09THE RESEARCH / OPEN QUESTIONS
CORTEX studies how task-relevant, cross-person and verifiable cognitive structure can guide intermediate AI computation. The loop is a next step, not a foregone conclusion.
First distill existing cognitive trajectories. Then actively discover which trajectories are worth distilling next. The aim is an explainable, verifiable cognitive-transfer infrastructure that can also decline to update.
Risk-calibrated human neural trajectory distillation · revised manuscript
Full paper · PDFStatistical theory, path-compatible calibration, independent risk audit and full protocols
Supplement · PDFTheory, task roadmaps, experimental ladders and governance
Research proposal · MarkdownValidate the state machine, debiasing, sentinels and rollback.
Change selection only; freeze the student and train offline.
Fix the query order; update a small adapter to isolate stability.
Combine selection and updates with independent promotion audits.
Compare population-first, personal-first and private aggregation.
Test whether cognitive operators reduce human trials on new tasks.
Interested in cognitive distillation, explainable AI or closed-loop experiments? Get in touch.