Introducing AlphaProof Code: a breakthrough in verifying complex software at depth

Selection record

AdmittedSum of both 164 ≥ twice the threshold 120

Source tier

Official, first-hand; this tier's threshold is 60

Pre-filter

passed：AI智能体框架用于形式化验证代码

Why it was chosen

DeepMind used a multi-agent framework to formally verify the actual code of the Signal protocol, completing in four weeks work that would previously have taken years, and open-sourced the verification artefacts.

A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.