The Free-Recipe Limit: research artefact (pre-registrations, analysis code, per-run results, and derived verdicts)

Frozen artefact for the paper The Free-Recipe Limit: every recipe effect measures which premise of an idealised learner broke. It contains the pre-registrations (each frozen in the job script or markdown note before the run it governs), the full analysis tree, every scored evaluation the paper draws on with per-problem score vectors retained rather than only aggregates, and the derived verdict files the paper quotes. Model checkpoints are deliberately excluded (the archive is 9.4 TB). Every claim in the paper is a function of the shipped per-run result files, so any number in the paper can be recomputed without a GPU: tools/verify_artifact.py re-runs each analyser against the shipped results and reports, per analyser, whether it reproduces its shipped verdict byte-identically. Companion papers in the same family are archived separately.