blogs.hn is for personal blogs! Non-technical blogs are okay too, but remember that the community is mostly folks from HackerNews.

Adding Blogs

This repo comes with a helper script refresh.ts to automatically grab blog info from a URL. But feel free to manually add/edit information to blogs.json!

1. Run the script (deno):

deno run --allow-net refresh.ts "https://taylor.town" "https://gwern.net"

2. Clean the output:

[
  {
    "url": "https://taylor.town",
    "title": "Taylor Troesh",
    "desc": "🐸🐸🐸🐸🐸🐸🐸🐸🐸🐸🐸🐸🐸🐸",
    "about": "https://taylor.town/about",
    "now": "https://taylor.town/now",
    "feed": "https://taylor.town/feed.xml",
    "active_at": "2026-08-01T10:00:00Z",
    "posts": 42,
    "cadence": 9.5,
    "hn": [
      {
        "created_at": "2023-04-04T11:42:33.000Z",
        "title": "How to be a -10x Engineer",
        "url": "https://taylor.town/-10x",
        "points": 875,
        "comments": 494,
        "id": "35438068"
      }
    ]
  },
  {
    "url": "https://gwern.net",
    "title": "Essays · Gwern.net",
    "desc": "Personal website of Gwern Branwen (writer, self-experimenter, and programmer): topics: psychology, statistics, technology, deep learning, anime. This index page is a categorized list of Gwern.net pages.",
    "keywords": "meta",
    "about": "https://gwern.net/about",
    "feed": "https://gwern.substack.com/feed",
    "hn": [
      {
        "created_at": "2019-01-21T12:08:15.000Z",
        "title": "On whether changes in bedroom CO2 levels affect sleep quality",
        "url": "https://www.gwern.net/zeo/CO2",
        "points": 576,
        "comments": 306,
        "id": "18959796"
      }
    ]
  }
]

3. Add to blogs.json in a pull request. To prevent merge conflicts, please don't append to the top or bottom!

A daily workflow refreshes a rotating subset of blogs: it fills in missing fields (title, desc, feed, about, now, and github/bluesky/x/mastodon profiles found on the blog's homepage or /about page) and updates HackerNews stories. It never overwrites existing values, so hand-curated edits are safe.

Every write also normalizes the whole file: fields are sorted, text is trimmed, and anything that cannot be attributed to the blog is dropped (off-site /about and /now links, feeds belonging to somebody else, HackerNews stories pointing at another domain). Run it by hand with:

deno run --allow-read=blogs.json --allow-write=blogs.json refresh.ts --fmt

Activity

active_at, posts, and cadence are read from the blog's feed: the date of the newest post, how many dated entries the feed carries, and the median days between posts. They refresh on every run rather than being filled once, and they are dropped if the feed goes away. Blogs that advertise no feed get the conventional paths (/feed.xml, /index.xml, /feeds/posts/default, …) probed before we give up.

What the site shows

blogs.json is the whole directory, and the OPML export carries every feed in it. The front page is smaller on purpose: a blog is eligible only if it has a feed, has posted within two years, and says something about itself, and the page then draws a weighted random sample of 1000 from that pool — better blogs are likelier to appear, but the page rebuilds hourly, so the rest get their turn. Being left off the page is not a judgement on a blog; it usually means we could not find a feed.