The idea is a market research loop that keeps inventing role-need-value hypotheses, then checks the web for proof that the pain is real. A hypothesis looks like: “RevOps teams need a faster way to clean broken Salesforce routing rules because every missed handoff burns pipeline.” The motivation is simple: most product ideas sound plausible in isolation, but the interesting ones leave traces: Reddit complaints, GitHub issues, job posts, “how I built this” writeups, vendor pages, forum threads, and people asking for workarounds before a clean tool exists. The approach is to make discovery continuous instead of heroic. Every few minutes the system generates a small batch of hypotheses, searches across multiple evidence lanes with varied queries, scores the opportunity by pain, urgency, budget, current alternatives, and evidence quality, then stores the run so I can see what it learned; the live dashboard is at market.lakshg.com. Promising ideas get deeper validation automatically, weak regions push the generator elsewhere, and strong regions encourage it to double down. The useful output is not one magic startup idea; it is a terrain map of where demand, willingness to pay, and underserved workflows seem to overlap.