Experiments in building things that probably shouldn't work.https://mightwork.io/https://mightwork.io/experiments/organizational-signal-intelligence/https://mightwork.io/experiments/organizational-signal-intelligence/I built a production agentic pipeline that ingests customer signals from five sources — NPS surveys, customer success notes, win/loss data, in-app feedback, and internal dogfood — clusters them semantically, runs them through 23 LLM agents, and delivers prioritized product specs as structured documents routed to the right PM weekly. The question wasn't whether AI could help a PM work faster. It was whether a PM could build the intelligence infrastructure an entire organization runs on.Sun, 01 Mar 2026 00:00:00 GMThttps://mightwork.io/experiments/seven-apis-one-binary/https://mightwork.io/experiments/seven-apis-one-binary/I built a production CLI tool in Go that migrates data from Asana, Monday.com, Trello, Jira, Airtable, Notion, and Wrike into Smartsheet — non-destructive, resumable, and full-fidelity. I am not a Go engineer. The question I was actually running was whether product thinking transfers directly to engineering problems when the implementation layer is handled by AI, and what that handoff actually looks like in practice.Sun, 01 Feb 2026 00:00:00 GMT