The availability of AI-assisted engineering makes it easier than ever to produce a convincing prototype. It does not make it easier to identify a durable problem.

We begin with a written thesis: what has changed, whose behavior is constrained, why existing solutions persist, and what must be true for a new company to matter.

Code follows when it can remove a meaningful uncertainty. Sometimes the correct first artifact is a workflow, a manually delivered outcome, a dataset, or a commercial commitment.

Speed is measured by learning velocity, not deployment frequency alone.

Companies are built through choices made before the market is paying attention.