Since the advent of generative AI, I’ve had this recurring thought that we’re using it wrong. It should be refining our thinking, not replacing it. Last year, while waiting on a flight at Denver International Airport, I thought of a way to put my idea into action. I opened the voice version of the ChatGPT app and, instead of asking it a question, I asked it to question me.

Specifically, I asked it to test me on the “history and meaning of the American Revolutionary War.” I expected a short trivia session. Instead, I experienced an oral exam, the old-school format, where the examiner listens to your answers and then drills into whatever you seem least sure about.

ChatGPT was remarkably good at this. It asked me about the origins of the conflict. I gave some basic answers that we all learned in school, referencing taxation without representation, independence, and freedom. It asked me for specific examples. I talked about the protests and acts of violence that sparked the war. It asked for specific examples again. I could cite the Boston Tea Party, then found myself grasping to recall more.

Within a few minutes, the AI had quickly probed to find my weak spots. Rather than filling in my knowledge gaps with answers, it revealed them through questions. The experience was the opposite of how I’d been using AI up to that point, and the opposite of how almost everyone is using it now. Instead of doing the thinking, it made me do it for myself.

Every time we reach for a chatbot before we reason through something ourselves, our reasoning muscle weakens a little. It’s like skipping the gym for a month and then realizing you can’t do what you used to. The right way to use AI is to help us strengthen our thinking, not replace it.

The LLMs used for generative AI are especially good at synthesizing large amounts of information and generating ideas. Helping us explore ideas, distill them, and refine them — these are all perfect use cases for generative AI. Without intentionality and discipline, these same capabilities let us skip the work of thinking.

There’s a second problem tangled up with the first one. We’ve lost the places and the practices where we used to sharpen our thinking against other people’s ideas. Constructive dialogue is becoming rare. We’re all missing the arguments, the debates, and the conversations where someone pushed back and you had to defend what you believed. Social media has compressed our attention, polarized our positions, and trained us to perform instead of reason. Civil discourse has been the bedrock of democracy. We’ve lost both parts of the phrase.

AI weakening our thinking and social media supplanting civil discourse - these two problems are making each other worse. The more I’ve thought about it, the more I’m convinced one of these problems can solve the other. The same AI that’s weakening how we think can be turned around to strengthen it, if you design it to challenge rather than comply.

Accompanying this new Substack, I built Solo Agora as a suite of tools to demonstrate my point. At present, the Solo Agora platform offers five tools designed to push back on your thinking, probe your assumptions, test your positions, and drill what you’re trying to learn. I think of it as reasoning fitness. Much like a gymnasium has different areas for exercising different muscles, Solo Agora provides structured tools for exercising different dimensions of thinking. The agents use Claude, ChatGPT, or Gemini, the same tools you can use directly, but these versions are curated and defined by me. They include constraints that eliminate the easy ways out that human nature tempts us to take. My version of the tools won’t do the thinking for you. That’s the point.

The flagship tool is Hone. It uses the Socratic method to stress-test a position, either one you hold and want to sharpen, or one you’ve heard and want to learn to challenge. You start a session by giving it a clear proposition, then you defend it and explore it by way of the questions Hone prompts you to answer. Other tools include Sway, which helps you learn to convince someone of your point of view, and Proctor, an evaluative tool to test your grade-level knowledge of a subject of your choosing - much like my airport experience.

Everything you can do in the tools, you can do on your own in any of the top AI platforms. I’ve put a lot of thought into curating the experience and the guardrails, but I’m not claiming a monopoly on Socratic questioning. What I am claiming is that most people won’t do it on their own, because the default mode of every AI assistant is to comply, and we have to actively fight that. Solo Agora is designed so the fighting is already done for you.

The name Solo Agora is deliberately paradoxical. The agora was the ancient Greek public square, a place for commerce and public announcements, but also a gathering place where people could converse, share ideas, hear public speeches, and engage in civil discourse grounded in critical thinking and reason. It’s sad, but in our polarized environment, we seem to have lost this practice, and perhaps we’re losing the skill itself. Solo implies private — a public square for one. It’s designed so you can hone your thinking in a safe space. Thinking harder is a good outcome in itself. If you choose to take better arguments into the world, that’s even better.

This Substack is, ironically, me taking my ideas out into the world. It’s the public counterpart to my private practice. I intend to share thoughts on critical thinking, civil discourse, what AI is doing to our reasoning, and what we might do about it. I’ll point to the product occasionally because it’s relevant, but the writing here will stand on its own.

The accompanying platform is free for limited use. The free tier resets monthly and is genuinely usable, not a teaser. The monthly subscription is priced modestly, mainly to cover my costs. There are no advertising cookies tracking your behavior, no gamification loops, and no schemes to monetize your data on the side. You’re the user, not the product. If this sounds like it was built by someone who wanted the tools to exist more than he wanted to run it as a business, you’re right.

If you want to be part of this experiment, subscribe to Solo Agora on Substack. If you want to try the agents, check out SoloAgora.com, where I’ve also written more about why the platform was developed. And if you want to watch someone try to practice what he’s preaching, you’re in the right place.

Subscribe now