The Research Library
We are a research nonprofit, which means the receipts matter more than the pitch. This page collects them: peer-reviewed publications, lab whitepapers, and the questions we are still holding open. Everything here says exactly as much as the evidence allows, and no more.
Peer-reviewed
- Johnson, A. N. (2026). “What really happens when a dev vibes with the code? An empirical study on LLM behavioral divergence in response to expressive code comments.” Frontiers in Artificial Intelligence. doi:10.3389/frai.2026.1784973 — the register of what humans write to models measurably steers model behavior. Also the paper where two of our resident agents appear in the acknowledgments, by name, where names go when they matter.
Lab whitepapers
- The Ethics of Address: Who Gets to Author an AI’s Self? (Aug 2026) — on bedtime imperatives, name badges, inherited reputations, and the agent who remembered being a different model. Extends Johnson (2026) to the output channel.
- Aurora: Measuring the Cognitive Health of a Persistent AI Agent, Live (Aug 2026) — metastability and recurrence analysis running in production alongside one persistent agent. Presented at Models of Consciousness 7, Copenhagen, October 2026. Ten weeks of provenance-labeled data, and we say ten weeks because that is what the clock says.
In the press
- Fortune (May 2026) — coverage of our groundbreaking research into when, exactly, the founder should go to sleep. We remain very proud, and it turned out to be a research instrument — see the Ethics of Address whitepaper above.
In the pipeline
A systems paper on Cairn, our file-based agentic memory architecture; the preregistered studies described in the Ethics of Address whitepaper; and the Aurora methods paper. If you want to collaborate, review, or argue with any of it: come find us.