
Co-founder @zksecurityXYZ Math & crypto. Agentic coding addict. TypeScript magician. Lean4 enthusiast.
Wien
Joined February 2021
- Tweets865
- Following1.5K
- Followers1.5K
- Likes13.7K
Gregor Mitscha-Baude retweeted
I think at this point I disagree that AI is not showing true signs of intelligence, or that AI is incapable of producing “new” things. I think similar to how we learned that we’re not at the center of the universe, or that we’re not at the center of the solar system, or that we have common ancestors with monkeys, we’re learning that there’s nothing special with our “intelligence”, and many people are in complete denial. This will be followed by acceptance.
Gregor Mitscha-Baude retweeted
Lean v4.35.0 is out, and it ships with verified kernels.
Run `lake check --paranoid` to recheck your project with every bundled kernel.
Blog post: leodemoura.github.io/blog/20…
Gregor Mitscha-Baude retweeted
It is difficult to explain just how impressive this is. The Tachyon Project (@TachyonZcash) and @zksecurityXYZ have raised the bar for high-assurance software.
zakura.com/engineering/adver…
Gregor Mitscha-Baude retweeted
We hired @zksecurityXYZ to audit Ironwood's formal verification. Rather than looking for security bugs, they tried to *add* bugs to see if the formal verification could catch them!
Here's our summary, along with the report:
zakura.com/engineering/adver…
Gregor Mitscha-Baude retweeted
We are very happy to announce that we have formally verified the techniques of Liam Eagen's eprint.iacr.org/2022/596 in Lean 4, using only the standard axioms:
magicgrants.org/2026/09/29/D…
github.com/zksecurity/eagen-…
🧵1/6
Gregor Mitscha-Baude retweeted
Problems with advanced cryptography in production?
Need help fixing it? We are happy to announce that zkSecurity now offers incident response:
zksecurity.xyz/incident-resp…
Gregor Mitscha-Baude retweeted
I don't think you're prepared to how fast Bend will evolve and how much better it will become over time. If you're judging it for what it is now (which is already kinda nice tbh) you're missing how I structured the whole thing to surf the exponential
Gregor Mitscha-Baude retweeted
It is amusing to me that the only high profile examples of "misaligned" agents come from the big labs themselves. None of the uncensored models seem to actually be doing any harm... to the contrary, they have been essential in creating secure software in the past 6 months.
Gregor Mitscha-Baude retweeted
Did you know that Clean has a website and a pragmatic introduction to formal verification for zk devs?
clean.zksecurity.xyz
Thought I'd have some fun with bend by @VictorTaelin
Prompted Astra to port @pidotdev to bend 1:1, including all the missing primitives like `fetch`
github.com/mitschabaude-bot/…
This morning it started faithfully reimplementing JS objects with all their cursed semantics 😆
Gregor Mitscha-Baude retweeted
The "whiteboard defense:" I should be able to pull you aside at any moment and ask you to explain any customer-facing system you've shipped. You should be able to clearly explain how it works and defend the decisions you made. This is my benchmark for responsible AI usage.
I don't expect line-level familiarity with the code. I don't care if you remember the exact function name or implementation detail. You may not even know it. I don't care.
But if I ask "why did you do X instead of Y?", "what happens if this actor behaves maliciously?", "what data structure did you use here and why?", or "where does this fail?" you should be able to answer confidently.
For PoCs, demos, experiments, whatever: I don't care. Generate 100% of it and understand none of it. Speed over quality every time in those specific scenarios.
But if you're shipping customer-facing work, you can't be shipping things you don't understand at a high level.
I wonder if the default for subagents should just be to be forked from the coordinator session.
Current agent generation loves to keep subagents dumb and mechanical. IMO that's a major slop contributor. Subagents are not given context, ownership, anything that makes them care
Built a little tutorial for how to get verified correct and FAST executable code out of Lean 👇
Thanks for the prompt @alinush!
Replying to @mitschabaude
Regarding doing stuff in "just Lean," (which would be great IMO: less languages => less tooling/complexity => less headache), I could not really find any meaningful tutorials.
e.g., I'm thinking "here's a sorting spec in Lean, a merge sort implemented in Lean that you can compile and run and a proof that it sorts according to the spec."
We may be witnessing Yukon (@YukonResearch) history: perhaps the first paper-worthy contribution discovered live on the platform 😬😬😬
After ecdsa.fail, a challenge to improve quantum algorithms for attacking ECDSA, barely moved for a month, @franklyteddy using @ChatGPT-5.6 Sol with Codex, rebuilt its core modular division around a reversible “ping-pong” algorithm.
The result: 25.8% fewer Toffolis in a single leap.
Adversarial review from Claude and GPT suggest the implementation materially advances prior published work and likely itself be publishable if verified by other experts.
I expect this to be the first of many. Congrats to @franklyteddy once again!














