diff --git a/CHEATSHEET.md b/CHEATSHEET.md
new file mode 100644
index 0000000..86f8b5e
--- /dev/null
+++ b/CHEATSHEET.md
@@ -0,0 +1,208 @@
+# BincioActivity — Cheatsheet
+
+## Daily workflow
+
+```bash
+# 1. Drop new .fit / .gpx / .tcx files into your input dir, then:
+bincio extract
+
+# 2. Rebuild the site
+cd site && npm run build
+
+# 3. Done — copy site/dist/ to your host
+```
+
+---
+
+## Extract
+
+```bash
+bincio extract                            # full run using extract_config.yaml
+bincio extract --since 2025-01-01        # only files newer than date
+bincio extract --file ride.gpx           # single file → JSON on stdout
+bincio extract --input ~/rides \
+               --output ~/bincio_data    # override config paths
+```
+
+Re-extraction is safe — unchanged files are skipped (hash-based dedup).
+To force a full re-extract: `rm -rf ~/bincio_data && bincio extract`
+
+---
+
+## Site
+
+```bash
+cd site
+
+# Symlink data (do once)
+ln -sf ~/bincio_data public/data
+
+# Dev server with hot reload
+npm run dev                              # → http://localhost:4321
+
+# Production build
+npm run build                            # → site/dist/
+
+# Preview production build locally
+npm run preview
+```
+
+---
+
+## Python / tests
+
+```bash
+uv sync                                  # install / update deps
+uv run bincio --help                     # CLI reference
+uv run pytest                            # full test suite
+uv run pytest tests/test_fit.py -x       # single file, stop on first fail
+uv run pytest -k "sport"                 # run tests matching keyword
+uv run pytest -v                         # verbose output
+```
+
+---
+
+## Data store layout
+
+```
+~/bincio_data/
+  index.json                    ← feed index (all activities, summaries)
+  activities/
+    2024-05-15T08:30:00Z.json   ← full detail + 1Hz timeseries
+    2024-05-15T08:30:00Z.geojson  ← simplified GPS track
+```
+
+Activity ID format: `YYYY-MM-DDTHH:MM:SSZ` (UTC, always Z suffix).
+IDs are stable — safe to use in bookmarks and links.
+
+---
+
+## extract_config.yaml — key fields
+
+```yaml
+owner:
+  handle: yourname
+  display_name: Your Name
+
+input:
+  dirs:
+    - ~/Activities          # scanned recursively for GPX/FIT/TCX/.gz
+  metadata_csv: ~/strava_export/activities.csv   # optional
+
+output:
+  dir: ~/bincio_data
+
+default_privacy: public     # public | blur_start | no_gps | private
+incremental: true           # false = re-process everything
+track:
+  rdp_epsilon: 0.0001       # GPS simplification — larger = fewer points
+  timeseries_hz: 1          # samples/sec in stored JSON (1 = 1 Hz)
+```
+
+---
+
+## Privacy
+
+| Value | Track served | Stats | In index |
+|---|---|---|---|
+| `public` | Full GPS | ✓ | ✓ |
+| `blur_start` | First/last 200 m removed | ✓ | ✓ |
+| `no_gps` | None | ✓ | ✓ |
+| `private` | None | ✗ | ✗ |
+
+Set per-activity in a sidecar `.md` file, or globally via `default_privacy`.
+
+---
+
+## Sports
+
+Canonical sport values: `cycling` `running` `hiking` `walking` `swimming` `skiing` `other`
+
+Sub-sports: `road` `mountain` `gravel` `indoor` `trail` `track` `nordic`
+
+FIT files: sport is read from the `sport` frame, with `session` frame as fallback.
+Strava CSV: `Activity Type` column overrides the FIT-detected sport (authoritative).
+Mapping lives in `bincio/extract/sport.py`.
+
+---
+
+## Patching activities (manual fixes)
+
+When re-running extract isn't practical, patch the JSON directly:
+
+```bash
+# Fix sport for a single activity
+python3 -c "
+import json
+p = 'site/public/data/activities/2025-03-16T113005Z.json'
+d = json.load(open(p))
+d['sport'] = 'skiing'
+d['sub_sport'] = 'nordic'
+json.dump(d, open(p,'w'), separators=(',',':'))
+"
+
+# Then update the index.json to match
+python3 -c "
+import json
+idx = json.load(open('site/public/data/index.json'))
+for a in idx['activities']:
+    if a['id'] == '2025-03-16T113005Z':
+        a['sport'] = 'skiing'
+        a['sub_sport'] = 'nordic'
+json.dump(idx, open('site/public/data/index.json','w'), separators=(',',':'))
+"
+```
+
+---
+
+## Common diagnostics
+
+```bash
+# Count activities by sport in the data store
+python3 -c "
+import json, glob
+from collections import Counter
+files = glob.glob('site/public/data/activities/*.json')
+c = Counter(json.load(open(f))['sport'] for f in files)
+print(dict(c.most_common()))
+"
+
+# Find activities with 0 distance
+python3 -c "
+import json, glob
+for f in glob.glob('site/public/data/activities/*.json'):
+    d = json.load(open(f))
+    if (d.get('distance_m') or 0) == 0 and d.get('sport') != 'other':
+        print(d['id'], d['sport'], d['title'])
+"
+
+# Find activities still tagged 'other'
+python3 -c "
+import json
+idx = json.load(open('site/public/data/index.json'))
+others = [a for a in idx['activities'] if a['sport'] == 'other']
+for a in others[:20]:
+    print(a['started_at'][:10], a.get('source','?'), a['title'])
+print(len(others), 'total')
+"
+```
+
+---
+
+## Key files
+
+| File | Purpose |
+|---|---|
+| `extract_config.yaml` | Main config (input dirs, output dir, privacy) |
+| `SCHEMA.md` | BAS format specification |
+| `CLAUDE.md` | Dev notes, gotchas, design decisions |
+| `bincio/extract/sport.py` | Sport name normalisation + mapping |
+| `bincio/extract/metrics.py` | Distance, speed, HR, elevation computation |
+| `bincio/extract/parsers/fit.py` | FIT file parser |
+| `site/src/components/ActivityFeed.svelte` | Feed page — card grid + sport filter |
+| `site/src/components/StatsView.svelte` | Stats page — heatmap + year totals |
+| `site/src/components/ActivityMap.svelte` | MapLibre GL map |
+| `site/src/components/ActivityCharts.svelte` | Observable Plot charts |
+| `site/src/lib/format.ts` | `formatDistance`, `formatDuration`, sport icons/colors |
+| `site/src/lib/types.ts` | TypeScript types mirroring BAS schema |
+| `site/astro.config.mjs` | Astro + Vite config (MapLibre GL workarounds) |