Applied AI, not models, is the real moat as labs move toward inference
illscience · x · 2026-07-22
Joshelman argues that models are no longer the moat: the real opportunity is building products that use models more deeply in context-specific experiences.
The quoted thread adds a key strategic point: labs are more likely to vertically integrate into inference than move up into applications, because the application market is highly heterogeneous in product design, go-to-market, and price-performance needs. For applied AI founders, the test is simple: if models get twice as good, your customer should get roughly twice the value and your business capture should scale with it.
Related event: AI Consensus: Models Are No Longer the Moat, Application Layer Is(2 posts)→
More from Companies & People
- Germany logs 3,053 new startups in H1 2026 as AI takes one-third — mgill25 · 2026-07-23
- Women hold only 10–16% of core AI lab roles, says Peter Diamandis — PeterDiamandis · 2026-07-23
- Vector Institute Hiring Applied AI Roles for Deepfake Detection and Privacy — annetgriffin · 2026-07-23
- a16z picks 65 inaugural FDE fellows from thousands of applicants — a16z · 2026-07-23
- AI job finder tool says it helped land another big-company interview in Texas — seanmcdonaldxyz · 2026-07-23
- A meme turns X into an “agentic tweet company” — var_epsilon · 2026-07-23