A finished small thing teaches more than an unfinished large one
💎CORE ARGUMENT
An unfinished thing can only be judged by its maker, and its maker is the person least able to see it. A finished thing, however small, meets the world, and the world answers with information I could not have produced myself. What teaches is the contact, and only finished things make contact. Size hardly enters into it.
So ten percent of a large project and a hundred percent of a small one are not comparable amounts of progress. The first has taught me what I already believed. The second has been used by someone, and I now know one thing that is true.
Two cases from this year. In October 2024 we chose to pilot the Exceptions Inbox with one customer where three had been proposed. One pilot, finished, showed us a dispatcher's list too long to read, which no plan had predicted and three half-run pilots would have blurred. And on 13 March, two days after the launch and a rollback at 02:40, I had nothing left for a proper issue. I sent 400 words. It said one thing, and it was read, which the long version I did not write would not have been.
The counter-example is mine too. In 2021 I shipped a carrier onboarding wizard, complete and polished, that eleven people used. It was finished and large, and it taught me late what a small version would have taught me early.
Hence the working rule: cut scope until the thing can be finished, then finish it. A small finished thing can grow. A large unfinished one can mostly be defended.
It fails where a thing only works at full size. Half a bridge teaches nothing about traffic. And the single pilot, right as it was, hid a scaling problem that only volume could show.
🧱BUILT FROM
🌍WHY IT MATTERS
When a piece of work stalls I ask what the smallest version is that someone else could use this week, and I ship that. The 400-word issue is now a format I allow myself, not an apology.
☑️NEXT STEPS
- Send the 400-word issue
- Add a smallest-usable-version line to every project note
- Sketch the first finished chapter of the book, not the outline
Test the risky assumptions before building: whether anyone wants it, can use it, and whether it can be built and sustained.
The cheap way to finish something small is a discovery test: 💎 Discovery is cheaper than delivery so do more of it. Knowing when small is finished is 💎 Good enough is a rational stopping rule.
Learning needs feedback from outside the maker's head, and only something released can receive it. A small complete thing closes that loop quickly and cheaply; a large incomplete one keeps it open, so beliefs go untested however much work is done.
Small finished things can add up to nothing. A sabbatical book is not thirty issues laid end to end, and a career of pilots is a way of never betting on anything.
The belief this note argues for, the short issue itself, and the pilot decision reviewed in February.