postmortem: 3 weeks shipping a native mac app on a driver + reviewer agent split (what held up, what i dropped)

The developer shipped v3.0 of Buffer, a native Mac clipboard manager, using a two-agent workflow where one agent drove implementation and another reviewed diffs. This reviewer-agent pattern proved effective, catching issues like silent history mutations and missing tests. Settled specs, derived from user input, also streamlined execution. Real-device testing was crucial for assessing gesture feel, which models cannot evaluate. The developer is now asking if others have used a dedicated reviewer agent for over a month and if its effectiveness was sustained.

Unlike typical postmortems, this one details the specific benefits of a two-agent workflow for app development, highlighting how the reviewer agent caught issues a single agent might miss.