Zhipu founder Tang Jie says company will prioritize AGI research over near-term monetization

An internal letter on July 11 said Zhipu’s new Touch High plan will prioritize AGI research for the next two years, focusing on autonomous agents, synthetic-data self-training and safety governance over short-term revenue.

46d ago

Fact Check

Multiple independent sources (BlockBeats flashes 355606 and 355628, The Standard Hong Kong, and Wall Street CN) all confirm the July 11 internal letter from Zhipu founder Tang Jie launching the 'Touch High' (摸高) plan that prioritizes AGI research over short-term monetization for two years. The four focus areas cited in the claim—autonomous agents, synthetic-data self-training/self-play, and safety governance (plus long-horizon complex tasks)—match the sources precisely. The only minor discrepancy is that some English sources describe Tang Jie as 'co-founder' rather than 'founder,' but this does not affect the substance of the claim.

Summary

Zhipu founder Tang Jie said in an internal letter released on July 11 that the company is launching a “Touch High” plan and will prioritize AGI research over near-term application monetization for the next two years. Tang described the approach as a “counterintuitive” path in the next phase of AGI competition, emphasizing long-horizon complex tasks, large-scale autonomous agent collaboration, fully self-training with synthetic data and self-play, and stronger safety governance. He also said the effort will include billion-yuan-level investment in mechanistic interpretability research, underscoring Zhipu’s focus on foundational capability development and model safety rather than short-term commercial returns.

Terms & Concepts

AGI: Artificial general intelligence, AI with broad human-like capabilities
autonomous agent: AI system that can act independently on tasks
mechanistic interpretability: Research into how AI models internally produce outputs