Writing
PRDs, teardowns and notes on building AI products that work.
14 Oct 2026·5 min readHow I evaluate LLM features before they ship
The 6-step framework I use for RankFlowHQ and HarshOnRoads to decide whether an AI feature is good enough to ship: what good means, a real test set, quality, failure modes, cost and launch gates.
AI Product ManagementEvalsFramework
14 Oct 2026·5 min readPRD: LLM-assisted client mapping for a lead platform
A sample PRD for adding AI to the lead management platform I built: LLM-suggested field and course mappings, approved by a human and validated by the system, to cut new-client onboarding from hours to minutes.
PRDAI Product ManagementB2B SaaS
14 Oct 2026·4 min readTeardown: Perplexity, and what an answer engine has to get right
A product teardown of Perplexity from someone who builds grounded AI content tools: the core loop, where citations earn trust, where it breaks, the metrics I'd watch, and three things I'd build next.
TeardownAI Product ManagementAI Search