Strengths & Weaknesses

Draws parallels between Amazon/AWS, NVIDIA, Moore's Law. Sees structural similarities before others.

Pivoted multiple times: chatbots → data labeling → self-driving → LLM data → enterprise AI → defense → Meta. Each pivot early and correct.

Equally fluent on RL curves and sovereign data strategy with heads of state.

Reviews every hire, hand-reviewed partner data.

“Quality is fractal... high standards trickle down.”

Language models with OpenAI in 2019, defense AI in 2020, enterprise AI before the wave. Recognized AI at age 16.

Frameworks show interconnected thinking: Virtuous Flywheel, Stacking Waves, Four Building Blocks.

Admits first 10 ideas were bad, agents were overhyped, young founders lack “sense of alpha.”

Frameworks elegant but lack implementation detail. “Think about your data reserves” is directionally correct but not actionable.

Mentions risks briefly then pivots to opportunity. Doesn't engage with displacement or concentration of power.

Post-Meta, noticeably more polished. Hard to distinguish personal conviction from corporate positioning.

Positions Meta/WhatsApp as trust exemplar without acknowledging documented trust challenges.

Frequent “sort of,” “in many ways,” “I think.”

“When you're young you're very impatient... that's both a great strength and a great weakness.”

“Personal AI that knows you” is now pitched by every major lab.