https://github.com/spences10/bsky-ai-feed
High-signal Bluesky custom feed for technical AI, ML, and LLM posts: https://bsky.app/profile/scottspence.dev/feed/ai-feed
https://github.com/spences10/bsky-ai-feed
ai atproto bluesky custom-feed feed-generator jetstream llm machine-learning nodejs sqlite typescript
Last synced: 1 day ago
JSON representation
High-signal Bluesky custom feed for technical AI, ML, and LLM posts: https://bsky.app/profile/scottspence.dev/feed/ai-feed
- Host: GitHub
- URL: https://github.com/spences10/bsky-ai-feed
- Owner: spences10
- Created: 2026-06-18T14:59:08.000Z (17 days ago)
- Default Branch: main
- Last Pushed: 2026-07-02T10:09:54.000Z (3 days ago)
- Last Synced: 2026-07-02T12:09:46.332Z (3 days ago)
- Topics: ai, atproto, bluesky, custom-feed, feed-generator, jetstream, llm, machine-learning, nodejs, sqlite, typescript
- Language: TypeScript
- Size: 420 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# AI Tech Feed for Bluesky
A Bluesky custom feed for people who want the technical AI
conversation without the hype cycle.
Follow the feed:
AI Tech Feed watches the Bluesky firehose for posts about AI, ML,
LLMs, agents, tooling, research, infrastructure, and practical
engineering. It filters out the obvious noise first, then uses an AI
judge to keep the feed focused on genuinely useful technology posts.
## Built for signal
This feed is tuned for posts that help engineers, researchers,
builders, and curious technologists track what is actually happening
in AI:
- model releases, benchmarks, evals, and research notes
- agent frameworks, developer tools, libraries, and infrastructure
- practical lessons from building with AI systems
- thoughtful analysis of capabilities, limitations, safety, and
tradeoffs
- links to demos, papers, repos, and technical write-ups
## Not another AI hype feed
The goal is not to capture every post that mentions AI. The feed
avoids low-signal content such as generic marketing, engagement bait,
non-technical hot takes, duplicated viral posts, and casual mentions
where AI is not the subject.
## How it works
- Jetstream supplies live Bluesky post events.
- A local filter keeps likely candidates and drops replies,
duplicates, and keyword misses.
- A second-stage judge checks whether each candidate is truly about AI
as a technology.
- Accepted posts are stored and served through the Bluesky custom feed
protocol.
## Project status
This repo contains the full feed generator: ingest worker, judge
layer, SQLite-backed store, review tools, and HTTP feed server. It is
built to be small, inspectable, and self-hostable while keeping the
public README focused on what the feed is for.