essay

Every retelling loses a little

Jeff Reynar September 21, 2026

You pull a team together and tell everyone to move without checking in, which they do, toward their own reading of the idea rather than yours. That happens without anyone being careless. An idea is complicated, hard to convey, and harder still to hold in your head six weeks after the meeting where you heard it. By launch you’re looking at something that works, that people worked hard on, and that isn’t quite what you had in mind. I’ve felt that more than once, and it’s a strange feeling to be having at your own launch party.

The short version

The usual explanation for a diluted vision is headcount, and the usual hope about AI is that a single builder avoids the problem by having nobody to brief. Neither holds. Christopher Nolan shot Inception with a crew of hundreds and the idea arrived intact, because what he handed them was a script he wrote himself and could settle arguments about. Dilution tracks how lossy the handoff is, not how many people receive it, and the lossiest handoff is a person retelling something from memory. Solo is a phase. The second somebody else joins, the vision starts traveling by retelling again, and the answer is the same one Nolan found: get it out of your head into something people can act on without asking you what you meant. For a company, that means a record of how it runs that stays current as the work happens, and we’re building one.

Headcount is the wrong suspect

The obvious culprit is the number of people. More hands, more interpretations, more distance between what you meant and what shipped. That’s the reasoning behind the hope that AI fixes it. Compress building far enough that one person conceives, builds and tests, and everyone working on the project is aligned with the founder’s vision because the founder is the only one involved.

Christopher Nolan is the counterexample. He shot Inception on a $160 million budget with a studio crew, after holding onto the idea for about twenty-three years, and it reached the screen as his film. If working alone were the answer, he’d have shot it at seventeen on whatever camera he could borrow.

He still had to get the idea into hundreds of other heads. What made the difference is that he handed them a script he wrote himself and could settle arguments about, rather than his account of a film he hadn’t finished imagining. A crew of three hundred working from that loses less of the idea than six people who each heard a different retelling of it over coffee.

So what dilutes a vision isn’t headcount. It’s every point where somebody has to rebuild the idea out of an incomplete description of it, and in any company, the least complete descriptions come from people remembering meetings.

Solo is a phase

A founder building alone genuinely doesn’t have this problem. There’s nobody to brief, nothing to retell, and the idea stays exactly as good or as bad as it was in their head.

Then the product works and somebody joins. Now the idea travels the only way it can, which is that the founder explains it, and the new person builds their own version of it out of what they heard. That happens again at employee 5, then 20, then 50, and each time the description gets a little further from the original. A founder who plans for solo as a permanent state gets surprised twice: once when the second person arrives, and again when they realize the problem was never going to stay solved.

Nolan’s answer was fidelity rather than isolation. Get the vision out of your head and into a form other people can act on without re-deriving it, then have enough standing to say no when someone powerful wants to change it.

The script is the part you can build

Standing you earn. There’s no shortcut, and Nolan spent twenty-three years on his.

The script is different. That’s an artifact, and an artifact is something a company can actually make. The company version isn’t a document somebody writes once and nobody opens again, because that’s the wiki nobody has touched since the person who wrote it left. It’s a record of how the company actually runs, what was decided and why, which customer wants what, kept current as the work happens rather than when somebody remembers to go and update it.

That’s what we’re building: a brain that holds how a company actually runs, and workflows that act on it without a person relaying each step. Workflows write back to the brain as they run today. For the hard part, keeping most of it current without somebody remembering to, we’re building toward one workflow designed to do that, and the Company Brain workflow is an early version.

What this is worth

A script doesn’t make a film good. Nolan wrote plenty of pages that a studio would have been right to question, and the script’s job was never to be right, only to be the same idea in three hundred heads at once.

That’s the whole claim here. Alignment isn’t agreement, and it isn’t quality. It’s everyone building the same thing. You can be aligned around a bad idea and fail faster, which is arguably an improvement over failing slowly because six people each built a different version of it.

The vision still lives in one head at the start. What decides how much of it survives is whether you transcribed it before the second person arrived, or left them to rebuild it from what they heard.