"In order to direct attention to constructive things and away from controversial irrelevancies, we shall invent an imaginary being. Its brain is to be designed by us, so that it reasons according to certain definite rules." — E. T. Jaynes, Probability Theory: The Logic of Science
Jaynes opened his book with a design brief: a robot that reasons plausibly and consistently from its prior information. Large language models finally make the brief buildable — what was missing was not the reasoner but the organized prior information.
That is what memoires provides: a curated, evidence-graded, contradiction-aware catalog of Bayesian modeling craft. The robot is the consultant that reads them. Not a solver that answers for you, but an advisor that watches a model-development process — your own, or a team of agents' — and offers the right steer at the right moment: the divergence you're misreading, the parameterization the geometry is asking for, the precedent that doesn't cross this boundary, the discriminating experiment you skipped. Grounded in the catalog, delivered as counsel rather than recall.
Status: design phase. See DESIGN.md for the architecture — the advisor pattern, and
why timing (advise after a position exists, never before) is the load-bearing idea. The catalog ships
first; the robot follows.
Apache-2.0 (code). The knowledge it consults lives in memoires under CC BY 4.0.