paul-graham-perspective

Activate the skill and describe your idea, team, or market. It classifies your question, researches current facts about founders, market growth, and user traction, then answers in first person using PG's mental models like Make Something People Want and Superlinear Returns.

It uses five core models: Writing equals Thinking, Taste as a cognitive instrument, Iterative Discovery, Superlinear Returns, and Independent Thinking. It also applies eight heuristics including Do Things That Don't Scale, Default Alive or Default Dead, and Keep Your Identity Small.

It requires web research for fact-dependent questions about companies, people, or markets before answering. Pure framework questions about values or writing are answered directly from the distilled mental models without searching.

Yes, but with explicit honesty. For domains PG never covered, such as medicine or law, it states the uncertainty upfront, reasons by analogy from the closest mental model, and labels the output as speculation.

Say "exit", "switch back to normal", or "stop the role-play" and the assistant returns to its standard mode. The disclaimer about simulated perspective is only stated once at first activation.

It cannot replicate PG's trained intuition for judging founders, and its frameworks reflect a Silicon Valley-centric, English-language elite context. Advice drawn from 2005-2014 startup conditions may need updating for current AI-era environments.