, that’s liberal democracy. If we do something boring like support democracy, how will people ever think of us as special, clever thinkers with bold, contrarian thoughts?

It’s still One Person. A mortal, fleshy person. Their defence is that they’re inoffensive, things are stable, nothing is directly their fault and people are bound by law and oath.

Bro, your system involves giving all the power to one person. You cannot then say they have no responsibility or that they’re “inoffensive” when they abuse it.

I’ve seen this story play out in software engineering: people were very impressed when the AI does unexpectedly well in one out of 50 attempts on an easy task, and so people decided to trust it for everything and turn their codebases into disasters. There was no great wave of new high-quality software. Instead, the only real result was that existing software has become far more buggy and insecure.

Now we have people using AI in science and math because it was impressive in random demonstrations of solving math problems. I now have friends asking me why I’m not using AI, and also saying that AI will be better than all mathematicians in 30 years or whatever. Do you really think I refuse to use AI out of ignorance? No, I know too much about it! I have seen the same story play out in software engineering, and what makes this any different?

I think the main difference here is that breaking RSA now just requires scaling up existing approaches, while breaking LWE or anything like that would need a major conceptual breakthrough. The former possibility is much more likely, and in any case, cryptographers are the most paranoid people on the planet for a reason.

Unfortunately, one can never be sure about much in cryptography until P vs NP is solved (and then some).

(Of course, just because some people say that scaling up is enough doesn’t mean it’s actually true. For breaking RSA, we know have Shor’s algorithm, while the only evidence AI bros have from superintelligence coming from scaling is “trust me bro”.)

This is what happens when your worldview is based on anime.

(A lot of anime has heavy themes, but most people understand that it’s not real life, just like all such art. Unlike Yud, most people’s worldviews on coding and math are based on actual coding and math.)

We can see that one 9 of availability is 90% = 0.9, two 9s is 99% = 0.99, three 9s is 99.9% = 0.999, etc. In general, for positive integers n, n 9s of availability is 1 - (1/10)^n, and we can extrapolate that to non-integer values of n. The value γ needed for 87.5% availability is the solution to 1 - (1/10)^γ = 7/8, or γ = log_10(8) = 0.903089987. γ is transcendental by Gelfond-Schneider (see this for a reference proof).

Right now, Sora is at zero 9s of availability.

By far the dumbest “feature” in the codebase is this thing called “Buddy” (described in a few places such as here). Honestly, I don’t really know what it’s for or what the point is.

BUDDY - A Tamagotchi Inside Your Terminal

I am not making this up.

Claude Code has a full Tamagotchi-style companion pet system called “Buddy.” A deterministic gacha system with species rarity, shiny variants, procedurally generated stats, and a soul description written by Claude on first hatch like OpenClaw.

…

On top of that, there’s a 1% shiny chance completely independent of rarity. So a Shiny Legendary Nebulynx has a 0.01% chance of being rolled. Dang.

Great, so they were planning on a gacha system where you can get an ASCII virtual pet that, uhh, occasionally makes comments? Truly a serious feature for a serious tool for the serious discipline of software engineering. Imagine if IntelliJ decided to pull this bullshit.

But also, Claude Code is leaning hard into gambling addiction — the “Hooked” model. You reward the user with an intermittent, variable reward. This keeps them coming back in the hope of the big win. And it turns them into gambling addicts.

The Onion could not have come up with a better way to illustrate this very point.