Cardano's governance paper is a full spec plus research wish list for proving the thing doesn't deadlock.

Research prototype for EasyCrypt lemma proofs, using an MCP agent to spit out admit-free scripts and replay-check them.

Paper formalizes look-ahead bias as temporal non-interference for backtests and trading, with a decidable fragment, naturally.

An essay arguing multi-agent codegen dies on vague specs, and only verification or invariants save it, as usual.

Formal impossibility result for prompt-injection defenses in shared-embedding models, where token reuse ruins the clean split.

C2PA's provenance story cracks under analysis, with timestamp, revocation, and validator bugs undermining its own goals.

Can you brick a CUDA collective by flipping its masks, lanes, or epochs? This paper says yes, then adds CICs.