NVIDIA's Jensen Huang Defines New AI Computing Paradigm: Agent = Large Model + Orchestration Engine

According to Beating Monitor, Huang Renxun systematically outlined the architecture of "AI Agents" in his GTC Taipei 2026 keynote: large language models handle thinking, reasoning, and planning, while an external orchestration engine (harness) acts like an operating system, connecting the model to tools such as spreadsheets, browsers, and databases, and managing both short-term and long-term memory. He defined this as a new computing paradigm—where the past involved “launching applications and clicking inputs,” and the future will involve “describing intent to AI, which then generates code, invokes tools, and delivers results.” Huang demonstrated multiple Agent examples live: generating complete application code from natural language prompts; producing a dynamic particle animation from a description that transitions from Taipei 101 to GTC Taipei 2026 to the NVIDIA logo; and even capturing a photo of a lost remote control’s battery compartment, after which the Agent directly invoked CAD tools to generate a 3D-printable replacement part file. He noted, “We’re using Claude Code here, but Codex performs equally well.” In response to claims that “Agents will cause software companies to go out of business,” Huang again refuted: “Quite the opposite.” He argued that as the world is no longer limited by human numbers, vast numbers of Agents will use software tools far more extensively than humans ever could—“Now is the best time ever to be a software company”—provided that software is designed to be callable by Agents. He also announced that NVIDIA’s CUDA X libraries are now fully accessible to Agents, with Agents utilizing these libraries more efficiently than human developers.