My favourite part isn't just that Claude found the seeds. It's how: reasoning about what a C programmer in the '90s might have done, recognising an old PRNG, and tracing its output back to Unix timestamps. Cryptographic archaeology at its finest 😂

One thing I love about cryptography is how ideas from decades ago can suddenly become relevant again. We've been looking at batch decryption for encrypted mempools, and it turns out that an idea from almost 40 years ago (Batch RSA) gives us a surprisingly elegant construction. Sometimes the old tricks are the best ones :) ia.cr/2026/2252 @muurshb @nicodoettling

Very excited to be collaborating with the QRL foundation on the future of post-quantum ledgers!

Why are we teaching AI agents to operate computers through shell commands? An agent should express what it wants, not how to do it. The OS should decide how to execute that intent, enforce permissions, and verify the outcome. That's the direction we're taking with NomOS. nomoslinux.org @nicodoettling

Omarchy's vision of vibe-coding your machine with AI agents is exciting. But who says "no" when the agent gets it wrong? Better models aren't a security boundary. Agents should express intent. The OS should enforce policy and verify what actually happened.

One uncomfortable thought about AI-driven operating systems: We're giving probabilistic agents permission to make real changes to our computers. Better models help. But prompts should not be the security boundary! The agent expresses intent; the OS enforces policy and verifies the result.

nomoslinux.org ☝️