Best match combines word and meaning matches. Phrases, hashtags, accounts and websites guide which posts fit; timely queries also favor recent posts. Match labels describe why a post appears, not its quality or accuracy.
What mathematicians should know about the Lean Theorem Prover: questions of reliability and AI. ~ Thomas Hales. terrytao.wordpress.com/2026/10… #LeanProver #ITP #AI4Math
terrytao.wordpress.com What mathematicians should know about the Lean Theorem Prover: questions of reliability and AI
Was talking w some colleagues the other day and we came to the conclusion that, at this point, AI-generated Lean code adds almost zero to our confidence that the stated results are correct and correctly proved. (We already knew it added zero to our understanding.) #MathSky
Ten years ago I dreamed of a world where dependant typed programming would become widely adopted for load-bearing software. Everyone told me that it was a pipe dream, even PhD mathematicians could barely learn to use Lean. Now we can leverage the type checker of lean without ever writing a line...
Sad news for #LeanLang Because of AI? Will AI become a negative force for Lean?
Dunno what the future looks like but being able to brainstorm with agents and verify my proofs in lean quickly means I'll do a lot more math in my work
You can spot who uses AI well when something breaks. One explains why the fix works. The other just says it works. The tool is the same. The difference appears on the next bug. If you can’t explain the fix, it isn’t finished. #SoftwareEngineering #AI #buildinpublic
“0.8 feels safe” 😹 That’s exactly what I wanted to avoid with changelog-bot. John and I expanded the WHY benchmark from 14 to 50 cases, then swept thresholds and compared precision, recall, and F1. Guess less. Measure more. #TypeScript #OpenSource
dev.to How Do We Extract the “Why” from a PR into a CHANGELOG with Jev? 🤔
A few thoughts from the last couple weeks of touching grass and talking to people (mostly talking to people): 1) we should spend more time talking with the median voter. And the 1st and 3rd quartile voters while we are at it.
You mean the Healthcare Plan, with empty pages, was written with invisible ink? 70+ million must have thought so, because they voted for it. And the say they didn’t? What DID THEY VOTE FOR? Hate?
“Here's the REAL conspiracy: Too many people are forced into car-dependent lives, with more health harms, crashes, air pollution, social isolation — and less space for everything…” @grantennis.bsky.social @tomflood.bsky.social @brenttoderian.bsky.social in @usa.streetsblog.org. #1HourCityConspiracy
usa.streetsblog.org Urban Truth Collective: The One-Hour City Conspiracy — Streetsblog USA
"You need to do incremental change" Nothing the current regime has done has been incremental. So why does fixing it need to be a slow walk when breaking it took no effort or time? "Hurr they are breaking the law tho" You care more about the rules than the people they affect.
The Conservative Way.
one of my weirdest little ai experiment ideas is that i would like a text-to-speech model paired with some other small model. the pair watches the lean proof state for a rapidly advancing lean repo and narrates the updates to the proof state like a sports announcer or an auctioneer
Yes on Prop 40. NO on Prop 41. NO on Prop 42. Read up on the rest of the props here: peterates.com/props-1126/ Yes. It’s biased. Republicans ruin everything. Stop voting for Republicans. If you’re not a multimillionaire or billionaire, they don’t care. www.instagram.com/p/DeNDyBwEtz...
instagram.com Instagram
I suspect that I'm still actually better at implementing things well and cleanly than the models are, but at least professionally the demand is usually for quantity over quality at this point. And admittedly, if I have seven ideas and don't know which is worth the effort, the slop-shotgun helps.
6 months of agentic AI with 75 senior devs. +75% stories, −40% bugs. And that is where the real problem started. Not the bugs. From my keynote to 3 articles, ending with our agentic software factory. Part 1: medium.com/@tpierrain/t...
medium.com The hardest part begins when it works.
I cut a video about it today. I like Mr. Levy and kept my mouth closed despite my suspicions but yeah, obvious what it is.
What if AI didn’t write the “Why” in a CHANGELOG at all? I’m using Jev to select rationale already written in the PR or return nothing when the evidence is weak. No invented WHY. 😸 dev.to/nyaomaru/how... #TypeScript #OpenSource
dev.to How Do We Extract the “Why” from a PR into a CHANGELOG with Jev? 🤔
the idea we're going to have like 100x software but with *fewer* people is very funny to me, very funny to think that and be like oh yeah thats going to be good
Search looks through the last 36 hours of public posts for your words, "exact phrases", #hashtags, @handles and websites, and also finds posts about the same thing by meaning. New posts join about every 10 minutes. We only include posts whose authors allow logged-out viewing (Bluesky) or search (Mastodon, indexable accounts), and we re-check Bluesky authors each time we show a page. Likes and follower counts never decide the order. For meaning matches, query text is turned into a fingerprint on Wildroot’s own server (EmbeddingGemma-300M); it is never sent to another service. Wildroot does not save query text to disk. Looking for posts like one you have? Similar posts. How to opt out.