Weekly Review: One Ship, One Application

Week 2 of my four-week build sprint had two main goals: apply to one job every day, and ship my agent project. The honest score: one application, one shipped project. The application count looks like a failure until you see the dependency — my recruiter session and I had agreed that for agent-engineer roles, a resume without a shipped agent project isn’t credible. So the ship gated the applications, deliberately. As of this week, the gate is open.

learn-to-ship went from in-progress to public-with-a-live-demo in one week, 71 commits. The mechanism that kept it moving while everything else waited: QUESTIONS.md-driven development — every open question becomes one small PR. The work was always small enough to finish, so it never stalled. That’s the practice I’m carrying to the next project.

The quieter win was environmental. A mid-week experiment — go out to a café every morning — turned out to fix two things at once: it skips the noon nap, and skipping the nap fixed my night sleep. I only noticed the causal chain while compiling this review, which is the point of reviewing at all.

Compiling the review surfaced something else: I could recall the things I shipped, but not the infrastructure that improved underneath me — my daily-review prompt versioned five times this week, my job tracker grew a dashboard, and none of it surfaced from memory until the commit log showed me. Review is output-driven; I’m practicing running my life review-first.

Next week: Kubernetes, one gap skill at a time, and one application every day — this time protected by a morning orient where the application is task #1, before any build work. Spoken English moves to lifestyle slots (work breaks, shows over lunch): the disagreement with my recruiter about it dissolved once we saw we were arguing about timing, not importance.