Teaching AI. Previously Director of AI for NYC.

NYC
This thing is an absolute beast of a theorem, notorious for its size, and well worth verifying even if the verification is bad (as one would assume it is). It is an amazing achievement that the thing is apparently completely correct.
The largest math proof in human history is now computer-verified. Today, Axiom announces a complete Lean auto-formalization of the Classification of Finite Simple Groups, the "periodic table of symmetry." About 15,000 pages, built by over 100 mathematicians across generations.
1
1
16
689
It’s involved in something called the monstrous moonshine conjecture which was proved by Borcherds in the 1990s. I can’t remember who but an expert on it once described the conjecture as the voice of God.
1
2
69
He had to pull out a bunch of stuff about vertex operator algebras for it, if I remember. Complicated, weird thing.
2
39
Counterexamples in Topology might be the most chaos energy book in all of math. But I agree, counterexamples are very important. It’s good to have a set of them in your head because sometimes you can immediately break a new idea by just testing against them.
I find useless and impractical counterexamples wildly exciting on average because they highlight that the questions themselves need refinement to capture what is truly meant. More interesting theories and questions will doubtlessly fall out of that refinement process
1
1
216
This is more or less the software engineering concept of test driven development, which has relevance far wider than just actual software or engineering.
1
1
85
It is a trick Feynman has talked about, that he had pet counterexamples in his head, and anytime someone proposed an idea he’d scroll through them mentally and quickly. Often suddenly broke something instantly and nobody could figure out how he was seeing it. But it was this.
1
53
There was the old fashioned way, where Feynman was on the committee. Didn’t always end well.
Replying to @gdb
somewhere a PhD student just whispered “please not my thesis”
2
178
16,787
“Did you try X thing off the top of my head that does your thesis better?” “No. I did … I … I did not. Sorry.”
1
29
1,832
Nobody wanted him on that committee. Insane risk.
1
22
1,911
We used to care to build things like this, and it made both our own and others admire us. (This is the Library of Congress.)
1
2
393
It does not read the same today and it does not read the same standing in that hall in front of him.
1
78
Now again we find ourselves on a great battlefield, in a great civil war, and we must revisit these things.
30
I think the fact that Marissa Mayer would voluntarily disclose this to a reporter is a pretty good demonstration of how deeply the culture of Silicon Valley has diverged from the rest of humanity
57
1,030
9,758
170,371
This is a very important and not automatic thing, not just that people love the Mayor but that the Governor does. That relationship is typically very contentious in NY and that he has been able to turn her into a fan and ally rather than an adversary is critical.
New Yorkers love them some Mamdani lol
3
25
597
18,559
You see them together all the time and there is enthusiasm you can’t fake. Bloomberg handled this by being Bloomberg, the 800 pound gorilla everyone was scared of. But if you’re not that you have to do it another way. This is the better way than just having like $80 billion.
2
16
784
I am not a fan of everything he does but this is a major thing to do on behalf of the city. Mysteriously, he somehow also charmed Trump, which is unfortunately also important. I have no clue what the hell he said, but he did something, and that something is doing his job.
11
681
Neal Parikh retweeted
"Everything we love is on the line" in the US election. I argue that President Trump has systematically undone the strategy that since the 19th century has made the US the world's leading country. He has eroded our sources of greatness. So what's at stake isn't just affordability or data centers or the Iran war. It's America's future: nytimes.com/2026/10/24/opini…
74
534
1,176
32,801
Neal Parikh retweeted
Ok I'll bite. AI can point out all the small mistakes a human may miss. But the author can already run it through that same AI first so what's the point? More importantly, AI sucks at assessing how interesting something is. It's god awful at it. We need human reviews still.
New paper: "AI Article Reviews Dominate Human Peer Reviews at Their Best: Evidence and Policy Choices" See: GaryKing.org/peerless. with Steven Worthington. Comments welcome!
16
6
105
5,590