Main idea

Every piece of work has an expected output, the one it was commissioned for. It produces a second one, uncommissioned, which is almost never collected because nobody asked for it.

A user interview produces a write-up, and also assumptions, reusable examples, sometimes a decision. An RFP response produces a client document, and also a memory of recurring questions, validated answers, evidence and disclosure limits. An article produces a published text, and also stabilized ideas that will later feed a training course or a discussion. A specification produces the means to build, and also the trace of trade-offs that will explain, six months later, why the product works the way it does.

The second output exists in every case. What varies is whether it is picked up.

Layer added by "Persist that" (2026-09-12). A working conversation with an assistant falls under the same observation, with an even less favorable ratio: most of what it produces — attempts, rewordings, abandoned leads, answers replaced ten minutes later — is disposable, and the second output there comes down to a few paragraphs. Collection is therefore less a pickup than a designation, made at the moment the paragraph appears.

Why it matters

This changes the nature of the effort required. It isn't about adding capitalization work on top of the real work, but about not throwing away what the real work has already produced. The expense is one of collection, not of production.

It also gives a place to look: the second output is at its maximum at the moment the first piece of work ends, and it evaporates afterwards.

Nuances and limits

Collecting isn't free, and not everything deserves to be kept. A second output picked up without sorting produces an archive, not capital.

And some second outputs only have value in a precise setting: a very local trade-off will serve nowhere else.

Open questions

• Who collects, when the expected output is produced by one person and the second output is useful to another?