Engineering
How I'd build an MVP for a startup
Ship something real quickly, without creating problems for later.
Define the smallest proof
What has to work for the first ten users to get value? The MVP should test one hypothesis
Choose proven foundations
Managed auth, a relational database, and predefined frameworks are a good starting point.
I still write typed, documented code.
Build for change
I produce clear module boundaries, config determined by environment, and a deployment pipeline from day one. This means that the codebase can be updated and extended without breaking the existing functionality.
I also add basic instrumentation early so you know when people sign up, convert, or drop off.
Plan the handoff
Even an MVP needs a plan: who maintains it, how it deploys, and what to do next.
The best MVPs are prototypes in production with real users, data, and documented shortcuts.