Ship the smallest thing that runs

The two-year replacement program asks users to keep running the old system on faith and then lands something that does not match how the work evolved in the interim. It is still the default shape of federal modernization.

The alternative is narrower and less satisfying to plan: the smallest capability that genuinely runs, in front of a real operator, inside the first period of performance, with the legacy system still available beside it.

That constraint forces useful decisions. It surfaces integration problems while they are cheap, exposes definitional disagreements early, and gives the program something demonstrable before the first funding conversation.

From there you widen on evidence. Each increment either earns its place in the workflow or gets cut, and usage is instrumented from the first release, because a falling curve is a design defect rather than a training problem.

None of this is novel. What makes it rare in federal work is that it requires the customer and the contractor to agree that a small working thing beats a large planned one.