What this is

Why

AI news is a flood. The same announcement gets repackaged across 14 newsletters, three trade publications, and a dozen X threads within hours. Most of it is the same story.

A good summary should be enough to decide whether to read further. A good headline should be enough to decide whether to skip an entire category. That's what this feed is built for.

How it works

1. Ingest. Polls labs, blogs, news outlets, podcasts, and social posts on 5–60 minute cycles.
2. Triage. An LLM filters items down to genuinely AI-relevant content — not ads, not summaries of summaries.
3. Cluster. Embeddings + LLM verification group items covering the same underlying story, regardless of how each outlet framed it.
4. Synthesize. The cluster is rewritten into a single card with a tight headline, a 2–3 sentence summary, and links to every source.
5. Theme. Cards are grouped into broader themes (model releases, policy, robotics, tools, business) so you can skip a category in a single glance.
6. Rank. Source authority, story velocity, and engagement signals decide what surfaces at the top.

Sources

A selection of feeds the system reads continuously:

• OpenAI Blog
• Anthropic News
• Google DeepMind
• Hugging Face
• Meta AI
• Microsoft Research
• Hacker News
• MIT Technology Review
• The Verge AI
• Ars Technica
• Stratechery
• Import AI
• The Batch (deeplearning.ai)
• Sequoia
• a16z
• Latent Space
• Simon Willison
• Karpathy
• Lex Fridman
• TLDR AI

…and many more. The full list grows as new high-signal feeds appear.

Privacy

The feed is fully readable without an account. Sign in (passkey or Google) only if you want preferences synced across devices. Analytics are aggregate and anonymous — privacy policy.