A reproducible study
Compare a baseline and one change, preserve all outcomes, and defend a bounded conclusion.
Prepare a protocol · P10 implementation studio · Original two-game research notebook
Version 1: a selected regression branch using three authored synthetic tasks. This is a scaled-down reproducible study, not a reproduction of a frontier paper. Data are public; “final” means held back by this workflow until freeze. Previously exposed final tasks remain practice.
Loading saved study…