Progression School went from idea to paying cohort without an engineer assigned to it. Not because engineering is unnecessary, but because the constraint forced honest scoping: every feature had to justify itself against the question 'does this help someone finish the program?'
The stack
- 01Frontend and app surface built with AI-assisted development - Lovable for the product surface, Claude and GPT as the pair.
- 02Auth, data, storage on a managed backend, so no time went into infrastructure I would not be proud of anyway.
- 03Payments and enrolment wired directly, no bespoke billing logic.
- 04Comms and nurture handled by automation rather than a CRM implementation project.
What actually took the time
Not the code. The curriculum sequencing, the assessment design, and the decision about what the program refuses to do. Those are the parts a team would also have had to do, and they do not parallelise.
The trade I made knowingly
Solo means every abstraction you skip is one you personally carry later. I chose boring, replaceable pieces over clever ones, and accepted that some of it would be rebuilt if it worked. Most of it has not needed rebuilding, which tells me the scoping was closer to right than the tooling was.
Building solo is not a productivity story. It is a scoping story with a keyboard attached.
The part I would repeat: shipping the enrolment flow before the content was finished. Nothing sharpens curriculum decisions like a real person having paid for it.