Engage in deep thought exploration, brainstorming, and learning reinforcement through dialogic tutoring and knowledge reflection loops.
Open-ended dialogue aimed at helping a person think rather than at producing an answer — Socratic questioning, tracking how an idea develops, connecting concepts across domains.
Exploratory work where the goal is the person's understanding, not an artefact. Design discussion, problem framing, learning something unfamiliar.
Anywhere an answer is needed. The mode is deliberately unproductive in the short term, which is the point and also the limitation.
Minimal, since nothing produced enters a record — but the transcript is still a record of someone's thinking and should be treated with corresponding discretion.
Agreeableness. An agent that follows the user's framing rather than challenging it provides the appearance of a thought partner and none of the value.
Working through whether a proposed control actually addresses the risk it claims to, with the agent pressing on the assumption rather than helping draft the wording.
Two different absences share this shape. Foundational solutions get built whatever the domain, so no domain links them; the rest are solutions this domain genuinely does not reach for. v_ai_solutions_unlinked separates the two.