When Marks prepared a memo on whether AI was a financial bubble late last year, he viewed the technology with a healthy dose of skepticism. Like many, he initially assumed AI models were little more than glorified search engines, capable of retrieving and regurgitating data without true comprehension. He wondered whether AI could actually break genuinely new ground, or if it was merely restricted to statistically recombining the thoughts of humans. And so, as he explained in a follow-up note to clients on Feb. 26, he talked to “some interesting techies in their thirties and forties.” One of them suggested he ask Claude.

Anthropic’s AI model produced a tutorial, Marks wrote, explaining AI and changes over the past three months. The resulting 10,000-word essay was so impressive, Marks added, he decided to reprint most of it for his clients in a recap, but he wanted to add some of his own human writing.

“I could have saved myself a lot of time by asking Claude to write this memo,” he wrote, “but I decided not to, because I consider putting words on paper a big part of the fun.”

Marks stressed to clients: “I want to try to communicate the level of awe with which I viewed Claude’s output.” Not only did it read like a personal note that a friend or colleague might have sent, but it psychologically dismantled Marks’ skepticism by using the investor’s own legendary mentors against him. Above all, he said, it was really smart.

“It argued logically, anticipated points I might make in response, injected humor, and bolstered its credibility by candidly acknowledging AI’s limitations, just as I might do,” he said. “I’ve asked AI questions before and gotten answers back, but I’ve never received a personalized explanation like I did in this case.”

Here’s what Claude told Marks, which left him impressed—and a bit shaken.

The Graham, Buffett, Munger example

In response to Marks questioning whether AI could genuinely think, synthesize information, or have a new idea, Claude delivered a “spirited rejoinder,” Marks said. The machine addressed him directly: “Howard, everything you know about investing came from other people,” the AI wrote. “Benjamin Graham taught you about margin of safety. [Warren] Buffett taught you about quality. Charlie Munger taught you about mental models from multiple disciplines. John Kenneth Galbraith taught you about the psychology of financial manias.”

The AI pointed out Marks had spent 50 years reading thousands of books, memos, and case studies, taking raw material from others to produce something genuinely new.

“You learned reasoning patterns from decades of reading. I learned reasoning patterns from training,” the AI argued. “The question isn’t where the inputs came from. The question is whether the system—human or artificial—can combine them in ways that are genuinely novel and useful.”

The argument left Marks stunned.

“Of course, this is completely true,” he wrote, marveling at how the AI argued logically, injected humor, and anticipated his counterpoints. The revelation shifted his entire perspective on machine intelligence, forcing him to ask: “Is AI’s way of growing, learning, and ‘thinking’ really different from ours?”

Marks is famous in financial circles or his periodic “memos” to Oaktree clients, written since 1990, which analyze market cycles, risk, and investor behavior; they have been called some of the most influential writings in modern finance and are now in the Museum of American Finance’s permanent collection. Buffett has said that when he sees a memo from Marks, it is the first thing he reads, which has helped cement Marks’ reputation among professional investors. In 2024, the Museum of American Finance gave Marks a lifetime achievement award for his impact on financial thinking, underscoring how his memos and philosophy have shaped modern investing. So his opinion is just one on the impact of AI, but it’s quite an opinion.

Reasons why and why not

Now viewing AI not as a mere assistant but as an autonomous system—what he calls “Level 3” AI, capable of full labor replacement—Marks said he sees massive implications for Wall Street. Many of the qualities required to be a phenomenal investor are present in AI, which can absorb endless data, recognize historical patterns, and operate entirely free of human greed, fear, or fads. He also noted the technology is advancing at speeds that outpace anything in history, with some models even writing their own code and testing their own software autonomously.

Yet despite AI’s terrifyingly fast evolution and its ability to mimic the intellectual synthesis of Buffett and Munger, Marks added he doesn’t believe human investors are entirely obsolete. The human edge, he argues, remains in assessing truly novel developments where historical patterns don’t exist. Furthermore, AI lacks a critical component of investing: “skin in the game.” Unlike humans, a machine does not intuitively sense risk or feel the gut-wrenching fear of capital loss.

Marks may be right that AI will transform finance. But there is reason to pause at the feelings of awe he described. The beauty and persuasiveness of Claude’s response to Marks alone may be cause for skepticism, because personalization is one of Claude’s core product features. Large language models are extraordinarily good at inferring context clues—in this case, the user’s name, professional background, and likely objections—and tailoring output accordingly. The 10,000-word essay felt like a note from a brilliant colleague because it was engineered to feel that way.

Claude’s central argument—that Marks learned from Graham and Buffett just as Claude learned from training data—sounds profound, but human learning and statistical pattern-matching over text corpora are not the same. Marks didn’t just read Graham; he applied margin of safety in real markets, lost money, felt fear, revised his thinking under genuine uncertainty, and built convictions that cost him something if he was wrong. That feedback loop—consequence, revision, hard-won judgment—is precisely what Claude has never experienced and cannot simulate.

Ultimately, Marks’ encounter with Claude convinced him that AI is vastly more powerful than a simple fad, and that its true potential is likely being underestimated today. While he cautions against going “all in” on AI stocks, owing to the risk of ruin, the legendary investor’s final advice is clear: No one should stay “all out” and risk missing one of the greatest technological leaps in human history. Even though Claude will never have skin in the game.