Brain Build
How a Brain Build is made, from the real work to a Glitch you can test.
AI is an average of a huge mix of information. That is fine when the average is right. It is a problem when your edge is the part that is not average. A Brain Build captures how you actually work, the claim that explains it, the evidence on both sides, and what would change your mind.
It is a living record. It is not a one-time PDF, and it does not retrain a model. You can download a Brain File later and use it as reference context. Empty fields stay empty. The product will not fill them with guesses.
Mind Node shows a cleaned version of the fields you saved. It does not show the raw typing. Capture still shows exactly what was typed. The cleanup tightens wording. It does not add facts, metrics, or a new Glitch. The Glitch stays the claim a person wrote, cleaned only enough to read clearly.
Saving Capture runs that cleanup when the AI gateway is configured. If the gateway is not configured, the save keeps your words and Mind Node does not invent a cleaned version. The screen says the gateway is missing.
A private optimization read stays on the team side. You do not see it, and it is not in your downloads. Recommendations show up in Fit only after Scott publishes them.
The truth of the work
Start with what happens, not the org chart.
- Staff, or a first pass if you are solo, describe a recent example.
- A manager, or a second pass, names what keeps repeating.
- Leadership, or a third pass, reads those notes before explaining what they might mean.
One person can do all three passes. Do them separately so the explanation does not overwrite the example.
The Glitch
The Glitch is a counterintuitive thesis. It is the surprising explanation for what is really going on, the one most people would miss. Keep it short enough to say without special vocabulary.
A useful Glitch has to be able to be wrong.
Evidence and the falsifier
Keep three things next to the claim:
- Evidence that supports it, with a specific example and a source.
- Evidence that pushes against it, including exceptions and missing information.
- The falsifier: what you would have to see before you dropped the claim. That is the fair test. Write it before you call the idea proven.
Ranked moves
Fit turns the evidence into a short list of 3–5 practical recommendations. Each move says what to do, why it matters, and how you will know it helped. It also has a rank, a reason, a cost, an effort, and a status such as Recommended, Review, Selected, or Not now. One of those moves is what an Install would pick up. "Do not buy this" is an allowed answer.
A living dataset
You keep editing. Meetings can add excerpts later. The measure, the starting point, and the target sit on the same record so a later result can be checked against the claim. Department Brain Builds can stay separate when the work is different, and still roll up into one company view.
A recording with no transcript text is not rewritten. Transcription is not part of this workspace, and the upload limit stays 4 MB. When a transcript is marked reviewed, the filled Capture fields are cleaned again. The cleanup does not turn the recording into new claims, and it does not search the web.
The five capture steps inside the workspace are: the real work, the Glitch, the evidence, the decision, and the measure. Screen-by-screen clicks are in the workspace tour.
Where to edit it
Open Brain Build in the client workspace. Use Capture to write. Use Mind Node to read the picture. Use Meetings and sources to attach a transcript. Use Fit and conclusions for the ranked moves. Use Brain File to download the markdown.
View-only samples can be read. They cannot be edited.