diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
new file mode 100644
index 0000000..2fe5c89
--- /dev/null
+++ b/CONTRIBUTING.md
@@ -0,0 +1,105 @@
+# Contributing
+
+This repository holds **reference baselines** and **independent RG optimizer
+experiments**. Contributions should keep those lanes separate and avoid
+unsubstantiated efficacy claims.
+
+This guide is for people sending **pull requests**. How you get a branch onto
+GitHub depends on your access (see below). The **layout and PR expectations**
+apply to external and write-access contributors alike; maintainers may use a
+shorter internal path for trivial fixes.
+
+## Who you are (access paths)
+
+| Role | Typical access | How you open a PR |
+|---|---|---|
+| **External contributor** | No push access to this repo | **Fork** under your account, push a branch to *your* fork, open a PR into `CalculatedContent/rg_optimizers` `main` |
+| **Collaborator / write access** | Can push branches to this repo | Branch on **this** repo from current `main`, push here, open a PR into `main` (no personal fork required) |
+| **Maintainer** | Admin / merge rights | Same as write access for reviewable changes; direct commits to `main` only per the project’s usual maintainer practice (prefer PRs for non-trivial work) |
+
+**Fork is not a social rank.** It is only the usual GitHub path when you cannot
+push to this repository. If you already have write access, do not fork just to
+satisfy a ritual.
+
+### External contributor (no write access)
+
+1. Fork `CalculatedContent/rg_optimizers`.
+2. Branch from current upstream **`main`**.
+3. Push the branch to **your fork**.
+4. Open a PR against `CalculatedContent/rg_optimizers` `main`.
+
+### Collaborator with write access
+
+1. Clone this repository (or add it as `origin`).
+2. Branch from current **`main`**.
+3. Push the branch to **this** repo.
+4. Open a PR against `main`.
+
+Prefer **small, single-theme** PRs (one package, or docs-only) in either path.
+
+## Layout (read before editing)
+
+| Path | Role |
+|---|---|
+| [`baseline/`](baseline/) | **No RG correction.** Matched baseline recipes and notebooks (MLP3, ViT, nanoGPT one-head, nanochat, …). |
+| [`optimizers/`](optimizers/) | **Independent** experiment packages (own notebooks, tests, hypotheses). Not drop-in replacements for baseline without reading each package README. |
+| [`OPTIMIZER_VARIANTS.md`](OPTIMIZER_VARIANTS.md) | Read-only map of actuators, support/geometry, and cadence for every optimizer package. |
+
+Do not mix baseline recipe changes and optimizer-intervention changes in the same PR unless the PR description explicitly justifies the coupling.
+
+## Before you open a PR
+
+From the repo root (with your usual venv/conda env):
+
+```bash
+# example: baseline-oriented syntax/tests (see CI for the full matrix)
+PYTHONPATH=src pytest -q
+```
+
+Also run the package tests that match the files you touched (for example CI
+workflows under `.github/workflows/` such as `baseline-tests.yml` or
+`wwpgd-local-delta-tests.yml`). Name what you ran in the PR body.
+
+## PR expectations
+
+These expectations apply to **any** PR author (external or write-access), unless
+a maintainer explicitly says otherwise for a given change.
+
+| Do | Don't |
+|---|---|
+| State docs-only vs behavior vs logging | Claim an optimizer “wins” without a stated baseline protocol |
+| Keep package defaults unchanged unless the PR is about defaults | Treat α = 2 as a universal point optimum (prefer **boundary** language when α is discussed) |
+| Point at `OPTIMIZER_VARIANTS.md` when adding a new optimizer folder | Unify all optimizers into one package without discussion |
+| Keep baseline test metrics **monitoring-only** (no test-set model selection) | Bundle unrelated refactors with a one-line fix |
+
+Suggested PR body checklist:
+
+```text
+## Type
+- [ ] docs only
+- [ ] logging / provenance fields
+- [ ] behavior (describe default impact)
+
+## Summary
+...
+
+## Test plan
+- [ ] pytest / named CI workflow
+- [ ] no scientific efficacy claim
+
+## Out of scope
+...
+```
+
+## Where to look
+
+| Topic | Path |
+|---|---|
+| Root overview + quick start | [`README.md`](README.md) |
+| Optimizer catalog | [`OPTIMIZER_VARIANTS.md`](OPTIMIZER_VARIANTS.md) |
+| Baseline qualification trail | [`baseline/FINAL_BASELINE_QUALIFICATION.md`](baseline/FINAL_BASELINE_QUALIFICATION.md) and sibling audit docs under `baseline/` |
+
+## License
+
+By contributing, you agree that your contributions are licensed under the same
+terms as this repository.
diff --git a/README.md b/README.md
index ea81c05..c7a5579 100644
--- a/README.md
+++ b/README.md
@@ -5,6 +5,8 @@ spectral renormalization-group program. The repository keeps the unmodified
 reference baselines separate from every RG intervention so optimizer claims can
 be tested against strong, restartable, statistically controlled experiments.
 
+
+See [CONTRIBUTING.md](CONTRIBUTING.md) for fork/PR workflow, baseline vs optimizer separation, and PR hygiene.
 ## Baseline status
 
 The baseline suite has completed a recipe audit, an executable audit, and a