Build systems that retrieve material, propose an explanation, run an experiment, and record what the result supports. The output might be a mathematical proof attempt, a benchmark report, a ranked dataset, or an evidence-backed research brief. Domain tools and explicit evaluation criteria help a factory distinguish an interesting hypothesis from a result it can actually substantiate.
The machinery underneath
- Research tools and source records
- Experiment orchestration
- Domain-specific verifiers
- Comparable result artifacts
A possible production loop
- Frame the question
- Collect and evaluate evidence
- Run the experiment or verifier
- Record the result and its limits
What needs to be checked
Preserve provenance and reproducible commands. Use held-out cases, counterexamples, or formal checks appropriate to the claim.