The engine wakes up within about a minute of an order.
Gate 1 runs on the engine: wide candidate phrases → only the ones with measured demand → store results, 1–2★ review themes, pricing, community → the direct rivals and whether they are winning → GO / MAYBE / KILL, and whether the searchers add up to the money you want. Needs DataForSEO and Claude under Connections → Validation. ~$0.50 per idea. The launch copy is step 2: written only when you decide to go ahead with an idea, from the run's cache, so only that one Claude call is spent.
Three steps, two stops. 1 Claude writes down the problem behind the idea; you confirm or rewrite it (≈ $0.01). 2 A wide hunt finds every product that might be a rival — App Store, Play, Google, products Claude knows of; you go through the list and keep the real ones (≈ $0.40). 3 The confirmed rivals are measured: store facts, prices, revenue, the phrases feeding them, their users' complaints, an economics sheet (≈ $0.60). No score, no verdict — the stats, for you to weigh.