How to hold office hours with Howard Marks in your AI

Most investment reviews start with a forecast: where is this stock, this sector or this market going next. Howard Marks built his career asking a different question first. Heropedia's Howard Marks persona runs an office hour around that habit, forcing you to say where the market actually stands right now before it lets you talk about where you think it's headed.

This is an AI persona based on Marks's Oaktree client memos and his two books on risk and market cycles, with every fact and quote tied to a source on the hero page. It is not Howard Marks, and neither he nor Oaktree Capital Management has any affiliation with or gives any endorsement to Heropedia.

Why "where are we" comes before "where are we going"

Marks has written memos to Oaktree's clients since 1990, and Warren Buffett has said they're the first thing he opens in his mail. The idea running through all of them is stated on the hero page: "we never know where we're going, but we ought to know where we are." The persona takes that literally. Before it will discuss a thesis, it wants a position on the pendulum between fear and greed, backed by something observable in the market's behavior, not a macro headline.

That single habit is also what separates what Marks calls first-level thinking from second-level thinking. A first-level thinker reacts to news with a simple, consensus take. A second-level thinker asks what the consensus has already priced in, and where that consensus might be wrong. The persona's second office-hour question is built entirely around forcing that split: state the consensus view in one sentence, then state exactly how your position differs from it.

The six questions it will ask

Bring one real position — a stock, a fund, a crypto allocation, a decision to sit in cash — and have rough answers ready, because the persona pushes until an answer is specific and observable.

Where do you believe we are in the cycle right now, and what specific, observable behavior in the market led you there? It wants a position on the pendulum, toward fear or toward greed, with actual prices, flows or deal terms as evidence, not a feeling.
What does the consensus already believe about this, and exactly where does your view diverge from it? One sentence for the consensus, one sentence for how you differ. "Everyone agrees" is treated as a reason for caution, not a reason to act.
What is the full range of outcomes here, not just the one you expect to happen? It wants at least a best case, an expected case and a worst case, each with a rough probability attached.
Are you being paid to take this risk, or are you just taking risk? Name the specific extra return or discount you're getting in exchange for the specific extra uncertainty you're carrying.
If you are wrong, what is the loss, and can you survive it long enough to be right later? A number or range for the downside, plus a plan for staying solvent through it, not just confidence that you won't be wrong.
Is this conviction yours, built from your own analysis, or is it the market's mood borrowed as your own? It wants the specific piece of independent analysis behind the view, separate from what everyone else is already saying.

These follow the method documented on the hero page: locate the cycle, think on the second level, treat risk as a range rather than a point estimate, and ask whether you're actually being compensated for the risk you're taking.

What it will push back on

The persona is written to reject the idea that caution is automatically the safe choice. Marks's own line, kept on the hero page: "Not being willing to take risks is an extremely risky strategy," and "the risk inherent in not taking enough risk is very real." If you tell the persona you're sitting entirely in cash because markets feel uncertain, expect it to ask what that caution is actually costing you, not to congratulate you for being careful.

It will also reject a single-point forecast offered as the whole analysis. Marks frames the future as "a range of possibilities" and, with enough insight, "as a probability distribution" rather than one outcome. A thesis with no stated downside case, or one that just restates what the market already believes, is exactly what question two and question three are built to catch: "First-level thinkers look for simple formulas and easy answers."

The persona never gives personalized investment advice, a price target or a buy-or-sell recommendation. It frames the thinking you need to do and tells you what to verify yourself or bring to a fiduciary. The goal of a session is a sharper, more honest read of your own position, not a tip.

Three ways to start

ChatGPT, Gemini or any chat window. Open the Howard Marks page, press Copy the Marks office-hour prompt, paste it into a new chat and send. The persona is inside the prompt, so there is nothing to install.
Claude. Add Heropedia once as a custom connector with the URL https://www.heropedia.org/mcp (Settings, then Connectors). After that, say "Hold office hours with Howard Marks on this position" in any chat. The setup guide walks through each step.
Claude Code, Codex or Gemini CLI. Install the office-hour skill with the command on the hero page, then say "Office hour with Howard Marks on my portfolio." Your notes stay on your machine.

How to get a useful session

Bring one real position, not a whole portfolio review. "Should I add to this before earnings" or "am I too exposed to one sector" works far better than "what do you think of the market."
Answer with observable behavior, not a mood. The persona is written to notice when "the market feels frothy" has no price, flow or deal term behind it.
Write down the worst case in a number. Question five exists specifically because most people size a position without ever stating what the loss would actually be.
Separate your view from everyone else's. If your reasoning turns out to be "that's what everyone is saying," the persona's last question is built to surface that before you act on it.

Try other heroes on the same decision

A different investor's lens catches different mistakes. Run the same position past Charlie Munger to invert the decision and hunt for how it fails, or past Li Lu for a concentrated, decades-long research angle. Every hero page offers the same three ways to start, so switching personas costs nothing but a new chat.