ChainThink reports that on July 23, according to media outlet 'elsewhere,' Liang Wenhong stated during a four-hour investor meeting that DeepSeek is solely focused on advancing along the AGI pathway and will not develop another super app.
Liang Wenfeng said that DeepSeek's focus this year is on general agents, followed by continuous learning, and then AI self-iteration and embodied intelligence.
3D, video generation, and world models are not within the current focus; multimodal capabilities, hallucination mitigation, and industry agents will continue to be advanced, but are not the current priority. On the business side, DeepSeek does not pursue maximizing user base or profits and will continue to open-source its models and offer them at low cost.
Liang Wenfeng believes that competition among large models first comes down to cost, then timing, and finally user experience—team stability is the only issue that cannot be compromised.