Recorded investigations¶
These examples let you inspect the research process: the operator’s question, what the agent chose to run, the resulting data and the review decisions. A preserved study can be read without an API key. Fresh agent runs need a configured backend; numerical reproduction and record verification are separate operations.
Plasma research¶
Start with the pictures: these plasma walkthroughs include actual field maps, evidence maps, short explanations of each measurement and downloadable figures. Their linked receipts distinguish completed work from outstanding tests.
Magnetic-island coalescence with FLASH presents a fresh v0.7.0 investigation that independently falsifies a bounded reconnection-rate scaling hypothesis. The explanatory repair remains unvalidated. Inspect the real fields, failed controls, refinement decisions and accepted counterexample.
A FLASH current sheet followed with WarpX presents a fresh v0.7.0 investigation with an independently supported local pressure prediction. It links a new MHD source to two prospective kinetic particle realizations, with numerical allowances and accepted claim/report reviews. Both examples retain the older studies in separate historical walkthroughs and immutable audit records.
Qin-1: from hardware design to circuits, fields and 3D plasma connects actual CAD changes, Gmsh, Elmer/RLC, native WarpX and PyVista. Explore the interactive fields, the stronger cold-conductor design and the five failed plasma flux-sensitivity checks. This operator-assisted host study demonstrates the broader toolchain; it has no formal harness-adjudicated scientific verdict.
A complete minimal-workflow example¶
Does energy conservation guarantee an accurate orbit? uses a fresh Kepler simulation study to exercise counterexample search and the repair completion policy. Ordinary CPU Python is enough. It falsified the original claim and supported a four-case repair after fresh validation. The record identifies human framing, agent work, independent reviews and the observed outcome.
Capability and toolchain reference cases¶
Case |
Scope and verification |
|---|---|
Operator-assisted CAD → Gmsh → Elmer/RLC → 3D WarpX → PyVista study with retained sensitivity checks ( |
|
Material-specific diamond-Si ( |
Earlier v0.1–v0.6.2 classic-workflow demo directories have been retired from
the active tree and remain available in the
v0.6.2 release tag
and research archive.
Research audits¶
Longer investigations also produce useful records when they leave a question open. The adaptive aluminium Z-pinch audit examines actual 3D trajectories, energy diagnostics, unresolved reconnection attribution and the cost of the research process. Its data and limitations are kept with the original dated report.
To contribute a study, include the operator input, version and source identity, actual run records, verification instructions and a clear account of who did which work. See documentation development.