diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
new file mode 100644
index 0000000..495878c
--- /dev/null
+++ b/CONTRIBUTING.md
@@ -0,0 +1,9 @@
+# Contributing
+
+1. Pick an issue and comment that you're working on it.
+2. Branch from `main`: `git checkout -b feat/short-name`.
+3. Set up as in the README, then make your change.
+4. Before pushing: `ruff check . --fix` and `pytest` must pass.
+5. Open a PR with `Closes #N` in the description. CI must be green.
+
+Keep PRs small and focused. Changes to the model should include evaluation numbers (`python src/lysis/models/evaluate.py`).
\ No newline at end of file
diff --git a/README.md b/README.md
index 5f93a04..c1e7093 100644
--- a/README.md
+++ b/README.md
@@ -1,7 +1,72 @@
 # Lysis
 
-Open-source tool that helps new contributors find beginner-friendly GitHub issues and repos, ranked with machine learning.
+Beginner-friendly GitHub issue ranker. Give it a repo and it ranks the open issues by how likely a first-time contributor is to get a PR merged for them.
 
-Status: early development.
+Labels like `good first issue` are often stale or wrong, so Lysis learns from what actually happened: issues that first-time contributors really closed.
 
-License: MIT
\ No newline at end of file
+## How it works
+
+1. **Ingest:** fetch repos, issues and merged PRs from the GitHub API (ETag-cached).
+2. **Label:** an issue is positive if a first-time contributor's PR closing it was merged.
+3. **Model:** gradient boosting on issue and repo features, evaluated with repo-grouped cross-validation.
+4. **Rank:** score open, unassigned issues, then decay by age so stale issues sink.
+
+## Results
+
+5-fold repo-grouped CV on closed issues (base rate 0.10):
+
+| Ranker | AP | AUC |
+|---|---|---|
+| random | ~0.10 | 0.50 |
+| labels-only baseline | 0.245 | 0.678 |
+| gradient boosting | 0.278 | 0.753 |
+
+This is an early prototype. The model beats the baseline, but the signal is modest.
+
+## Quick start
+
+```bash
+git clone https://github.com/OpenRootdev/Lysis.git
+cd Lysis
+python -m venv .venv
+.venv\Scripts\activate        # Windows (macOS/Linux: source .venv/bin/activate)
+pip install -r requirements.txt
+cp .env.example .env          # add a read-only GITHUB_TOKEN
+```
+
+Build the dataset (once, takes a while, cached):
+
+```bash
+python src/lysis/ingestion/fetch.py
+python src/lysis/features/labels.py
+python src/lysis/models/evaluate.py
+```
+
+Rank issues:
+
+```bash
+python src/lysis/cli.py psf/requests
+streamlit run src/lysis/app.py
+```
+
+## Project layout
+
+```
+src/lysis/
+  ingestion/   GitHub client and repo fetching
+  features/    labels and feature building
+  models/      baseline, gradient boosting, evaluation
+  cli.py       command line ranking
+  app.py       Streamlit UI
+tests/
+```
+
+## Known limitations
+
+- Repo stats come from fetch time, not issue time (a small leak; removing them costs about 0.025 AP).
+- Only Python repos so far.
+- Labels depend on PRs linking issues with `closes #N`.
+
+## Contributing
+
+See [CONTRIBUTING.md](CONTRIBUTING.md). Look for issues labeled `good first issue`.
\ No newline at end of file