Can agents resolve vague human references like 'the change that broke token refresh'?

yoavgo · x · 2026-09-09

Humans routinely refer to things without naming them — "the change that broke token refresh" or "the issue Amy opened yesterday" — and expect others to follow. Keren Fuchs asks whether agents can do the same: infer what the user means from context and actually locate the right item using their tools.

It's a novel angle on combining reference resolution with tool use as a measure of practical agent capability.

Original post →

More from coding & agent

coding & agent channel →