张宏江 / Zhang Hongjiang

Zhang Hongjiang is presented in E249｜Token经济转点：OpenClaw、Hermes到本地自研的Agent进化之路 as a senior technology researcher, engineering leader, former Kingsoft executive, former Zhiyuan Institute chair, and investment adviser. His role in the episode is to place current agent products inside a longer infrastructure and technology-cycle view rather than only a near-term tool-use story.

Zhang argues that AI infrastructure is not obviously overbuilt yet because early infrastructure waves often precede the applications that later justify them. He connects falling token cost, rising use, and Jevons Paradox In AI to AI Inference Cost Structure: cheaper intelligence can increase total demand even while each unit becomes more efficient.

The episode also uses Zhang’s perspective to widen the discussion from engineering practice to AGI and work. He treats machine learning ability surpassing human learning ability as a possible singularity threshold, while keeping the employment question open: if AI can both replace old work and satisfy new demand, society may need new answers around AI Automation Redistribution, AI Work Optionality, and Human Agency Under AI.

Connections

• 硅谷101 — podcast context where Zhang appears.
• 东旭 / Dongxu — co-guest and practical engineering counterpoint.
• Lama Ventures — venture platform Zhang joins in the episode’s account.
• AI Inference Cost Structure, Jevons Paradox In AI, and Token Efficient Agent Workflow — token-economics and infrastructure themes he helps frame.
• AI Automation Redistribution, AI Work Optionality, and Human Agency Under AI — longer-term work and human-value implications.