The Methods
A framework for thinking and working with AI
In the last few weeks, I’ve had a breakthrough in understanding how I actually practice this work and how to pull a few different pieces together into something teachable.
Let me connect the dots.
A few weeks ago, I came across an MIT Sloan study that identified metacognition—thinking about your own thinking—as the key variable in getting real results from AI. Not prompt engineering. Not tool selection. Metacognitive strategies.
I’ve known this intuitively. But the study gave me a different lens for analyzing my own three years of work. So I exported my archive—over 9,000 messages, hundreds of conversations, daily AI collaboration documented across that entire period—and started mining it for patterns.
Not “what prompts work.” Something harder to see: What does it actually look like to work in partnership with AI? What do I do when I’m thinking well with this technology? What moves do I make? What sequences show up again and again?
When I first pulled this, I identified 29 distinct methods. That was too many. As I examined them more closely, what emerged was a combination of methods—each with their own set of moves. A move is the metacognitive strategy. The thing you’re actually doing.
These organized into six buckets. Six methods. Not sequential, you don’t work through them in order. They’re different ways you can engage with your AI depending on what you’re trying to do. Sometimes you start with a handshake. Sometimes you jump straight into building something. The framework is modular. It’s a way to work deliberately instead of by accident.
This expands my original “6 moves”. Here’s what I landed on:
Method 1: The Handshake. Alignment before engagement. How you configure the partnership so the AI stops guessing and starts building from what you actually need.
Method 2: Voice Dump Synthesis. Getting raw thinking into the conversation at the speed of speech. Most people filter their best ideas before the AI ever sees them.
Method 3: The Pattern Hunter. Finding meaning in large datasets through synthesis, not summarization. The difference matters.
Method 4: Blueprinting. Building structure before substance. Separating the architectural decision from the content decision.
Method 5: The Stress Test. Using AI to fight your own blind spots. Not about AI’s sycophancy problem—about your inability to evaluate your own work objectively.
Method 6: Recursive Drill-Down. Iterative refinement. First draft is never the point. AI handles mechanical generation; you handle evaluation and direction.
Over the next six weeks, I’m going to walk through each of these. What they are. Why the default fails. How to actually use them. The specific moves that make each one work.
I’ll share the thinking publicly. For paid subscribers, I’ll include the prompts themselves—the actual language you can use to put these methods into practice.
On prompts.
One thing I realized in building this out: there are three kinds of prompts, and most guidance only addresses one of them.
Instructional prompts are the kind you find in public libraries. Structured templates you can copy, paste, adapt. “Act as a senior strategist. Review this document for logical gaps. Be direct.” These tell the AI what to do. They’re useful—especially when you’re starting out. They give you a reliable way to perform a step intentionally.
I used to say don’t worry about prompts because the AI will write them for you. That’s still true. But I’ve also realized these instructional prompts are good seeds. They help you understand what a move is actually doing. And as you practice, they become less about copying and pasting and more about knowing what works.
Dialogic prompts are the natural evolution. This is where your personal style starts to enter. You stop copying and pasting because you know what reliably works and you just do it now—with your voice, in your way. The conversation becomes, funnily enough, conversational. Your specific nuance enters the chat. This is where augmentation really starts to happen.
Questions for your practice are different still. These aren’t instructions for the AI. They’re prompts for you. “What kind of thinking do I actually need help with right now?” When you engage with that question—out loud or in writing—what you say becomes the prompt. The question prompts you. Your response prompts the AI.
Over time, the line between these blurs. The instructional prompts teach you the moves. The dialogic prompts become how you naturally work. The questions become habits of thinking. That’s the goal: not better prompts, but better thinking that happens to produce good AI interactions as a byproduct.
Let’s start with Method 1.
Method 1: The Handshake
I’ve written about the handshake before. It was one of the first things I taught—a way to introduce yourself to your AI so it stops giving you generic, default answers.
What I’ve learned since then is that the handshake is much bigger than I originally understood. Going back through my archive, I can see how I handshake threads at different levels, in different ways, across different kinds of work. The method has expanded. It’s not just introducing yourself anymore. It’s everything you do to configure the partnership before the real work starts.
Most people start their relationship with AI the same way: they open a chat window, type a question, and hope for the best. It’s the equivalent of walking into a meeting with no agenda, no context, and no shared understanding of what success looks like. Then they’re surprised when the output feels generic, off-target, or weirdly corporate.
The problem isn’t the AI. The problem is the missing first step.
The Handshake is how you give the AI its instructions, its role, what you’re doing, the context it needs. At its simplest, it takes thirty seconds: state your purpose, set a role, name your constraints. At its most developed, it becomes a living agreement that deepens over time.
Think of it this way. When you hire a new team member, you don’t hand them a task on day one and walk away. You onboard them. You explain how you work, what matters to you, where the landmines are. The Handshake is onboarding for AI. Skip it, and you spend your time cleaning up misunderstandings instead of doing actual thinking.
Why the default fails.
Without a Handshake, the AI defaults to the statistical middle. It writes for no one in particular. It hedges. It gives you the answer that would be least offensive to the widest possible audience.
This creates two problems. First, the generic trap: you ask for feedback on your strategy document, and the AI doesn’t know if you’re a first-time founder or a twenty-year executive. It doesn’t know if you want encouragement or a hard critique. So it splits the difference and gives you something that sounds helpful but says nothing.
Second, the misalignment problem: you want a structural outline and the AI writes a full essay. You want a harsh critique and it gives you a polite compliment. You want it to push back on your assumptions and it agrees with everything you say. These aren’t failures of intelligence. They’re failures of alignment.
The Handshake fixes both problems at once. It replaces the AI’s generic defaults with your specific context, your preferences, your rules of engagement.
The architecture: four tiers.
The Handshake scales to match the scope of the work.
Tier 1: The Setup. Every new conversation or task. Thirty seconds to two minutes. You state what you’re doing, what you need from the AI, and any guardrails that matter. “I’m preparing a board presentation. Act as a critical reader who’s seen a hundred of these. Push back on anything vague or unsupported.” Three sentences. Completely changes what you get back.
Tier 2: The Install. Once per AI account, updated periodically. This is where you encode your preferences for how the AI should challenge you, how much preamble you want, what kinds of hedging annoy you. Most people have never spelled out their cognitive preferences for another human, let alone for an AI. The act of doing it is itself valuable.
Tier 3: The Introduction. Early in working with a new AI system, or when starting a major project. This is where you share your blind spots, your thinking patterns, your professional context. When you tell an AI “I tend to over-complicate things and lose people in the details,” you’ve given it a standing instruction to flag complexity.
Tier 4: The Recalibration. When quality drifts or the AI starts sounding repetitive. Every long-running AI relationship hits a plateau. This tier resets and re-aligns the working relationship.
The moves.
The Handshake has three core moves: Role Definition, Context Loading, and the Mutual Interview.
Role Definition is how you tell the AI who to be for this conversation. Context Loading is how you give it what it needs to know. The Mutual Interview flips the dynamic—instead of guessing whether the AI has enough information, you make it identify its own gaps. “Before you respond to my question, ask me 3-5 clarifying questions that would help you give a better answer. Do not answer yet.”
That last instruction—“do not answer yet”—is key. Without it, the AI will ask questions and answer simultaneously, which dilutes both.
What’s underneath this.
When you run a Tier 1 setup, you’re forcing yourself to articulate what you’re trying to do before you do it. Most people skip this. They start generating before they’ve finished thinking about what they want.
When you build a Tier 2 install, you’re articulating your own cognitive preferences. How do you like to think? When do you need support versus challenge? Most people have never made these explicit, and doing it changes how you work—not just with AI, but with everyone.
When you deploy the Mutual Interview, you’re inviting external monitoring of your own thinking. You’re saying: I might be missing something, and I want you to find it before it costs me.
The Handshake isn’t just a technique for getting better AI output. It’s a practice of becoming more deliberate about how you think.
That’s what makes it foundational.
The Moves in Practice
What follows are the prompts themselves—the actual language you can use to put the Handshake into practice.
For each move, I’ve included instructional prompts from public libraries—structured templates you can copy and adapt. I’ve also included what this starts to look like in practice, based on my own archive.
The instructional prompts are seeds. They help you understand what the move is actually doing. As you practice, you’ll naturally adapt them. Your voice will enter. The way you think about these things will start to shape how you engage. That’s the shift, from copying and pasting to just knowing how it works and doing it your way.
