I had a vague idea: automation for small businesses. A week later I had a concept, a market, pricing, and a go-to-market outline. The model accelerated the thinking. It did not replace the calls.
The week#
Days 1–2: narrow the problem. Landed on automated invoice reconciliation for freelancers. Not “automation.” A job someone already hates.
Day 3: competitors, gaps, a rough market size. A ranking, not a promise.
Day 4: survey and interview scripts. Reached 20 freelancers on LinkedIn. Eight replies. That is the document. The model did not send the DMs.
Day 5: features, a wireframe in words, a thin architecture.
Day 6: pricing tiers, unit economics, a pitch outline.
Day 7: launch notes — content, email, possible partners.
One hard decision#
Do not let the model invent the customer. It will fill the hole with a freelancer who does not exist. The week is calls. The outline is what you can write after eight people answered.
The idea is still in progress. I know it is worth pursuing because the replies were specific, not because the chat was long.
What I would not do again#
Spend a month of nights polishing a pitch before a single call. That is a dump of the model I call research.
Ask the model for a market size and treat it as a number I can ship. A ranking. Not a promise.
The bar#
A week that produces a brief someone will answer. The last mile is still the same: eight replies — not a dump of the chat I call validation.