Updated as I go
How I build products with AI, written up as I go: what works, what doesn't, and what I'd do next time.
Making things became cheap. Deciding what not to make became the job.
My first plan for Whereto had four categories in a dark iOS app. Reality subtracted: within a day it was eat and drink only, on the web, because a link can be texted in a second and a beta app cannot.
From: How I built Whereto using AI →AI will generate a design from one line. It will look fine and feel generic, and it cannot tell you which design is right.
That is taste, and you build it by looking at a lot of good work. Collect references first, then give the model three to five with a sentence on what works in each.
From: How I built Whereto using AI →Claude does not remember yesterday. Keep a README and an architecture doc in the project, including the known weaknesses, which is the section a newcomer needs most.
From: How I built Whereto using AI →One focused session on your product, your AI strategy, or the idea you can't stop thinking about. Bring the problem; leave with clear next steps.