Do you really need Fable level 99% of people don't need to solve 600 year old math problems. People use the best model available because they are scared a cheaper model won't be as good but in the end it rarely matters for 99 percentile of coding.

Bad decisions compound over time, so I want my model making as few of them as possible. Opus 5.5 is the current GOAT, and it's fairly affordable on a subscription. (I wish it was cheaper, but nothing else delivers comparable results for my price/time investment.)

It's FOMO and the never ending treadmill.

The newest model has people go "oh I thought models sucked until THIS one came out." Then when some other company makes a cheaper model that is comparable to it people go "nah this model sucks who would use it over Claude/Openai? Noone can beat them at all because look at this one shot"

This greatly depends on the type of coding you are doing. For easy webdev, sure. For me and my colleagues, LLMs have been quite unhelpful for programming until about five months ago. Mistakes can be devastating, especially at scale. If the simpler stuff works for you, then great. But don't pretend to speak for everyone.

I'm not foolish enough to say that AI is not economically viable. But these two companies specifically, OpenAI and Anthropic, just seem totally screwed. They're making great revenue, but their expenditure is unbelievably irresponsible (actually, their behaviour across the board is irresponsible). Training, testing, and marketing should not be this expensive.

I'm looking forward to the new wave of companies that will likely pick up after them, and engage in much more responsible business practices. These two have got to go.

$72B of Anthropic's loss was on the paper value of financial instruments, though. Their operating loss, including non-recurring engineering/development costs, was under $9B. That's still not great, given $16B in revenue over half a year, even if that was 15x the prior-year revenue. Future infrastructure spending is the one place that they look overexposed: $600B in promised spend over the next decade, mostly not cancelable.

Those financial instruments are the circular financing that's feeding the beast.

I think AI is going to be like television: a hugely important technology, that you can't make much money from.

There's not enough moat and the open models just have to be good enough for people to cheap out.

A lot of the costs are going to compute aka GPUs aka Nvidia. Guess who is funding AI startups so they can buy more GPUs.

And unfortunately, anyone who uses DeepSeek for serious math applications is extremely foolish, at least right now anyway. Disappointingly, the Chinese models are so far behind on math capability that it is a genuine moat at this point.

Totally different story for coding of course.

The financial value you can extract from math capabilities is minuscule compared to software engineering.

Yes, that's about right. It is much easier for it to roughly copy other code that it has seen, than to understand the code it has been given enough to make targeted fixes.

You can definitely still get PCSX2 working on it though, and it's fantastic. PS2 and PS3 games are among my favourite uses for handheld PCs.

I'm normally not the cynical type, and I am appreciative of Karpathy's efforts to educate people on developing language models, but this is exactly how it reads.

Every once in a while, everyone needs to touch grass.

tell that to the c-suite!

Why do you believe that learning Lean is a better use of time when the AI is clearly better at writing and interpreting Lean than it is writing quality papers? At this point, Lean is for autoformalization, no one is really supposed to read it.

Well, you don't necessarily need to read the proof, but the proof is useless if you don't read the specification.

That's fair, but I would argue that reading the specification is very easy by comparison. A quick one hour tutorial is usually enough judging from my students' experiences.

The Lean proof to Fermat's Last Theorem is 13 mil lines.

The one for the quasi-Riemann Hypothesis is half a million.

Indeed! Almost none of these most math folks are likely to read. The only thing to read is the statement, which is only a handful of lines in both cases. The point of Lean is that if that statement compiles and is validated by hand to be equivalent to the natural language statement, then it is true. That was what I was trying to say here.

Do you care to share then? Because anyone I know with any sense right now is aware that "planning for the next forty or so years" is genuinely impossible at this point, especially in mathematics, and the right approach is to remain on your toes, ride the waves, and diversify as much as possible. Heck, even planning for the next five years seems absurd. This is a time of immense uncertainty, and to think otherwise is foolish.

A few years ago I was in a pure math degree, but I switched and graduated with a combined degree in math and computer science. Once I finish my masters degree I intend to work on industry applications of formal methods, which I anticipate will be a high-growth area for the next few years. The idea is to take advantage of the growing ability to generate proofs/code and work on keeping the messy human side of the equation up to speed. So I plan on addressing the social and operational aspects which come with delegating more work to AI. I'll work on integrating these gains into the industry, on building trust in the work that AI produces, and I'll work on the interface that lets us ensure that this trust is deserved.

But nothing is certain. I am trying to avoid overspecialization and remain flexible, just as you advise. If this path fails I might become a paramedic or physiotherapist.

Yes. Most probably do not understand the notation involved in, and the statement of, the Lindeberg-Levy Central Limit Theorem. But every scientist uses this theorem in one way or another. These ideas have a way of trickling down because to people who work thanklessly to do so.

"Plausible" text was preferred over rational text when we trained LLMs using RLHF. It's rapidly shifting the other way now with RLVR, which enforces correctness by default.