The founder of Zhipu, an AI company, has revealed an internal strategy dubbed 'Touch High,' which aims to position the company at the forefront of Artificial General Intelligence (AGI). In a recent communication, the founder emphasized that the industry is rapidly moving toward commercial applications, and Zhipu is determined to break through to new heights in AI development.

According to the founder, the company will focus its efforts on four core engines over the next two years, rather than seeking immediate financial returns. The first engine is 'long-range tasks,' which aims to transition AI from simple question-and-answer capabilities to undertaking complex projects. The second engine is the 'autonomous intelligent system,' which envisions a society of digital employees with diverse skills and personalities capable of self-debate, collaboration, and resource management, achieving a level of productivity akin to autonomous driving.

The third focus is on 'complete self-training,' where computational power is transformed into evolutionary fuel. This involves building high-quality synthetic data factories and enabling AI systems to reconstruct their own code through self-play mechanisms. The fourth engine, which the founder highlighted as particularly crucial, is 'extreme safety governance.' As AI capabilities grow, the safety constraints must also be robust to prevent misuse.

Recently, Zhipu has also introduced its most powerful open-source model to date, GLM-5.2, which supports a million-context capability. This model continues to lead in long-range tasks and is open to all users under a permissive MIT license, allowing for download, deployment, and commercial use without restrictions.