Tang Jie's Internal Letter: Zhipu to Go All-In on AGI for Two Years
新智元 · wechat · 2026-07-12
In an internal letter, Zhipu founder Tang Jie stated that the company will focus heavily on the AGI/ASI route over the next two years, breaking down the goal into several engineering pillars: long-horizon tasks, long-term memory, continuous learning, self-judgment, agentic society, synthetic data factories, and self-training.\n\nKey highlights from the letter include:\n- A commitment to relentlessly pursuing "super artificial intelligence" over the next 2 years.\n- Envisioning an agentic society comprised of thousands of digital employees collaborating 24/7.\n- Emphasizing that "AI training AI" is already taking shape, and recursive self-improvement will drive faster iterations.\n- Defining AGI as "the sum total of human intelligence," believing the ultimate goal is to generate original knowledge on the level of the "Theory of Relativity."\n\nThe article also notes Zhipu's dual approach to safety research and open-source products: strengthening interpretability and safety research on one hand, while fully open-sourcing GLM-5.2 under the MIT license with 1M context support on the other. Overall, this serves as a very clear strategic signal for the company.
Related event: Zhipu's Tang Jie Internal Memo Leaks: Going All-in on AGI(8 posts)→
More from AGI Musings
- Bindu Reddy says the industry still lacks a way to train 20T models and scale post-training RL — bindureddy · 2026-07-22
- Advanced AI Models Are Becoming Impossible to Plug and Play — emollick · 2026-07-22
- The Thimble and the Waterfall: AI's Data Bottleneck and Feedback Loops — dyamins · 2026-07-22
- Researcher Admits Kurzweil Was Right About AI Scaling Laws All Along — davidmanheim · 2026-07-22
- AI is still not at a maturity plateau, the author argues — generativist · 2026-07-22
- Essay argues LLMs are externalized metacognition, not standalone intelligence — lnsip9reg · 2026-07-22