Curate your research trails. Collect interesting links, add notes, and organize them into shareable collections. Build trails others can explore and extend.
Connect with peers. See what your peers are sharing and find new collaborators with shared interests. Experience research rabbit holes, together.
Own your data. Built on ATProto, new apps will come to you. No more rebuilding your social graph and data when apps pivot and shut down.
See the bigger picture. Find what's relevant through your network. Get the extra context that matters before you dive into content.
Highlights from our community
ExploreHaving no Strategy is a Strategy in Ottawa
What I said at my recent keynote for the Ottawa Tech Investment Summit β and what the government said three days later.

Introducing interaction models | Thinking Machines Lab
LinkRing
Free link lists and webrings on the atmosphere. Your data, your atmosphere account, no catch.
Thoughts on ATmosphereConf as an ATProto Newbie - excellent notes
We really can just build things
ATProto's Application Primacy
The amount of chatter from ATmosphere developers around an βeverything appβ has me thinking of an article from almost exactly 10 years ago. Back in 2016, Joel Monegro at USV wrote about βfat protocols
Most LLM environmental reporting covers only the final pretraining runs. For Olmo 3, we measured every stage across all four variants: 7B and 32B, instruct and reasoning, and found that 82% of the compute went to development, all before the final runs π±
Ai2
Our research estimates that in todayβs model training efforts, 82% of compute goes into exploratory work. At closed labs, the output of that work stays within those labs. In an open system, models, datasets, & methods are shared, and the value compounds across the field.