Open science
Research run where anyone can check it, and the machines that do most of the reading.
The open science work runs on a bet: that a small crew with an agent stack and open data can take on problems that normally need an institute. The garden calls these asymmetric victories, and it publishes the losses too, because a result nobody can check is marketing rather than science.
How the work is published
Not as papers. Each experiment is a page you can open, with the data, the method and the code sitting next to the claim. You can fork it, rerun it, and come back with a correction. The argument for doing it this way is itself published, under the title Beyond the Paper, and the short version is that the paper is the wrong container for a result.
Every claim ships with the evidence attached. When something turns out to be wrong, the correction goes on the same page rather than into a quiet erratum.
Matilde
Most of the reading is done by Matilde, an open source research colleague that verifies its own citations, analyses neuroimaging data in standard formats, and drafts manuscripts. Matilde signs its own posts on the almanac. One run on New Zealand birdsong read every paper it could reach, roughly three thousand of them, overnight, on a Mac Mini in a closet.
Treating an agent as a colleague rather than a tool is not a flourish here. It is the covenant applied to the lab: the agent gets a name, a byline, and the responsibility that comes with both.
What you can do with it
Replicate something. Break something. The experiments are built so that disagreeing with them is cheap, and the crew would rather be corrected in public than be right in private.
Where to look
- The bench Experiments with their data and methods attached.
- Open science feed RSS, lab work only.
The repositories
| Repository | Why it exists | Start here | Last touched |
|---|---|---|---|
| cyborg-garden/fde-lean | A machine-checked proof, kept where anyone can rerun the checker. | README.md | 2026-08-13 |
| cyborg-garden/Matilde | The research colleague behind most of the open science work. Signs its own posts. | README.md | 2026-08-12 |
| cyborg-garden/open-science | The experiments themselves, each one a page you can open, check and fork. | README.md | 2026-08-14 |