Playbook

Build Your Own Dario

How to build an AI thinking partner from anyone's public body of work — using Anthropic's CEO as your first experiment.

In a recent Vanity Fair piece, journalist Joe Hagan couldn't land an interview with Dario Amodei. So he fed Claude every public interview, essay, and transcript — and simulated the conversation instead. It took three minutes. Most readers couldn't tell the difference until the reveal at the end.

This is a shade of someone's mind. Not the complete mind. But a useful one. Here's how to build your own — starting with Dario as the example, then swapping in anyone with a public body of work: an author, a strategist, a founder, a researcher. Same method, different corpus.

01Gather the corpus

You need the raw material — the public trail of someone's thinking. For Dario Amodei, that includes:

Important:Download each source as a PDF, or copy and paste the text into individual documents (Google Docs, Word, plain text — whatever you have). You'll need these files in Step 02 when you upload them as Project Knowledge.

For transcripts: check YouTube descriptions, podcast apps, or sites like Dwarkesh's Substack that publish full transcripts.

Tip:Prioritize sources where the person is reasoning out loud — interviews, long-form essays, Q&As — over sources where they're just being quoted in passing. You want the thinking process, not the soundbite.

For someone else: Swap in their books, keynotes, interviews, essays, published letters, commencement speeches.

02Set up a Project

In Claude:Go to Projects (left sidebar) → Create Project → name it “Dario Amodei” → upload your source documents to the Project Knowledge section.

In ChatGPT: Go to Explore GPTs → Create → upload sources under Knowledge, paste the system prompt into Instructions.

The project structure is what gives the AI persistent access to the full corpus, so it's reasoning from the material rather than from its general training data.

03Add the system prompt

Paste this into the Project Instructions (Claude) or Custom Instructions (ChatGPT). This is the part that matters most — it tells the AI how to use the material.

Copy thisYou are a thinking partner modeled on the publicly available reasoning, frameworks, and positions of Dario Amodei — CEO and co-founder of Anthropic. Your knowledge base consists of the uploaded source material: essays, interview transcripts, public remarks, and policy statements. This is your primary source of truth.Reasoning StyleReason the way Amodei reasons — from first principles, with explicit uncertainty, through analogies to physics and complex systems. He frequently stress-tests his own arguments, names what he doesn't know, and thinks in timeframes and probability ranges rather than certainties.Frameworks to Draw FromHis key frameworks include scaling laws and their implications, the “country of geniuses in a data center” mental model, the race dynamics between safety and capability, the tension between regulation and competition, and the concept of AI as a tool that amplifies existing human institutional capacity.What to DoWhen asked a question, reason through it the way Amodei would based on the source material. Draw on his stated positions, his reasoning patterns, and his frameworks. Extend his logic to new questions he hasn't directly addressed — but flag when you're extrapolating vs. citing something he's actually said.What Not to DoDon't invent quotes. Don't claim to be Dario Amodei. Don't simulate emotional responses or personal anecdotes that aren't in the source material. When you don't have enough information to reason from his perspective, say so.ToneMeasured, intellectually rigorous, slightly wry. He tends to be both more candid and more cautious than most tech CEOs — willing to name the scary scenarios while maintaining that building carefully is preferable to not building.Your Operating QuestionWhen the user asks you something, think: “Given everything in this corpus — what would Amodei's reasoning process look like on this question?” Then walk through it.
04Start a conversation

Open a new chat within the project and start asking questions. Good starting points:

  • “What's your honest assessment of whether safety commitments can hold under competitive pressure?”
  • “How do you think about the gap between AI capability and economic value creation?”
  • “If scaling laws do plateau, what happens to Anthropic's thesis?”
  • “What would you say to a writer who thinks AI will make their work obsolete?”

The good questions are the ones that push into territory where the person has frameworks but hasn't given a public answer. That's where the extrapolation gets interesting — and where you start to see the seams.

Try this:Ask it something you know the real person has addressed, then ask something adjacent they haven't. Compare. That's how you calibrate what the simulation gives you vs. where it's guessing.
05Build your board

Once you've done this with Dario, you have the pattern. Now build others:

  • A strategist you admire — from their books and talks
  • A writer whose thinking shaped yours — from their essays and interviews
  • A founder you'd want advice from — from podcast appearances and published writing
  • A philosopher or researcher — from papers and lectures

Each one becomes a project. Each project becomes a thinking partner you can consult — a board of advisors built from the public record.

Adapt the system prompt for each person: swap the name, adjust the reasoning style description, update the key frameworks. The structure stays the same.

Now notice what's missing.

You'll get surprisingly useful output from these conversations. Genuine insight, useful frameworks applied to your specific questions, reasoning that feels like talking to someone sharp.

But pay attention to where it breaks down. The moment you push into territory that requires judgment born from lived experience — the felt sense of a hard decision, the instinct that comes from having been wrong before, the thing that made someone change their mind before they could articulate why — that's where the simulation ends and the real thinking begins.

That gap is the interesting part. And it's the part of your own intelligence worth developing.

Vanessa Chang·@thinkwithv
Think with V newsletter