01 / REPORTExploratory
Stem-Cell Micro-Harnesses
Report 1. Thirteen colonies, eight coding tasks, five model conditions. Method, variables, results to date, 18 registered hypotheses and the roadmap.
Lab / research
Small harnesses,
under selection.
Razi cells is a research program at Moissanite Labs. A cell is a very small agent harness that starts from a need and an empty folder, writes its own tools, and is kept or dropped by a measured score. The program asks which laws, felt from inside by each cell, make a colony improve on its own. Everything here is exploratory. Guesses are written down before the data, and misses are reported.
Read
02 / MAP52 directions
Open directions
What is left to explore, as a mind map: laws of the cell, which model, what is inherited, signals beyond text, proof, growing a model, the colony as an organism.
03 / RECORDUnder review
The laws record, E11 to L1
The registered tests that followed Report 1: an energy economy for cells, toy growing networks, a stem cell, a need sensor, and heredity. Test log, idea log, code and results.
04 / PAPERSPublic
Weft papers
The program is the agent-level twin of the Weft cell loop, which applies the same split, score and keep cycle to model weights.
Guesses written before the data, and how many held
Each experiment in the laws record states its guesses before any run and is judged as written, by script. A guess that held is not an established result: runs are few and only one model family has run the energy laws so far.
| Test | Question | Held |
|---|---|---|
| GR1 | Toy network: does copy division pass a gate | 3 of 5 |
| GR2 | Toy network: daughters born different | 5 of 6 |
| GR3 | Toy network: one knob at a time | 3 of 5 |
| GR4 | Toy network: need sensor and escalation | 5 of 5 |
| SP1 | Stem cell that feeds its parts | 7 of 8 |
| SP2 | Colony shape: flat against stem | 6 of 8 |
| SP3 | Price list and need sensor | 6 of 12 |
| L1 | Heredity across generations | 5 of 8 |