Method4 November 2026
How I run a study
Freeze the plan, save every output, generate every number from what was saved, and report what didn’t work. The routine behind each paper on this site.
Every paper here follows the same six steps. It’s slower than running an experiment and writing up whatever looks good, but it means you can check each claim against what the study actually produced.
- 01
Freeze the plan
The question, the predictions, the measurements and the rule that decides each one are written down and hashed before any result comes in.
- 02
Save every output
Each run writes to its own folder, which is never rerun or overwritten. Checks fail if a saved file changes.
- 03
Generate the numbers
Numbers in a paper aren’t typed in. They’re generated from the saved outputs, and a check fails if text and outputs disagree.
- 04
Report what didn’t work
Studies that stop early and predictions that fail go in the paper, with the reason. Unplanned analyses are labelled post hoc.
- 05
Version, don’t overwrite
A revision is a new version with a changelog. Earlier versions stay as they were.
- 06
Release everything
Code, per-run outputs, the frozen protocols and checksums, archived with a DOI, so anyone can rerun the checks.
A note on tools
I use language-model assistants throughout, under my direction: to help design studies, write and run code, draft text and check references. The steps above are part of why that works. Whoever wrote a line of code, the frozen plan, the saved outputs and the checks decide what a paper can say.