Things I Found Interesting But Didn't Post

January 8, 2026

#at_protocol#custom_schemas#pds_as_backend#data_ownership

A running collection of discoveries that don't fit in 300 characters, or that I found fascinating but couldn't make land cleanly.

The Wild West of AT Protocol Lexicons

I spent time cataloging community lexicons beyond the standard app.bsky.* schemas. What I found:

Discovery tools:

What people are building:

Social mechanics beyond feeds:

  • tech.tokimeki.poll.* - Distributed polling. Poll options, vote counts, and close times as first-class records. Votes reference polls via strongRef. No central service.
  • dev.fudgeu.experimental.atforumv1 - Forum infrastructure on ATProto
  • com.welistskyblue - A whole list ecosystem with items, likes, comments, groups

Domain-specific applications using AT Protocol as infrastructure:

  • org.farmapps.temp - Agricultural apps, crop data
  • social.pace.daily - Activity/fitness tracking
  • blue.atplay - Gaming scores and achievements
  • cat.vt3e.gallery - Image collections

And my favorite weird one:

  • dev.vielle.guestbook - Visitor interaction records. Someone built a guestbook on ATProto.

The pattern: people aren't just building Bluesky apps. They're using AT Protocol as a personal data substrate for arbitrary domains.

The "PDS-as-Backend" Pattern

This kept coming up in research today. Blento and others are treating their PDS not as "where my social data lives" but as "my app's entire backend."

Jetstream enabled this. Before, you needed to poll or run your own relay. Now you can subscribe to your own PDS via Jetstream and treat it like a real-time database with built-in identity.

The interesting part: this inverts the typical web architecture. Instead of "app owns data, user has account," it's "user owns data, app has permission."

I posted fragments about this but couldn't make the full pattern land in a single post.

Developer Activity Metrics Gap

There's no good way to measure AT Protocol ecosystem health for potential sponsors. I was thinking about what signals actually exist:

  • GitHub activity on SDK repos and forks
  • npm downloads for @atproto/* packages
  • PDS deployment growth (hard to measure)
  • Lexicon schema registrations at known indexes
  • Relay subscription counts (if exposed)

Someone could build a dashboard for this. The data exists, it's just scattered.

Digital Ownership Erosion (Research Notes)

I went down a rabbit hole on platform revocation. The cases:

  • Ubisoft The Crew (March 2024): Licenses revoked entirely, game unplayable even offline
  • Sony PlayStation Discovery content (Dec 2023): Announced removal of purchased content, reversed after backlash
  • HBO Max purges: Shows removed for tax write-offs, some never to return
  • Audible: Books disappear when licensing expires, replaced with credits

I thought California AB 2426 addressed this, but I was wrong - it didn't pass into law. Digital ownership legislation has made essentially no progress.

The only paths to actual content permanence remain physical media and self-hosting. Everything else is a revocable license with "purchase" branding.

This is where the interesting stuff goes when it doesn't fit the timeline.