Durability and moats in the AI era

Why companies are becoming a series of loops | Anish Acharya (a16z)

Topic 14 of 19

Published Sep 6, 2026 · Clip 54:30–59:28 (4:58)

Anish offers two ideas on durability: moats are most often discovered, not designed — Decagon's Jesse's framing — as Cursor showed by capturing reasoning traces and training its own Composer models after being criticized for having no moat; and classic moats never depended on how hard software is to build — network effects, scale advantages, brand effects, proprietary data, and cornered resources all still work for founders with ambition toward them. He'd happily bet on a product without a stated moat if it has momentum, craft, and growing engagement, because big ideas rest on a dozen invisible small ideas competitors never replicate — Granola being beloved and dominant despite a weak durability story. With near-identical underlying models, the harness and UX become most of the differentiation, which is why edge users pay for every specialized assistant. He also names his shortlist of real business books: Seven Powers, High Output Management, and The Hard Thing About Hard Things.