In a 26-minute interview, Jensen Huang did not mention GPUs or computing power, instead focusing on the shift from writing code to building Agents. He spoke with LangChain founder Harrison Chase, emphasizing that engineers should write less repetitive code and build more functional Agents. Huang highlighted "Harness," which encompasses prompts, tool specifications, memory systems, context management, task decomposition, retry logic, evaluation pipelines, and access control. LangChain optimized its workflow using Nemotron 3 Ultra paired with Harness, pushing its Deep Agents benchmark score to 0.86 — just 0.01 behind the top result — while slashing the per-run cost from $43.48 to $4.48. The model itself stayed the same; only system prompts, tool descriptions, and middleware layers were adjusted. The benchmark score climbed to 0.86 from its baseline, while the top closed-source model sits at 0.87, with a gap of only 0.01. The per-run cost dropped from $43.48 to $4.48 — nearly a 10× reduction. Huang compared coding to typing, stating that generating code only solves the "write it down" part, while Agent engineering must ensure the system runs correctly, behaves reliably, recovers from errors, and leaves a traceable audit trail.