About¶

This section covers two things: how the investigation was attributed and methodologically anchored, and what the open follow-on question is.

• Attribution — Every artifact in this repo is a canonical output of the ml-lab workflow (hypothesis → adversarial critique → PoC → empirical resolution → peer review). Short page noting the methodology lineage.

• v3 idea — The follow-on investigation idea, captured explicitly as an idea only — not a spec, not a pre-registration, not a commitment to run. Documents the gap v2 leaves open (single-turn measurement of multi-turn-designed prompts), what a v3 design would need (multi-turn tasks, user simulator, session-level judges, reframed conditions), and what v3 would and would not settle.

Neither page is necessary to read the findings or apply them. They exist for readers who want methodological context or who are considering picking up the v3 thread.