Design the failure first
What an AI answer should do when it times out, gets it wrong or runs out of budget, designed before the happy path.
Every AI feature I have reviewed had a beautiful happy path and nothing else.
Read the noteNotes on designing AI products.
Short essays from my work at Incorta and on HiPharmacist: how AI should fail, how people review what it proposes, and how to prototype inside the real product.
3 notes, latest Oct 2026Follow by RSS
More notes
One row per proposed change
Why reviewing AI suggestions fails when the screen mirrors the database, and the one-sentence reframe that fixed it.
Prototype inside the real product
Why I put low-fi screens inside a copy of the live product, at real scale, so stakeholders decide without imagining anything.
You bring the problem.
I bring the prototype.
A few lines about what you want to fix. The reply comes from me, within two days.
Start a conversation