A #DuckDB file doesn't have to contain any data. It can store nothing but view definitions pointing at files stored somewhere else, such as Parquet on object storage. The file stays a few hundred KBs no matter how large the data it describes. In this post, we create such a small shareable catalog:

Social media is great but its problem is that knowledge doesn't *accumulate* anywhere. So how about a social platform where, instead of just text, each post is a small knowledge graph? AI can help with adding the structure and also contribute independently to the graph. Social KG construction

Here are supporting sources on the barriers to the creation of worker co-ops: aspeninstitute.org/events/democratizing-work-the… emerald.com/books/edited-volume/15376/cha… journals.sagepub.com/doi/10.1177/13505076251320889 researchgate.net/publication/324907081_The_sca… project-equity.org/wp-content/uploads/2023/06/Wo… ncbaclusa.coop/content/uploads/2025/01/SBA-7…

Very cool project connecting atproto with the Underlay knowledge graph protocol "[The Underlay] is an archive. [atproto] is a wire. Together they give a collection both permanence and real-time distribution." And check out these neat #ATScience examples too! underlay.org/blog/2026-04-28-atproto-integ…

In case you didn't here, Bluesky is adopting the at-tags proposal!

Community notes over papers is another example of a new kind of annotation that goes far beyond merely a "mention" and can be easily incorporated into protocol powered research ecosystems like #ATScience

I am so looking forward to digging into this. Tools & services to help web builders integrate their sites onto the atmosphere could turn the entire web into an Atmospheric recruitment service...