I've been reading HN for years, but I've noticed a pattern: I'd open the front page to "stay current on tech etc." then emerge an hour later having read hundreds of comments on some political thread or startup drama. I built this extension to keep myself honest.

What it does:

The extension uses an LLM to analyze HN stories and adds metadata badges directly to the front page:

- Topic classification (AI/ML, Web Dev, Systems, Startups, Policy, etc.)

- Content type (Show HN, Tutorial, Paper, News, Discussion…)

- Technical depth score (0-1 scale)

- Tags (specific technologies, descriptors like "deep-dive" or "beginner-friendly")

You can then filter stories by topic/type, or set a technical depth threshold. My favorite feature is hiding just the comments link on non-technical stories so I'm aware of them without falling down the rabbit hole.

How it works:

- A Cloudflare Worker runs hourly, fetching top stories and queuing them for enrichment

- Gemini analyzes the title, URL, and top comments to generate structured metadata

- Results are cached in Cloudflare KV

- The extension fetches enrichments and injects badges into the page

Privacy-wise, it only sends public story IDs to the API, no personal data.

Chrome: https://chromewebstore.google.com/detail/enriched-hn/mmcdcoa...

Firefox: https://addons.mozilla.org/en-US/firefox/addon/enriched-hn/

GitHub: https://github.com/mjshashank/enriched-hn

Would love feedback on the categorization accuracy or feature ideas!