Liang Wenfeng Outlines DeepSeek's AGI Focus and Tech Roadmap

During a 4-hour meeting with investors, Liang Wenfeng comprehensively outlined DeepSeek's strategic roadmap and core philosophy. He explicitly stated that DeepSeek's sole focus is AGI; it is currently not pursuing user scale, rushing commercialization, or building the next super app. 3D, video generation, and world models are not current priorities. While multimodality is important for the C-end, it remains a secondary track, with Coding Agent currently being the top priority.

Confirmed

**Strategy and Commercialization**: Liang pointed out that while DeepSeek will commercialize, it does not seek profit maximization. He emphasized that low pricing is not a subsidy strategy, but the result of improvements in model architecture and computational efficiency. Under equal compute power, lower costs better support the training of larger models.

**Team and Culture**: Liang dismissed the external narrative that DeepSeek relies on a "genius myth," describing the team as ordinary people doing extraordinary things. He believes that as long as the team remains stable, DeepSeek has a chance to achieve AGI. Maintaining open source, keeping prices low, and not chasing short-term profits are fundamentally aimed at increasing the probability of achieving AGI.

**Technological Roadmap**: Liang provided a clear AI development path: from CoT to Agent, then to continuous learning, and ultimately to AI self-iteration and embodied intelligence. He believes that what current AI lacks most is not taste or intuition, but the ability to learn continuously.

Why it matters

Against the backdrop of an LLM industry widely chasing commercial monetization and C-end traffic expansion, DeepSeek has demonstrated a starkly different strategy of "restraint." Clearly concentrating resources on foundational model capabilities and the AGI goal not only provides the underlying logic for understanding its open-source and low-price behaviors, but also serves as a crucial industry bellwether for observing the evolution of AI technology toward Agents and continuous learning.

2026-07-26 ~ 2026-07-26 · 6 related posts

• [source] DeepSeek’s Wenfeng Liang says restraint, open source, and low cost are the path to AGI — FinanceYF5 · 2026-07-26
• [source] DeepSeek says AGI is the only mainline, and coding agents are the priority right now — FinanceYF5 · 2026-07-26
• Liang maps AI’s next steps from CoT to agents, continuous learning, and embodied intelligence — FinanceYF5 · 2026-07-26
• [source] DeepSeek says it will monetize, but cost efficiency matters more than profit maximization — FinanceYF5 · 2026-07-26
• Liang says DeepSeek is built by ordinary people focused on one extraordinary goal: AGI — FinanceYF5 · 2026-07-26

1 near-duplicate retellings: FinanceYF5