Naka-pin na Tweet
it's 2025, and i'm still writing code that isn't formally verified
English
Raaid Tanveer
425 posts
@AdhvikKanagala well, master is default on git, so using main is cringe (since that means you cared enough to change branch name)
English
@bcherny feature request: make cc request all permissions it could ever need upfront so that i can look over and approve most things it might need permissions for at the start of the session
English
@simonw Coding agents are also great at writing AST walkers.
My biggest gripe with writing AST walkers manually is potentially having to case through every node type. Coding agents trivialize solving this problem.
English
@raaidrt I really like that idea - I've been using auto formatters and a tiny bit of linters but setting up more elaborate AST walking tools sounds like a really good move
English
@simonw re: simonwillison.net/guides/agentic…
One point that you missed that I think more people should be thinking about now is the value of static analysis and lint rules on code quality.
One of the best ways I've ensured that model output variance stays within bounds is by keeping the variety of bad code that can be written to a minimum.
You can simply create an agent skill for introducing new lint rules, or make AST walkers for detecting code smells, and then enforce that these checks be run on every commit.
English
Looking for suggestions on clean proofs that I can write on inductive datatypes.
Here's a simple one relating the size and depth of binary trees.
English
@radokirov probably depends on when a good enough “programming DSL” exists for Lean. I know there are projects like cslib.io, and even the recent github.com/verse-lab/velv… - but still seems much less than general Math support in Lean.
English
Anyone porting software to lean and adding correctness proofs? Wrestling with the compiler/proof checker is not so bad when the LLM is doing.
Andrej Karpathy@karpathy
I think it must be a very interesting time to be in programming languages and formal methods because LLMs change the whole constraints landscape of software completely. Hints of this can already be seen, e.g. in the rising momentum behind porting C to Rust or the growing interest in upgrading legacy code bases in COBOL or etc. In particular, LLMs are *especially* good at translation compared to de-novo generation because 1) the original code base acts as a kind of highly detailed prompt, and 2) as a reference to write concrete tests with respect to. That said, even Rust is nowhere near optimal for LLMs as a target language. What kind of language is optimal? What concessions (if any) are still carved out for humans? Incredibly interesting new questions and opportunities. It feels likely that we'll end up re-writing large fractions of all software ever written many times over.
English
Raaid Tanveer nag-retweet
Velvet, our automated Dafny-style verifier embedded into Lean, has moved to a new repository as a standalone Lean library:
github.com/verse-lab/velv…
Give it a try!
English
@dystopiabreaker you don’t need TypeVar’s in Python 3.14 anymore, just write ‘class Ok[T]:’
English
@AlterranIRL regular languages can be solved with O(1) space in a TM so i feel like this is a moot point - i think what you're asking for is a language with a lower bound space complexity of lg lg n as well right?
English
@TaliaRinger This compiler was an experiment and not intended for people to use. You are right that at this point of history it is premature to use a compiler written end to end by an AI agent. We would probably have to wait until June or maybe even next year.
English
Correctness actually matters for a C compiler, why would you ask a bunch of AI agents to write one with no oversight at this point in history
English
@amorriscode was the original inspiration Claude Shannon? Or Claude Monet?
English
Did y’all know Claude is a Martian?
anthropic.com/features/claud…
English
@bcherny exclusively on VSCode? is this not supported in the CLI?
English
Just shipped in v2.1.19: rewind & fork message history in the Claude Code VSCode extension
English
well, if everyone used this tool and everyone’s SAT scores improved, then the scores would remain the same I think due to fitting to the normal distribution - I suspect it will become more bimodal due to there being groups of people who take advantage of these tools, and those who don’t
English
Average SAT scores are about to not be statistically significantly different from previous years🔥🔥🔥
Google@Google
We’re launching full-length, on demand practice exams for standardized tests in @GeminiApp, starting with the SAT, available now at no cost. Practice SATs are grounded in rigorously vetted content in partnership with @ThePrincetonRev, and Gemini will provide immediate feedback highlighting where you excelled and where you might need to study more. To try it out, tell Gemini, “I want to take a practice SAT test.”
English
Ok this is really cool
I had an HEIC image (Apple's Image format) for a component I wanted to use with my Raspberry Pi, but Claude is unable to examine anything other than JPG. So, it installed a python library for converting the image to JPG, all from the desktop app!
I wonder if it did this locally or on a sandbox somewhere.
English
@patrickc @cursor_ai i’d say opus 4.5 was probably the coolest ai breakthrough since gpt-4 - maybe you wanted to replace “since gpt-4” with “since opus 4.5”?
English
This work by @cursor_ai is, I think, the coolest AI breakthrough since GPT-4. (And there are plenty of candidates!)
simonwillison.net/2026/Jan/19/sc…
English
@TaliaRinger Turn on pageless mode in Google Docs: voila, it's one page!
English
Finished a beautiful four-page research statement for my third-year review. There's just one problem: it needs to be three pages max
English
@collision excellent sequel - but not for the same reasons as you i'd imagine
English
Thoroughly enjoyed the new Knives Out movie, Wake Up Dead Man
• Felt like a sequel in the style of Top Gun 2: just figure out what worked well about the first movie and make a very similar one again. (It’s much better than Glass Onion, which tried to mix it up more.)
• Others have observed this, but these are the most Protestant-acting Catholics I’ve ever seen.
• Was tickled to discover that Daryl McCormack (playing Cy) is a fellow Nenagh native and Gaelscoil Aonach Urmhumhan alum.
English
@ilyasergey @HarmonicMath Hmm for math lemmas needed for algorithmic solutions this is ok, but I think for more general solutions we’ll need to wait for CSLib (which should be at least a couple of years out), right?
English
Yeah, proving Erdős problems with @HarmonicMath Aristotle is great, but have you tried using it to verify LeetCode-style solutions?
We have. It works (invariant inference is... not fast yet).
English