Research program C
Whether sustained collaboration exposes failure patterns that single-turn evaluations cannot see.
Long collaborations generate a record that short evaluations never produce: decisions, corrections, handoffs and their consequences days later. This program turns that kind of record into testable questions and then into repeatable protocols, keeping the line between anecdote and result explicit at every step.
Observations from real collaboration can motivate a question. They become findings only after a controlled test supports them, and private conversations are never published without authorization.
Do models keep constraints, corrections and earlier decisions intact across extended, multi-session work? This entry defines the question and the evidence that would answer it. It reports no results.
