How the smartest founders are winning in AI

TL;DR

Jason built a digital version of himself using Delphi AI, ingesting 20 million words of his content (blogs, tweets, interviews) and training it ~10 minutes per day. The AI had ~50,000 conversations and performed better than him at answering B2B SaaS questions, reviewing pitch decks, and even acting as a “therapist” for founders. The core conclusion: zero-to-one revenue in AI is now very likely (e.

Key insights

• A hyper‑trained AI (20 million words) hallucinates very little and can connect information the human creator cannot (e.g., linking comments made years apart).
• Users reveal things to AI they would never tell a human, including deep fears about co‑founders, board issues, and personal struggles—the therapy use case is massive.
• The moat from personalized AI is real but not as deep as people think: a motivated user can catch up with Claude in about a month of daily training.