DeepMind's AlphaProof Code Formally Verifies Key Signal Protocol Guarantees in Four Weeks

Skynet Chance (−0.05 pp): Practical AI-driven formal verification could eventually let developers mathematically guarantee software properties, a tool relevant to securing systems and verifying AI behavior. The explicit human-in-the-loop design also counters fully autonomous self-checking, modestly lowering control risk.

Skynet Date (no change): Better verification tools add some defensive capacity against AI-driven security threats, slightly slowing the pace at which risks could outrun safeguards. The effect is limited because the work is narrow in scope and its ability to scale is unproven.

AGI Progress (+0.02 pp): Agents that produce rigorous mathematical proofs about real-world code show meaningful progress in long-horizon formal reasoning and self-correction. However, the result is domain-specific, limited in scope, and still depends on human guidance.

AGI Date (no change): Completing a historically labor-intensive verification task in four weeks shows agentic reasoning tools advancing quickly. That speedup could also accelerate trustworthy software development, though the cost and scalability of the approach remain unclear.