Developer releases Nyxis, a Lean-based backend framework claiming mathematically proven authorization safety

A developer has released Nyxis, a backend framework written largely in Lean that uses formal verification to prove an authorization safety theorem: no reading, modification, deletion, or permission-granting beyond an authenticated subject's legitimate rights occurs for any sequence of HTTP requests. Developers only manage permission rules, letting AI agents write the rest. The author notes limitations: code outside the proof and the runtime must be trusted, and miswritten policies void protection. The author contrasts Nyxis with Microsoft/GitHub's PDD Kit and an Anthropic–Theorem collaboration, and says it was released early due to recent security incidents.

✨ AI summary of the linked source — not editorial content.