Strategies or Learning with AI by Duncan Duke and GPT5

Part 1: Our Philosophy of Learning

Let’s start with two reminders.

First, our definition of learning: to change the structure of your brain. The goal is to improve how you think, not just what you know (and will soon forget). True learning is about building mental models, not just memorizing facts.

Second, our method for learning: Read, Write, Do.

• You read to acquire new information and perspectives.
• You write to intensively process that information, find connections, and structure your own thoughts.
• You do to practice applying your knowledge, whether it’s through analysis, discussion, negotiation, or decision-making.

This process is where the heavy lifting of learning happens. It’s often difficult, but it’s what actually changes your brain.

Part 2: How AI Short-Circuits Learning

The main pitfall of using AI is that it offers a tempting shortcut around the hard work of learning. When you let AI do the work, the AI is learning, you are not.

This happens when:

• You don’t read the materials; you ask AI to summarize them for you.
• You don’t write to process your thoughts; you ask AI to write for you.
• You don’t do the analysis, discussion, or decision-making; you let AI do it for you.

I know I can’t stop you from using AI, nor will I always be able to detect it. It’s your responsibility to use it correctly—to deepen your learning, not to avoid it. If you cheat with AI, you are only cheating yourself out of an education.

Part 3: Using AI for Deeper Thinking

Used correctly, AI can be an incredibly powerful tool. Think of it as a collaborator, a personal editor, or a tireless sparring partner. Here’s how to use it strategically.

My Tips for Effective (and Trustworthy) AI Use:

1. Start with a Blank Page. Before you prompt the AI, do the initial cognitive work yourself. Write down your own first draft, initial thoughts, or outline. Use the AI to refine your ideas, not to generate them from scratch. This is the single best way to avoid the learning trap.
2. Use the Best Tool for the Job. Use the latest, most powerful models, which are often available through paid subscriptions (e.g., ChatGPT Plus, Gemini Advanced). These “frontier” models are vastly more capable than the free, simpler ones and will give you more nuanced and reliable results.
3. Write Detailed and Complex Prompts. Give the model as much context and as many steps to follow as you can. A lazy prompt gets a lazy answer. A detailed prompt forces the AI to engage its more powerful reasoning capabilities.
4. Argue with the AI. Don’t passively accept the first response. Push back. Ask for clarification. Tell it where you think it’s wrong and ask it to defend its answer. Treat it like a pupil whose thinking you need to sharpen.
5. Demand Options, Not Answers. Never let the AI give you a single “right” answer. Ask it for a variety of options, to rank them, and to explain its confidence in each one. Then, interrogate how it assigned those confidence ratings.
6. Verify and Distrust. This is critical. All AI models can and do “hallucinate”—they invent facts, sources, and data with complete confidence. Treat every factual claim from an AI as an unverified assertion. The burden of proof is always on you to check its work against reliable sources. Ask the AI to show you its sources, then review them yourself.
7. Use AI to Play Devil’s Advocate. Before you submit or finalize anything, ask the AI to destroy your argument. Prompt it with: “Act as a skeptical investor. What are the three biggest holes in this business plan?” This is one of the best ways to pressure-test your thinking.

Part 4: The Carbon Footprint of Your Curiosity

Using powerful AI models consumes a significant amount of energy, which has a real carbon footprint. As a responsible user, it’s worth understanding this impact and knowing what you can do about it.

Putting AI Emissions in Perspective: Let’s put this in context. A complex query on a frontier AI model might have a total lifecycle footprint of 1-4.5 grams of CO2e. For comparison, producing and printing a single sheet of black-and-white paper on a laser printer has a footprint of about 6-10 grams of CO2e. While you shouldn’t ignore the impact of AI, it’s useful to have a real-world benchmark.