# CLAUDE.md — WP-AI-0001: The Neural Organ Transplant Experiment

**Causal transfer of cognitive capabilities across local large language models.** Research home: UQO
Research (`Recherche-UQO/ia/wp-ai-0001`). Language: English for all code, docs, figures, paper, logs.
Status: protocol + implementation in progress; **no transfer result has been established**.

The full research mandate (sections 0–18: mission, estimands, RQ1–RQ8, task design, model matrix,
pipeline phases 1–8, metrics, statistics, architecture, gates A–F, figures, paper, registry, licensing,
roadmap, autonomous work rules) was supplied by the author on 2026-11-09 and is reproduced verbatim in
`docs/MANDATE.md`. It binds every session. Operational rules that follow from it:

1. **Never fabricate** runs, metrics, citations, hardware or references. Every number in `docs/`,
   `experiments/summaries/` and `paper/` must trace to a run id in `experiments/registry/runs.jsonl` and a
   manifest in `experiments/manifests/`. Pending measurements are explicit placeholders.
2. **Gates in order** (A repository readiness → B baseline integrity → C mechanistic sanity → D transplant
   validity → E confirmatory evidence → F paper integrity). A failed gate is diagnosed, not reinterpreted.
3. **Exploratory ≠ confirmatory.** `docs/PREREGISTRATION.md` v0.1 is registered; the confirmatory freeze
   (v1.0) must precede generation of confirmatory splits. Design changes go to `docs/DECISION_LOG.md`
   with "before/after freeze"; failures go to `docs/EXPERIMENT_LOG.md`. Unfavourable runs are never deleted.
4. **Persistence vocabulary.** Distinguish single-token intervention, within-sequence effect,
   cross-prompt reuse, and saved checkpoint/adapter. Transient steering is never called learning.
5. **Controls are mandatory** for any transfer claim: unmodified recipient/donor, sham, norm-matched random,
   unrelated-skill donor, shuffled-token/layer grafts, prompting, steering, matched adapter fine-tuning.
6. **Resources.** Laptop (M5 Max 48 GB) and the existing MacLustr SSH access only. Ask before paid compute,
   checkpoints > 20 GB, publishing/submitting, or touching shared infrastructure. No secrets in the repo.
7. **Backends.** Tier 0 = PyTorch toy models (hooks exact, CPU-testable). Tier ≥ 1 = MLX layer wrapping
   with a no-op test and a Transformers equivalence test; never claim a hook MLX does not expose.
8. **Remote.** `origin = gitsrv:srv/git/wpai0001_uqo.git` (spbgit on M1M32). Commit after each sprint
   deliverable; `paper/main.pdf` is committed when it exists.

Quick start: `python3 -m pytest tests -q` · `python3 scripts/audit_environment.py` ·
`python3 scripts/run_pilot.py --quick --device cpu` · see `docs/REPRODUCIBILITY.md` and `docs/TASKS.md`.