Henry Blanchette
187 posts
Henry Blanchette
@rybl4
Programming Languages PhD student at the University of Maryland. Interested programming languages, formal verification, game dev, game theory.
United States Bergabung Kasım 2019
699 Mengikuti97 Pengikut
@rybl4 The deeper bet is that LLMs handle this better than humans do, because they can reason about the whole slot environment in one pass rather than tracking it mentally. Whether that’s true at scale is genuinely open. Would welcome thoughts if you’ve seen better approaches. (4/4)
English
I built a programming language you're not supposed to read. Vera is designed for machines to write, veralang.dev. No variable names. Compiler errors are instructions for the model, not diagnostics for you. Point a model at it. #AI #OpenSource negroniventurestudios.com/2026/02/28/a-l…
English
Fontgen, a Gemini setup for generating fonts based on natural language descriptions gemini.google.com/gem/1YrAYEpQcE…
English
@Altimor Would you say now that AI programmers are 95% as good as humans?
English
We’ve started planning with the team for the fast-approaching day when AI programmers will 95% as good as humans
If it’s not obvious to you that it’s coming in 1-2yrs, you haven’t been paying attention
Nat Friedman@natfriedman
This is going to be an insane year for AIs writing code
English
@jonatanpallesen Your points are contradictory. If insider trades are revealing accurate info to enemies, they are also revealing accurate info to everyone, since these are public markets. Therefore making the prediction markets more accurate.
English
This is bad. (Assuming it truly is insider trading as it looks like.)
1.
It is literally giving secret military information to the enemy for money.
2.
Insider trading erodes the willingness to place bets on such events, reducing the utility of prediction markets.
Lauren@cabsav456
Something is going on with Polymarket, and we need to talk about it. Last night, a brand new account made over $515,000 betting on the U.S. strike against Iran. The trade was placed 71 minutes before the news broke publicly. For years, Polymarket was effectively banned for U.S. users and under active criminal investigation. In 2025: - Those investigations were closed. - The platform was legalized. - It saw $21 billion in trades (up from $9 billion in 2024). I believe in free markets. But when the President's own family sits on the advisory board of a platform that had its legal hurdles cleared overnight, we need to be asking questions.
English
Henry Blanchette me-retweet
"Package Managers à la Carte, A Formal Model of Dependency Resolution" preprint out today: a new package calculus to describe the cambrian explosion of systems that exist today arxiv.org/pdf/2602.18602
English
Fun bonus tip: regular Claude chat (not Claude Code) has the ability to clone repos from GitHub these days, which means you can ask it to checkout ANY public repo and answer questions about it or even use it as a starting point for an artifact!
Simon Willison@simonw
Added a feature to my blog I've wanted for ages - it now shows my content from elsewhere, including TILs, releases, museums, tools and research, as little badges in the various blog timeline views simonwillison.net/2026/Feb/20/be…
English
In particular, I cover ChopChop: A Programmable Framework for Semantically Constraining the Output of Language Models arxiv.org/abs/2509.00360
English
I wrote about new developments in constrained decoding for LLMs here: rybl.net/post/chopchop-…
English