Weak-form proofs mostly eliminated DoS bugs in httpd-rs — a verifiable RL-training target

jd_pressman · x · 2026-11-22

jdpressman discusses httpd-rs: the proofs that (mostly?) eliminated DoS bugs are among the weakest proof forms in the codebase, but even weak proofs can force coherence when applied correctly. Responding to skeptics, he argues models are bad at this only because they haven't been trained on it — it's a verifiable task of exactly the shape trainable via RL, linking the repo's proof-strategies.md.

Original post →

More from coding & agent

coding & agent channel →