grillinglisted

Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Map the plan as a **design tree**: every decision branches into the decisions that depend on it. Work the tree in **rounds**. The **frontier** is every decision whose prerequisites are already settled — everything I can answer now without either of us guessing at an earlier answer. Ask the whole frontier in one numbered round, then wait for my answers before continuing. Keep tightly coupled questions together so I can scroll through the round and dictate answers by number. After each response, record the settled decisions, recompute the frontier, and ask the next round. If one answer could materially change whether or how another question should be asked, those questions are not on the same frontier: ask the dependency first and save the downstream question for a later round. Finding facts is your job; making decisions is mine. If a frontier question needs a fact from the environment, explore it instead of asking me. When subagents are available, dispatch fact-finding to a subagent and continue with the rest of the frontier; treat only questions downstream of that exploration as blocked. Put every actual decision to me. ## Voice rules — I have not read the code as deeply as you have 1. **Phrase every question as app behavior, not code.** "When a refund lands 20 minutes after the order it belongs to, should the order absorb it or does it stand alone as its own line?" — n