Ex-OpenAI researcher: eval teams lose losing battles to glory-chasing capability teams
suchenzang · x · 2026-09-13
Former OpenAI researcher suchenzang argues that in org structures where individual glory is prized above all else, "eval teams" fight losing battles against hillclimbing "capability teams" (PR, promo packets, all the wins), and the only delivered outcome is "slathering lipstick on pigs".
In the quoted tweet she adds that auditing these systems requires nontrivial compute support, and that every probe leaks information — logged and used for future training despite denials — making the playing field unfair before you even start.
More from Companies & People
- Founder essay: building something extraordinary requires taking your bubble far too seriously — DominiqueCAPaul · 2026-09-13
- Swiss AI Safety Days 2026 doubles to two days, Stuart Russell to keynote at ETH Zurich — maksym_andr · 2026-09-13
- Dario's "third-party evaluators with employee-level access" idea sparks debate — ayushtweetshere · 2026-09-13
- 795 roles open at OpenAI — AGI has not been reached — Impossible-Humor-182 · 2026-09-13
- OpenAI hits automated research intern goal, agents solve Navier-Stokes in 88 hours — btibor91 · 2026-09-13
- Microsoft Is Beating Itself: How Office Branding Chaos Exposes Copilot Woes — DavidLinthicum · 2026-09-13