Antechamber

Add friction to YouTube distractions and reward focus — adaptive, and fully on-device.

Get it on the Chrome Web Store →

---

Antechamber is a browser extension that puts a small, deliberate pause in front of distracting YouTube videos and rewards you for staying focused. It learns what you consider productive vs. distracting, and does all of its thinking locally — no servers, no tracking, no data leaves your machine.

---

How it works

• On-device classification. A small machine-learning model (MiniLM via transformers.js) runs entirely inside your browser. It reads video titles and channel names on YouTube and decides whether each one is productive, a distraction, or neutral — by comparing them against the interests, example videos, and channels you set during onboarding.

• Friction, not blocking. When you land on a distracting video, Antechamber shows a calm "gate" — a brief, dismissible pause with a countdown — instead of hard-blocking it. You stay in control; the goal is intention, not restriction.

• Rewards for focus. Productive viewing earns points, levels, and streaks, with gentle on-page reward toasts.

• Adaptive & self-tuning. It calibrates to your real YouTube homepage during onboarding and keeps learning from your corrections — re-label any verdict and the whole extension re-checks live, no reload needed. Channels you rate consistently get auto-labeled and short-circuit classification.

• Private by default. All your data — profile, settings, stats, and classification cache — stays on your device in storage and IndexedDB. See PRIVACY.md for the full story.

---

Privacy

Antechamber collects nothing. There is no analytics, no remote logging, and no backend server. The only network requests are:

1. A one-time ~20 MB model download from the Hugging Face CDN (no personal data).
2. Optional calls to the YouTube Data API — only if you paste in your own API key — to improve classification accuracy.

Full details in PRIVACY.md.

---

Tech stack

• WXT (Manifest V3, Chrome + Firefox)
• React 19 + TypeScript
• @huggingface/transformers for on-device embeddings (run in an offscreen document)

---

Development

pnpm install

pnpm dev            # Chrome, hot-reload
pnpm dev:firefox    # Firefox

pnpm build          # production build (Chrome)
pnpm build:firefox  # production build (Firefox)
pnpm zip            # package for the Chrome Web Store
pnpm compile        # type-check only

After pnpm build, load the unpacked extension from .output/chrome-mv3/ via chrome://extensions (Developer mode → Load unpacked).

---

License

MIT