It’s not every day that an American company — valued at $60 billion no less — says it was out-innovated by a Chinese startup, but DeepSeek seems to have managed to do just that.
In a candid conversation, Mike Krieger, Chief Product Officer at Anthropic and co-founder of Instagram, admitted that DeepSeek’s rapid rise and innovative product features, particularly around “Chain of Thought” reasoning and a speedy product launch, had caught his attention. Krieger’s remarks offer a rare glimpse into the competitive landscape of AI, highlighting the pressure even well-funded giants face to keep pace with nimble newcomers.
“It’s interesting; our plan was already to show Chain of Thought when we launched our reasoning model,” Krieger said on a podcast. “So that was not a reconsideration, but it was interesting to see somebody else do it.” He went on to praise certain user interface details in DeepSeek’s implementation.
DeepSeek had been the first major model to show a model’s chain-of-thought reasoning to users. This feature, which allowed users a glimpse into how the model was thinking about the problem, had been one of the major reasons why DeepSeek had gone viral.
“The other piece,” Krieger continued, “from a product perspective, there were two things. I think that’s the under-talked-about piece of DeepSeek: They were able to go from nobody knowing about them to them being, frankly, in many circles, better known than Claude.” He recounted a surprising anecdote: “My great aunt was calling me about DeepSeek. I’m not even joking. It was a cliché actually happening. ‘What do you think about DeepSeek?’ she said. (That’s when you know) it’s broken through.”
Krieger further emphasized the speed of DeepSeek’s product development: “They went from not having a product to having an iOS app that actually had a lot of good details. For me, it was a good lesson.” He admitted, “We need to be getting some of these ideas out to market quicker. We should be willing to put it out there and learn, because sometimes the novelty of the experience is itself valuable. (With DeepSeek) it was the first time most people experienced this sort of live Chain of Thought. That’s interesting, and I wish we had done that sooner because it would have been novel for people to experience that.”
Anthropic has been keeping a close eye on DeepSeek since it was released. In its early days, Anthropic CEO Dario Amodei had said that while DeepSeek was being marketed as cheap, they did have 50,000 GPUs. He had also said that DeepSeek had no safety blocks against generating harmful information.
Anthropic CPO Mike Krieger’s latest observations on DeepSeek reveal a key dynamic in the AI race: the tension between perfecting a product internally and rapidly iterating in public. While Anthropic prioritized a more polished and safe initial release with Claude, DeepSeek’s swift deployment of Chain of Thought, coupled with a functional iOS app, allowed them to capture user attention and generate valuable feedback. This approach, Krieger acknowledges, offers inherent advantages, particularly in a nascent field where user expectations are still forming. The speed of DeepSeek’s ascent clearly demonstrates the power of early adoption and rapid iteration in capturing mindshare, even against established, well-funded players. It’s a clear signal that in the rapidly evolving world of AI, speed and agility can be as valuable as resources and reputation. The willingness to sacrifice some polish for the sake of early user engagement and feedback may be a crucial factor in determining the ultimate winners in this competitive landscape.