DeepSeek puts AGI research ahead of products and commercial growth

DeepSeek, the Chinese AI developer behind the open-source R1 reasoning model, is prioritizing artificial general intelligence (AGI) research over building a consumer platform or maximizing near-term revenue, according to a report by IT Home based on a circulated transcript of a four-hour investor meeting involving founder Liang Wenfeng. Liang reportedly said DeepSeek is not focused on competing for consumer traffic, enterprise revenue, or becoming the next ByteDance or Tencent, and sees continual learning as the next major technical challenge after AI agents.

According to a report by IT Home https://www.ithome.com/0/980/367.htm , DeepSeek, the Chinese AI developer behind the open-source R1 reasoning model, is prioritizing artificial general intelligence AGI research over building a consumer platform or maximizing near-term revenue. The report is based on a circulated transcript of a four-hour investor meeting involving founder Liang Wenfeng. Why it matters: The reported strategy presents DeepSeek as a research-led AI developer whose products and commercial services are secondary outcomes of its AGI work. - DeepSeek’s open-source reasoning models have drawn global attention by combining strong performance with relatively low-cost development. - Liang reportedly said DeepSeek is not focused on competing for consumer traffic, enterprise revenue, or becoming the next ByteDance or Tencent. - The approach helps explain the company’s emphasis on open-source models, lower API prices, and improvements in training and inference efficiency. - It also distinguishes DeepSeek from AI companies that are building standalone products and enterprise businesses around their models. Details: Liang reportedly sees continual learning as the next major technical challenge after AI agents. - He described a development path from chain-of-thought reasoning to AI agents, continual learning, and eventually embodied intelligence. - Current AI systems generally need users to provide the relevant context again for each task. Liang reportedly believes future models should be able to learn continuously from experience. - Coding agents are a priority because they could improve both DeepSeek’s products and its own research process. - DeepSeek is not prioritizing 3D generation, video generation, or world models, which Liang reportedly views as useful applications but not central to the development of intelligence. - Multimodal capabilities will still be developed for consumer and enterprise products. - Liang reportedly said DeepSeek’s API pricing is designed to recover hardware costs in around 10 months, rather than maximize profit. - DeepSeek also reportedly plans to keep its strongest models open source and use the same models internally and externally, rather than release weaker versions to the public. Context: DeepSeek’s public profile has been shaped by its open-source and low-cost approach, but the reported comments suggest that the company still sees itself primarily as an AI research organization. - Consumer products, enterprise services, and API revenue are reportedly intended to support the company’s longer-term technical goals. - The comments offer a view into how DeepSeek may be positioning itself as the AI industry shifts from model releases toward agents and systems that can assist with AI research.