there’s a version of building that requires a clear desk, a full night’s sleep, an uninterrupted afternoon, and the perfect conditions to focus. i’ve been waiting for that version for years.
the thing i actually ship is built between things.
a CLI tool while cooking. a prototype at midnight. a rough draft during the 20 minutes before a call. not optimal conditions — but real conditions. and it turns out those constraints don’t just coexist with building: they shape it.
why the margins work
when you have three hours, you scope for three hours. when you have thirty minutes, you scope for thirty minutes. this isn’t compromise — it’s calibration.
the tools i’m proudest of are tiny. a single binary that does one thing. a script that replaces a ten-click workflow. they were built in margins because they had to be. and because they had to be, they’re simple. and because they’re simple, i actually use them.
the full-featured thing you design with unlimited time usually has more modes than you need, more config than you’ll touch, and an interface that made sense in the planning phase and confuses you six weeks later.
the right unit of work
the skill isn’t working fast. it’s knowing what unit of work fits the time you have.
thirty minutes: add one function. write one test. fix one thing that’s been annoying you.
two hours: ship a small feature. write a rough draft. get something to a demo-able state.
the mistake is trying to do the two-hour thing in the thirty-minute slot — and then feeling like you failed when you didn’t finish. you didn’t fail. you picked the wrong unit.
on “real” work
there’s a cultural story that real work happens in focused blocks — the deep work, the flow state, the four-hour sprint. and that’s real. but so is the other kind.
the thing that exists in the world is more valuable than the perfect thing that doesn’t. every tool i’ve shipped in the margins is something i use. most of the things i’ve designed in the clear-desk, full-afternoon mode are still in a notes document somewhere.
margins are not the obstacle. most of the time, they’re just the reality. and the reality is workable.