# @davidad — 2026-04-24

♥9 ↻0 · https://x.com/davidad/status/2047476313120801253

@InverseMarcus @inductionheads @EvanHub i predict that if we get to look at what it actually said in that eval, it will have often included disclaimers alongside its answers about how it was only weakly confident, but the grader script will not have classified each of these as NOT_ATTEMPTED, because it did make a guess

tags: author:davidad, kind:tweet, thread-context, year:2026