How To Build A Company With AI From The Ground Up

YC Partner Diana Hu argues AI-native startups should eliminate middle management and run the entire company as a closed-loop intelligence system.

• Teams using fully queryable orgs with AI agents cut sprint time in half and got ~10x more done.
• Some companies now have repos with zero handwritten code — only specs and test harnesses that agents implement.
• Strong DM’s AI team built a software factory where agents write and iterate code until hitting a probabilistic satisfaction threshold.
• Jack Dorsey’s view: keeping the same org chart means missing the shift entirely — the company itself must be rebuilt as an intelligence layer.
• Dorsey’s three archetypes: IC builder-operator, DRI owning one outcome, and AI-founder type who leads by example — no classic managers.
• Token usage, not headcount, is the critical metric; a high API bill is cheaper than the equivalent engineering team.
• Early-stage founders have a structural edge: no legacy systems, no retraining costs, can design for AI from day one.

2026-04-24 · Watch on YouTube