Every week brings new AI breakthroughs, but it's hard to know what's truly been achieved versus hype. Researchers and developers waste time evaluating dubious claims.
A Wikipedia-style platform where experts can submit and vote on milestones, linking to reproducible benchmarks. Categories could include natural language, coding, reasoning etc.
AI companies might sponsor to showcase their achievements. Researchers would value accurate benchmarks.
MVP: Simple list of key milestones with voting buttons and citation links.
Risk: Requires critical mass of expert participation to be authoritative.