Hi TabulaRasa, thanks for sharing this!
From an HCI angle, I’d suggest tightening the validation first: expand beyond 2 subjects (at least 6 to start, ideally 10-20), and define measurable comprehension metrics (e.g., accurate recall, misinterpretation rate, and transfer to new questions). Also, concepts like, lossless re-encoding, might need a concrete integrity check plan.
BTW, if you wanna move from concept to evidence, Spark could be a fit for a small PoC/MVP: a simple CP test + document translation demo + a small community study with a short report. You can also recruit community volunteers as study participants.