TLDRocket
19 September 2026
AI’s big throughline today wasn’t better models—it was better ways to find out what they can do, and what they shouldn’t. Vals, backed by Andreessen Horowitz, raised $40 million to replace stale benchmarking with evaluations grounded in real industry tasks and measurable impact. It’s expanding from 8 to 25 staff, relocating, and starting work with federal agencies—an implicit admission that leaderboards don’t cover the failure modes executives actually worry about.
Read the full briefing →