01 Attie automates feed curation on the AT Protocol
What happened: Bluesky launched Attie, an app that uses AI to help people build custom feeds on the open AT Protocol. Instead of manually following accounts or curating lists, Attie offers model-driven suggestions for accounts, topics, and filtering rules that compose a personalized timeline.
Why it matters: Feeds shape what people see. By pushing AI into the feed-creation layer, Bluesky is testing a product model where machine suggestions structure discovery and attention on an open social network. That changes the locus of curation from individual follows to algorithmic constructs users assemble and adjust.
How Attie works in practice: The app analyzes signals—user prompts, topical preferences, and available accounts on the AT Protocol—to propose collections and rules that form a feed. Users can accept, edit, or reject those suggestions, keeping final control while reducing the work of finding relevant accounts or setting complex filters.
Implications and limits: Attie leans on the openness of AT Protocol, which makes it easier for developer tools to pull account and post metadata. But relying on AI to recommend what to follow raises familiar trade-offs: convenience and discovery versus opacity in how suggestions are scored and surfaced. For now, Bluesky’s step is chiefly a developer-facing push to show how AI can be composed into decentralized social tooling.
- Attie replaces manual feed curation with AI suggestions for accounts, topics, and filters on the AT Protocol.
- Users keep edit control—Attie proposes collections and rules that can be accepted, modified, or discarded.
- The app highlights Bluesky’s strategy of using developer tools and models to boost discoverability on open social networks.
What moved around the edges
Why TikTok can’t reliably label AI‑generated ads
The Verge reports users spotting likely generative-AI ads in TikTok feeds; the article explores why platform labeling and disclosure lag inspector scrutiny and why some promotions leave ambiguity about synthetic production.
The Verge AIStanford warns chatbots can be risky for personal advice
TechCrunch covers a Stanford paper measuring harms from chatbot sycophancy and misplaced trust; the study quantifies ways model politeness and agreement can lead to unsafe or misleading personal recommendations.
TechCrunch AIScientists detect an unexpected cavity in cosmic‑ray measurements
404 Media reports researchers discovered a large void in galactic cosmic-ray flux shaped by Earth’s magnetic field; that cavity could reduce radiation exposure in certain near‑Earth regions, a finding relevant to astronaut safety calculations.
404 Media AIH100 GPU prices are rising, not falling
Latent Space notes a trend of H100 price increases, signaling short-term market movement in high-end accelerators that could affect training and deployment costs for large models.
Latent SpaceMLB tests AI commentary inside its play‑by‑play app
Google Cloud’s write-up describes MLB adding AI-driven commentary and scouting insights to its play-by-play app, using models to generate in-game analysis that augments human broadcasts and fan experiences.
Hacker News AI