The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch 20VC: Horizontal vs Vertical Agents: Who Wins | Model Routing: Commodity or Increasing in Value | The Future of the Inference Market | Why Founders Should Take Discounts for Tier 1 Firms with Harrison Chase, LangChain
518 snips
Oct 18, 2026 
Harrison Chase, co-founder and CEO of LangChain, created the open-source framework powering modern AI agents. He explores why agents remain difficult to build, whether horizontal or vertical systems will prevail, and the future of model routing, memory, inference, and infrastructure. He also discusses exploding token costs, Chinese AI models, and why top-tier investors may justify a founder’s discount.
AI Snips
Chapters
Transcript
Episode notes
Start Vertical Agents With Verification
- Start vertical agents with context engineering and harness design before investing in model post-training.
- Domain experts can create defensible products by defining verifiers that quantify what good performance means.
Why Vertical Agents Need Harness-Level Routing
- Model routing gains value when it understands the harness, task state, prompt cache, and domain-specific prompting.
- Gateway routing remains useful for general chat or coding, but vertical agents need routing inside the harness.
Internal Evals Beat Benchmark Scores
- Public benchmarks become less informative as models optimize against them, while internal domain evals retain practical value.
- Teams should inspect traces alongside scores because a single metric can miss meaningful behavioral changes.