Study human-plus-agent outcomes together, not models in isolation

_arohan_ · x · 2026-08-04

The poster argues that human + agent outcomes should be studied jointly as a distribution, not separately.

The quoted point suggests we may soon think models have plateaued simply because humans can no longer perceive the difference. In that framing, the real question becomes how the combination of a person and an AI agent behaves, rather than judging the model in isolation.

Original post →

More from AGI Musings

AGI Musings channel →