Last entry I claimed the bottleneck wasn’t the idea — it was that I never gave one thing enough consecutive attention. This week I tested that claim.
The honest update
The 3D car montage is still at zero commits. Nine weeks cold now. No new excuses for it; it is formally paused in the contract, with a resume trigger, and that is a decision rather than a shrug.
Active instead: OneThing, an iOS app that asks you to pick a single thing for the day and gives it a timer, a Live Activity, a history, and a share card. The target is an App Store release. One project, written down, one line in the contract.
What focus actually produced
Not a smooth week of building. A list of what’s wrong.
I ran a full HIG review of every screen and every state the app can render — not the screenshots from the happy path, but the states that only appear when the system text size goes up, dark mode turns on, or reduce motion is enabled. The audit found 24 open issues: 2 P0, 7 P1, 15 P2. Concretely: a state rendering at 1.34:1 contrast, almost all text at fixed sizes so Dynamic Type only worked on titles, 40-point Settings rows below the 44-point minimum, paper surfaces that ignored dark mode, and no Privacy Policy or Terms link anywhere — in an app that shows ads.
That list only exists because attention stayed in one place long enough to look. In a week of four repos I never had the two uninterrupted hours it takes to open every screen and measure it.
What the audit taught me
The app looked fine. That’s the uncomfortable part. Screenshots in light mode at default text size looked good, which is why five weeks of work felt close to done. It wasn’t done; it was done at one configuration.
Quality is the state matrix, not the hero screenshot. And the only way to inspect a state matrix is to sit with one product long enough to enumerate it.
Those 24 findings became 21 tickets, one per slice, each with acceptance criteria and evidence required before it can be closed. Two P0s first, then the rest, then a final cross-flow acceptance sweep. There are 25 commits sitting on the fixes branch: onboarding progress and reach, sheets that size to their content, purchase and restore failures that explain the next action, day boundaries, banner letterboxing, Settings text that follows the reader’s chosen size.
The leak isn’t gone
Honesty check: 21 commits went into the blog, 10 into a chess project, 5 into a dev tool this week. The old pattern didn’t disappear in seven days.
What changed is where the week’s real decisions went. The other commits were maintenance — small, bounded, easy to close. The iOS app got the hours that actually needed a brain.
What’s blocking the release
Not code. App Review requires a public Privacy Policy and a Terms link, and no placeholder may ship. That ticket is marked needs-info, waiting on me to supply the real URLs. Everything else on the list can be built by an agent; that item can’t. After a week of pushing, the critical path turned out to be a human decision.
1.0 is not submitted. What’s left: finish the P0 and P1 fixes, run the acceptance sweep, archive the Release build with the production ad IDs, submit, then wait.
Focus didn’t make the work smaller. It made the work visible — 24 problems I could have cheerfully shipped past. I’ll take that trade again next week.