Research

Research and evidence

Research that proves the product.

Research outputs are judged by reproducibility: controlled runs, explicit uncertainty and evidence a reader can check independently.

1. Research planning

   The AI Research Scientist agent frames hypotheses, builds experimental plans, selects metrics and lists risks before any run starts, so every study has a defined finish line.

2. Source synthesis

   Comparison of sources into evidence maps with explicit uncertainty marking. Where evidence is incomplete, the gap is stated rather than filled with a guess.

3. Paper workflow

   Structure, text, citations, diagrams and export into professional documents, keeping the same numbering, claims and figures consistent across drafts.

4. Repository research

   Architecture, tests and experimental results are cross-linked, so documented claims tie back to the code and runs that produced them.

5. Product evidence

   A compact V271 interface map presents the documented product areas — agent navigation, workspace and controls — in a clean, readable form.