News
News & field notes.
Launch coverage, benchmark write-ups, and field notes on Scholialang — what we are shipping and what the evals actually show.
Latest.
Field note
Why reasoning needs a protocol layer
The launch field note: why agent work needs a visible reasoning layer, with the early v0.6 pilots — a three-model replay that matched 135/135 trace-seeded cells, plus a 41% input-token compounding result.
Read the field note → Press releaseDoug Fir Labs launches Scholialang
The launch press release — an open protocol for portable AI agent reasoning: content-addressed traces, DAG-backed reuse, and lazy context preludes, with early three-model replay and compounding pilot results.
Read the release → Release notesWhat’s new in v0.6
Content-addressed atoms, the lazy prelude, and the registry — the substrate that lets reasoning traces compound across sessions.
See what’s new →