Baidu CEO Robin Li declared an industry-wide pivot from foundation model competition to AI agent dep...
The AMW Read
Baidu is a canonical case-study player in Segment 01; the explicit pivot from models to agents meaningfully updates its strategic trajectory and the CN AI landscape, but the agent thesis is already active in the substrate. Cross-substrate capital or compute conditions are not the headline of this ar
Baidu CEO Robin Li declared an industry-wide pivot from foundation model competition to AI agent deployment at the Create 2026 developer conference in Beijing, themed Agents at Scale. Li argued that breakout AI products this year succeeded not due to underlying model prowess but because they function as agent systems capable of continuous online operation and task execution. Baidu introduced the Daily Active Agents (DAA) metric as a successor to DAU, predicting global DAA could exceed 10 billion. The company also unveiled enterprise agent products and positioned AI as evolving from chatbots into digital workers, task agents, and autonomous collaborative systems.
Why it matters: This signals that Baidu, a top-tier Chinese AI lab, is formally endorsing the thesis that the AI industry's value has shifted from model training to agentic application layers — a structural force already visible in the West with products like ChatGPT's agent features and enterprise copilots. By introducing DAA as a successor to DAU, Li is proposing a new measurement framework that frames agent task completion, not user engagement, as the monetizable unit. This echoes the context-engineering moat pattern in which model commoditization drives differentiation toward execution reliability and closed-loop verification.
Grounded expert take: Li's framing of "self-evolution" across three levels — agent, individual, and enterprise — directly updates the open debate about whether AI's primary economic impact will come from labor substitution or labor augmentation. His "super individual" thesis (one person plus an agent fleet as the smallest productive unit) leans heavily toward augmentation and flattens organizational structures. The risk, as with previous platform shifts, is that enterprises underestimate the data-integration and workflow-reengineering costs required for agents to operate in verifiable closed loops, a point Li himself acknowledged. Baidu's ability to execute on this vision depends on its existing search ecosystem as a distribution layer — a hyperscaler-distribution advantage that few Chinese competitors can match.
