Research and Learning Lab
DraftThe subject sections hold reference material written in an assertive register. This section holds the work behind it, written in a different register: what is being learned, what is uncertain, what was computed, and what did not work.
Keeping them separate is the point. Confident prose about a half-understood topic is the failure mode this structure exists to prevent.
The workflow
Nothing reaches a subject section without passing the test in the middle. A page is promoted when the topic can be explained end to end, when the numbers have been computed rather than recalled, and when the content does not depend on how it happened to be learned.
Conventions across the lab
| Convention | Reason |
|---|---|
| Date everything | Understanding changes; the chronology is how conflicting notes are resolved |
| Cite every source | A note without a source cannot be checked and is close to worthless |
| Record what was not understood | The most useful part of a note, and a pointer to what to study next |
| Write the prediction before the result | Converts a demonstration into a test |
| Keep failures | A record containing only successes has a survivor bias that defeats its purpose |
| Never edit old entries to match current understanding | Write a new one and link back; the change is the interesting part |
The last two are the ones most often violated and the most costly. An experiment that failed, an idea that was refuted, and a note that turned out to be wrong are the entries that make the collection useful for judging one’s own reliability.
Pages
Where a topic starts. One note per session, with its source, what was understood, and specifically what was not.
02Questions I am exploringOpen questions stated precisely enough to be answerable, with what is known and what would settle them.
03Paper reviewsClaim, method, assumptions, result and an honest assessment of whether the argument supports the conclusion.
04ExperimentsHands-on work with the hypothesis written before the result, the setup recorded well enough to repeat, and failures kept.
05SimulationsRuns recorded with their model, assumptions, validation against a known case, sensitivity, and the archived state to reproduce them.
06CalculationsDerivations shown in full with units carried through and an independent sanity check, because a wrong number rarely announces itself.
07CodePrograms indexed with what they do, how they were validated and where the repository is. No listings, for reasons the page gives.
08ModelsGoverning equations, assumptions and validity range, the difference between calibration and validation, and known failures.
09Architecture diagramsMermaid conventions used across the site, and the test for whether a diagram earns its place.
10Ideas and hypothesesSpeculation labelled as such, with what would falsify each entry, and the refuted ones kept.