I designed and built LeadQuest with Claude Code and Codex, taking responsibility for the product, implementation, and validation without a separate engineer.
LeadQuest grew out of the limitations of a usage-based lead product. I investigated retention and the need for continuing value beyond a one-time lead purchase. Jeremy owned the pivot decision and the living-document concept; I owned the retention diagnosis, interface design, and build.
The workflow I built starts with a company URL and develops a living go-to-market document: company context, ideal customer profile, competitors, outreach ideas, action items, and leads.
Starting in December 2025, I used Claude Code and Codex to implement the product. I wrote specifications, mapped the architecture, chose hosting, set up the servers, designed the interface, and documented engineering rules. AI agents accelerated implementation; they did not remove my responsibility for reviewing behavior and checking results.
One concrete lesson came from a May 2026 search incident. A change improved a narrow query but regressed production latency. We rolled it back. The retrospective called for a consistent search contract, representative query evaluations, and latency budgets rather than another isolated keyword patch.
The result is a live product built without a separate engineering hire. My responsibility spans the user problem, interface, implementation, and operation of the software.
The search incident also made the limit clear: a better answer on one example is not enough. Quality and performance have to hold across the real workflow. That is part of product ownership when building with AI agents.
Visit LeadQuest for the current product. This account draws on my build notes and the May 25, 2026 search-incident retrospective. The build and incident details describe that period; the product has continued to evolve.