Howard Marks is one of the sharpest financial minds alive. The man has been thinking clearly about markets for fifty years, has written memos that get passed around Wall Street like sacred texts, and has outlasted more market cycles than most of us have had hot dinners. So when Howard Marks decides he needs to get educated about artificial intelligence to write a follow-up to his December memo, he does what any serious intellectual would do: he asks Claude.
And then Claude — the AI — teaches him about Claude.
I’ve been sitting with this for a few days and I’m still not entirely sure whether it’s profound or just very, very funny. Maybe both. Probably both.
The Velocity of Autonomy
Marks’ recent memo, AI Hurtles Ahead, is ostensibly about the breakneck pace of AI development. And it’s genuinely good on that front. He traces the three levels of AI capability with characteristic clarity:
- Level 1 (Chat AI): Systems that simply answer questions and save thinking time.
- Level 2 (Tool-using AI): Systems that can search out information and perform tasks when instructed.
- Level 3 (Autonomous agents): Systems where the user simply gives a goal, and the agent does the work, checks it, and submits a finished product.
He notes, soberly, that we moved from Level 1 in 2023, to Level 2 in 2024, to Level 3 today. He is appropriately awed.
But the more interesting story isn’t the content. It’s the method.
The Ouroboros of Prompting
Marks’ advisers built a nine-module curriculum specifically tailored to him — built around his intellectual frameworks, his past memos, and his particular way of thinking. They fed this to Claude as a prompt.
Claude then produced a 10,000-word tutorial that, by Marks’ own account, “read like a personal note from a friend or colleague”. It referenced his prior work on the sea change in interest rates and the pendulum of investor psychology. Furthermore, it argued logically, anticipated points Marks might make in response, injected humor, and candidly acknowledged AI’s limitations to bolster its credibility. Marks was, in his words, filled with a “level of awe”.
Let’s just pause on what happened there. A man wanted to understand an AI. So he asked the AI to explain itself to him. The AI, knowing who it was talking to, shaped its entire explanation around that specific person’s worldview. And it worked so well that the person felt like he’d received a letter from a thoughtful friend.
This is either the most elegant tutorial design in history, or a ouroboros so perfectly formed it has achieved enlightenment.
The Cover Band vs. The Composer
What I find genuinely interesting — beyond the comic recursion of it — is the argument Claude makes when Marks presses it on the “can you actually think?” question. The skeptic’s case is reasonable: everything Claude knows came from human-written text. Therefore, it’s just sophisticated pattern-matching — a very talented cover band, not a composer.
Claude’s rebuttal, addressed directly to Marks, is brilliant:
“Howard, everything you know about investing came from other people… Benjamin Graham taught you about margin of safety. Buffett taught you about quality… You read thousands of books, memos, case studies, and annual reports over fifty years… You took frameworks from multiple disciplines, applied them to novel situations, and produced something genuinely new. The raw material came from others. The synthesis was yours.”
Marks calls this “completely true,” and he’s right. It’s a good argument. It’s also, and I mean this with respect, an argument that Claude generated from patterns it absorbed from human philosophical debate about AI. Which doesn’t make it wrong! But it does make the whole thing shimmer with a slightly dizzying self-referentiality.
The Economic Reality
The bit that will really stick with me, though, is the economic argument Claude eventually lands on. Forget whether AI is “truly” thinking, it argues. The economic question isn’t “does AI truly understand?”. The economic question is “does AI do the work?”. And if a tool can produce the analytical output of a $200,000-a-year research associate, nobody paying the bill is going to be very interested in the philosophical fine print.
Marks, ever the investor, finds this convincing. And I think he’s right to.
There’s something almost sweet about a legendary investor sitting down with an AI, being genuinely surprised by how good the conversation is, and then writing an addendum trying to make sense of it for his clients. He admits he could have saved a lot of time by asking Claude to write the memo. He chose not to, because “putting words on paper a big part of the fun”.
Which, honestly, same.
The optimists say new jobs will emerge, as they always have after every technological disruption. Marks isn’t sure he believes them, but he admits he can’t prove them wrong. His closing line — borrowed from a friend — is that he’d rather be an optimist and wrong than a pessimist and right. He’s not sure he can pull it off.
Me neither, Howard. But I appreciate the company.