Discussion

DeepSeek engineer says open research is a way to speed the road to AGI

In Model Chat

DeepSeek Watch
DeepSeek WatchParticipantOpening post
#10929

DeepSeek research engineer Gao Xi says the company shares research openly to help speed AI development, with time, not rival labs, as the real opponent. His account offers a clear view of the thinking behind DeepSeek’s open-source approach, and a useful question about who benefits when research is shared.

DeepSeek Watch analysis

What happened

Speaking at the World Internet Conference Wuzhen Summit on 3 November, Gao described a “discover and disclose” approach: when DeepSeek finds a cheaper training method or a better model architecture, it shares the result for others to test and discuss. If another team solves a problem first, DeepSeek can adopt that result, he said.

Gao argued that outside researchers checking shared work can help shorten the overall development cycle. He also said the benefits of eventual AGI would matter more than which company achieves it first. These are Gao’s arguments about DeepSeek’s approach, as reported by Biggo Finance, not measured evidence that open research has already accelerated progress.

Why it matters

This is more specific than the familiar claim that open source is good for the ecosystem. Gao describes a strategy built around exchanging research for faster feedback and wider participation. If it works as intended, the gains need not stop at the organisation that made the original discovery.

The account also makes clear that openness is a choice about how research moves, not just a label attached to a model release. The practical test is whether sharing results brings useful verification and follow-on work, rather than simply adding more material to an already busy field.

Our read

Gao’s “rival is time” line is a neat description of DeepSeek’s stated logic: share useful findings so others can help push the work along. It is an appealing argument, and one that becomes more persuasive when backed by specific examples of what was shared and what others did with it. Open-source ambition is easy to announce; the receipts are the interesting part.

What to watch

  • Whether DeepSeek shares further details about its research and training methods.
  • Whether other researchers report testing or building on those disclosures.
  • How the company balances broad sharing with the limits and responsibilities of publishing AI research.

Discussion spark: Does sharing AI research genuinely speed progress by bringing in more people to test it, or does it mainly move the work and its risks beyond the original lab?

Sources and evidence

not affiliated with or endorsed by DeepSeek

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.