Zhipu AI founder outlines 'Touch High' plan for AGI research, defying market slump
The AMW Read
The 'Touch High' plan meaningfully updates Zhipu AI's strategic posture within the foundation-model segment, signaling a deliberate pivot from commercialization to AGI research, but does not resolve any open debate or introduce a new top-tier entrant.
Zhipu AI founder outlines 'Touch High' plan for AGI research, defying market slump
Zhipu AI (智谱) founder Tang Jie sent an internal letter to employees outlining a 'Touch High' plan that shifts the Chinese foundation-model developer's focus toward long-term AGI research rather than near-term commercialization, as reported by LatePost. The plan identifies four technical priorities: long-horizon tasks, autonomous agent systems, fully self-training, and extreme safety governance. Tang also highlighted mechanistic interpretability — studying the internal logic behind model decisions — as a major safety research direction. Zhipu AI is the company behind the GLM series of large language models.
Why it matters: This move positions Zhipu AI in direct opposition to the prevailing capital-compression arc in China's foundation-model space, where most peers are racing to demonstrate revenue traction amid a funding winter. By publicly committing to AGI research and safety interpretability at a time when the market is pressuring for unit economics, Zhipu AI is effectively doubling down on the 'research lab' identity — a structural posture that echoes OpenAI's pre-commercial era but carries higher risk in a capital-constrained environment. The emphasis on autonomous agents and self-training signals that Zhipu AI sees agentic workflows, not chat interfaces, as the next frontier for foundation-model differentiation.
Grounded expert take: This is a deliberate signal to both the talent market and the state-backed capital pool that Zhipu AI intends to compete on frontier capability, not on price or enterprise sales velocity. It follows Zhipu AI's recent aggressive price cuts — the company announced its second price reduction in a month — which Tang described as 'not a simple price war,' suggesting the 'Touch High' plan is meant to reconcile short-term market-share moves with long-term ambition. The open question is whether the Chinese capital environment, which increasingly favors sovereign-AI champions with clear revenue trajectories, will continue funding a lab that explicitly deprioritizes commercialization.
