Process
How We Ship a Production MVP in 5 Weeks (The Actual Playbook)
When we tell founders their MVP will be in production in five weeks, the reaction is usually polite disbelief. That’s fair; the industry average for “simple app” is two quarters and a follow-up invoice. So here’s the actual playbook, week by week.
Week 0: The Discovery Sprint
Before any code, one intense week of scoping. Three artifacts come out of it:
- The user flow map. Every screen in the core loop, sketched and agreed
- The data model. Entities, relationships, and permissions on one page
- The cut line. A written list of what we are explicitly not building
That third one does the heavy lifting. Every blown timeline we’ve ever inherited traces back to a missing cut line. Scope doesn’t creep when the scope is a signed document.
Week 1: Skeleton with organs
We don’t build screens first. We build the spine: auth, database schema with row-level security, payments plumbing, deploy pipeline. By Friday there’s a deployed application with real signup and a working Stripe checkout, even if it only does one thing.
This inverts the usual demo-driven order, and it’s why the last week isn’t a death march. The risky integrations are done first, while there’s still time to react.
Weeks 2–3: The core loop, demo every Friday
With the spine in place, features go fast. This is where AI-assisted development pays compound interest, because the patterns are established and the tooling can follow them. We ship the one workflow that makes the product worth paying for, then widen.
Every Friday: a demo on the production URL. Not a staging link, not a video. Clients click things themselves. Feedback lands while it’s still cheap to act on.
Week 4: Hardening and handoff
The unglamorous week that separates products from prototypes: error states, empty states, mobile passes, monitoring, backups, load sanity-checks. Then handoff: docs, credentials, a recorded architecture walkthrough, and source code ownership transferred completely.
Why this works
The honest answer isn’t the AI tooling, though it helps enormously. It’s the constraint stack: a fixed timeline forces a real cut line, the spine-first order forces the risk to the front, and Friday demos force the truth out weekly.
Speed isn’t the goal. It’s the evidence that the decisions were made.