Tolerance
Feeds are toxic. Build Tolerance.
A browser extension that scores content by how hard it's trying to manipulate you, then makes the feed progressively higher quality so you spend precisely the amount of time you wish to.
Website · Chrome · Firefox · Releases
How it works
-
Every post gets scored. AI analyzes each piece of content for manipulation signals: outrage bait, curiosity gaps, tribal triggers, engagement farming.
-
The feed gets boring over time. For the first 15 minutes, you browse normally. Then Tolerance starts blurring the most manipulative content. The longer you scroll, the more gets blurred.
-
You leave because you're bored, not blocked. Blocking creates craving. Boredom creates freedom. You reclaim your feed and close the tab on your own terms.
Supported platforms
- Reddit (old.reddit.com is better)
- Twitter/X
- YouTube
- Instagram
- Facebook (Chrome only; on Firefox, Tolerance only tracks time there)
- TikTok (time tracking only)
Installation
From the stores (easiest)
From release
Download the latest .zip for your browser from Releases.
Chrome:
- Unzip
tolerance-chrome.zip
- Go to
chrome://extensions
- Enable "Developer mode" (top right)
- Click "Load unpacked" and select the unzipped folder
Firefox:
- Unzip
tolerance-firefox.zip
- Go to
about:debugging#/runtime/this-firefox
- Click "Load Temporary Add-on"
- Select
manifest.json from the unzipped folder
Build from source
git clone https://github.com/gtech/tolerance.git
cd tolerance
pnpm install
pnpm build # Chrome
pnpm build:firefox # Firefox
pnpm build:all # Both (outputs to dist-chrome/ and dist-firefox/)
pnpm build and pnpm build:firefox output to dist/; load that folder as an unpacked extension.
Setup
Tolerance uses AI to score content. In the dashboard, choose one of:
- Free Tier. Try it without setup, with limited daily usage. Tolerance issues you an OpenRouter key; scoring calls still go straight to OpenRouter.
- Your own key. Get one from tolerance.lol/account.html or directly from OpenRouter (Keys → Create Key, then add credits; $5 is plenty to start), and paste it into the dashboard.
- A custom endpoint. Any OpenAI-compatible chat completions URL, including a local one such as Ollama (
http://localhost:11434/v1/chat/completions).
Posts are scored by anthropic/claude-haiku-5.5 with thinking off (text, images and video thumbnails; Instagram reels use google/gemini-2.5-flash-lite). With a custom endpoint, you choose the text and vision models yourself.
Cost: ~$1-2/month typical usage.
Features
Cheap-Payoff Scoring
Every post gets a 0-100 score for how hard it grabs attention for how little reading or thinking it asks, and how little it leaves behind. Outrage bait scores high, but so do a clever one-liner, a cute dog or a relatable meme; a dense explainer scores low. Scores appear as badges; hover over a badge to see the reason.
The scoring rubric lives in one place (src/shared/scoring-spec.ts) and is versioned: when it changes, older cached scores are re-scored. pnpm eval:scoring runs a set of test posts through it and reports which land outside their expected range (needs an OpenRouter key in OPENROUTER_API_KEY or test_key.env).
- Video counts as at least 40. Autoplaying video is a hook in itself, so video posts score at least 40.
- Blur until scored. New posts stay blurred until their score arrives (up to 15 seconds by default), so you never glimpse bait before it's judged. A gate at page load stops the feed flashing before Tolerance starts.
- Failed scores count as 50. If the API fails, the post is treated as a 50 and the error shows on hover.
- Threads and replies (Twitter). A reply under a bait post is folded with it, and a thread continuing a bait post inherits its score. Search results are left unfiltered.
- Debug a post. Right-click a tweet and choose "Tolerance: debug this post" to see its score, the reason, and why it was or wasn't blurred.
Progressive Boredom
| Time | What happens |
|---|
| 0-15 min | Normal browsing |
| 15-45 min | High-manipulation content blurred |
| 45-75 min | More aggressive blurring |
| 75+ min | Almost everything blurred |
Resets daily at midnight.
Quality Mode
One toggle. Instantly blurs everything above score 20. Only genuine content remains.
Hover to Reveal
Blurred content reveals after hovering for 3 seconds. Friction, not blocking.
Adaptive Calibration
Daily feedback ("too restricted / balanced / too easy") adjusts thresholds automatically. The dashboard shows what's being blurred right now, including any adjustment from feedback, and lets you reset the adjustment.
Grounding Breaks
After a stretch of continuous scrolling (20 minutes by default), Tolerance offers a short grounding exercise: a few slow breaths, 5-4-3-2-1, or feet on the floor. It's always optional, and a 5-minute break resets the clock.
Narrative Awareness
Tolerance tracks recurring narratives in your feed: themes you define, plus emerging ones it discovers each day for you to confirm or dismiss. The dashboard shows today's exposure to each. A counter-strategy (thesis, antithesis, synthesis) for a narrative can raise or lower the scores of posts that match it.
Hide Facebook Reels
On by default. Removes the Reels tray and posts that link to reels from your Facebook feed.
This is the beginning of the README. Read the rest at the original. Read the full README ↗
