The runtime
your research runs on.

Records, typed relationships, reasoning, and expression — four parts with distinct jobs, so you can see where an answer came from and where its authority ends.

Query stored knowledge, follow the explanation path, and check known conflicts. When the graph lacks a relationship, Lois returns unknown instead of a guess.

Five layers — literature, molecular data, computation, the source contract, pathways — plus WetHands experiments. Every result names its database, or it is an error.

PubMed to AlphaFold to KEGG on one console, the compute bench, the WetHands lane, and pure C++ on the metal — instruments that were never meant to connect, talking.

Reasoning runs on local Ollama models by default; nothing leaves your devices to train any model; external models are opt-in; and the record — graph, ledger, notebook — is files on your disk that work with the cable pulled.

Evidence maps, pathway review, hypothesis comparison, and continuity across a lab — the proposed research workflow we want to evaluate with partners.

Released datasets and models, the prototype's real scope, the public papers, and the validation path — what is built, what is next, and what still needs a test.