On August 14, a confrontation on X brought the most fundamental philosophical differences in the AI industry to the forefront.

On one side is Sholto Douglas, a post-training researcher at Anthropic, a key figure responsible for the technical core of reinforcement learning. On the other side is Gavin Baker, founder of Atreides Management, managing a multi-billion dollar technology hedge fund, who has systematically articulated a framework for AI infrastructure investment this year on the All-In Podcast, being regarded as one of Wall Street's most knowledgeable investors in AI.

The catalyst was Baker's statement on the All-In Podcast. He mentioned that he heard from several trusted sources that Anthropic CEO Dario Amodei had internally stated that Anthropic might ultimately become "the only remaining private company in the world". Baker's exact words were: "If this is true, I strongly advise Dario not to say such things to anyone else."

Douglas directly countered on X: "Completely false. I appreciate Gavin's perspective, but whoever told him that is lying, intending to make you believe a narrative that certain people desperately want you to believe."

He then added, “The same group of people will tell you in one breath that Anthropic has no moat, and in the next breath, that it might be powerful enough to be dangerous. These two judgments cannot coexist.”

Baker did not push back harshly. He replied by thanking Sholto for clarifying the facts but pointed out a larger issue: "Many serious people in Silicon Valley have indeed heard some version of similar statements, and they believe this because it aligns with Dario’s public statements."

The real value of this confrontation lies not in whether Dario actually said that statement. Instead, it ignited a deeper divide concerning whether AI should be centralized or decentralized, about what security truly means—control or competition.

Baker's Framework: AI is too Dangerous to be Centralized

Baker provided the cleanest statement in the entire debate on the All-In Podcast:

"Anthropic and those in the effective altruism movement believe this technology is too dangerous to be widely distributed. Whereas Zuckerberg, Elon, and Jensen believe this technology is too dangerous to be centralized."

Baker's logical chain is as follows: Anthropic believes AI could be extremely powerful and dangerous, hence it requires cautious regulation and centralized control by a few "responsible" labs. Baker argues that this logic ultimately leads to one outcome: setting regulatory barriers that give a few companies de facto monopolies, restricting open-source efforts, and weakening competition. History repeatedly shows that absolute power concentrated in a few has never resulted in good outcomes.

He likened AI to gun rights: Americans have the right to bear arms, even though guns can be abused, because this right serves both defense and resistance against tyranny. Baker stated that he wants an AI that serves him personally, one aligned with his own values, rather than one that aligns with what Dario Amodei believes is "best for humanity."

Baker is a strong proponent of open-source models. He cited Dwarkesh Patel’s perspective, pointing out that Anthropic's Claude, built with a "constitution" and alignment system, operates under the logic that Anthropic deems correct. He does not believe the people at Anthropic are bad, but he considers it dangerous to allow any company to define what’s "best for all humanity."

He even provided an unconventional investment judgment: the rise of open-source models does not necessarily harm the value of cutting-edge labs. His analogy is the Manhattan Project. Oppenheimer and Feynman required thousands of ordinary scientists to perform the work, and the elite became more valuable due to a greater pool of "intellectual labor." Similarly, open-source models ("150 IQ") can be orchestrated and scheduled by cutting-edge models ("250 IQ"), with open-source focusing on quantity and cutting-edge focusing on value.

Douglas’s Counter: The Market is Resolving Your Concerns through Competition

Douglas’s rebuttal did not get tangled in the specific gossip but directly attacked the logical foundation of Baker's narrative.

He linked a lengthy piece by Dario Amodei and wrote: "The AI market is currently the most fiercely competitive market on Earth, with every major global company striving to provide you with smarter, cheaper models. If all goes well, we will successfully reduce the cost of everything to the cost of energy. This is incredible, but it threatens many people's old moats. They are afraid."

The implications of this statement are rich. Douglas suggests: Baker and those behind him spreading the narrative that "Anthropic wants to monopolize the world" are often precisely those stakeholders whose old moats are threatened by AI. They exaggerate Anthropic's "ambitions" and "regulatory capture," with the real aim being to promote open-source, serve their own portfolio, or purely out of fear of change.

Douglas also pointed out a core contradiction in Baker's argument: if you believe Anthropic has no moat and will be crushed by open-source, then you should not simultaneously worry about it monopolizing the world. If you believe it will become powerful enough to be dangerous, then you cannot say it lacks competitiveness. These two judgments are mutually exclusive.

In another tweet, Douglas wrote: "Indeed, if AGI is achieved, capitalism will become very strange, and the concept of 'company' might look entirely different." He linked Dwarkesh Patel's article on corporate forms in the age of AI, suggesting that this issue is far more complex than the simple dichotomy of "centralized vs. decentralized."

A Missed Core Issue

The brilliance of the debate between Baker and Douglas lies in the fact that neither of them made obvious logical errors; their disagreement stems from underlying assumptions.

Baker's underlying assumption is that AI is powerful enough to be weaponized, thus dispersing power is safer than concentrating it, just like democratic systems are safer than enlightened autocracies. He adopts a realistic view of human nature: no one, no matter how well-intentioned, should possess unchecked superintelligence.

Douglas’s underlying assumption is that the intensity of competition in the AI market itself is the best balancing mechanism. There is no need to "disperse" AI power through political narratives, as the market is already doing that. Google, Meta, xAI, OpenAI, DeepSeek, Kimi, GLM, the world’s largest companies and smartest teams are all rushing in the same direction. In this competitive landscape, no company could “become the only private company in the world.”

Both avoided a sharper question: if AI truly, as Dario said, might lead to the disappearance of 50% of white-collar jobs, then is "competition" and "decentralization" enough? Competition ensures efficiency, but it does not guarantee fairness. Decentralization ensures there are no single points of failure, but it does not guarantee that systemic risks will not arise.

Baker stated that he wants an AI "that serves him personally." But what would happen if everyone had their own AI that served them? This is no longer a technical issue, nor even a business issue; it is a political philosophy issue.

And on this question, the divergence between an Anthropic researcher and a hedge fund manager may precisely reflect humanity's most fundamental anxiety divide when facing AI: are humans more afraid of a gatekeeper that is too powerful, or are they more afraid of a world without gatekeepers?

This debate will have no winners, but it at least clarifies both fears simultaneously.