← All writing Final · Oct 2025 · Updated Sep 2026

Cyclic Rationality

Prediction only works on possibilities we've already imagined. A method for recovering imagination before verification: dream, then test.

Introduction

As a kid I took things apart and tried to put them back together. When my dad showed me a relay, I knew at once it was the part I needed for my tree house security system on my grandparents’ farm.

My childhood was also a bit of a disaster. My mom’s and dad’s families were at war, and the world was a scary, violent place that didn’t make sense. I craved control, and I had the curiosity and drive to go after it. I left home early and taught myself instead of going to school and college. In hindsight that was fortunate: my mind was shaped less by the modern era than by ancient philosophers, biographies, and the logicians of the 1600s and 1700s, who thought differently than we do.

Slowly I worked the world out: religion, philosophy, psychology, culture, politics, economics, computer science, and AI. I wanted reality to be safe, yet still surprising. That craving for control was really a craving for predictability, a small version of the universal human drive to reduce uncertainty, which Claude Shannon taught us to measure as entropy. I’m gripped by a simple idea: understanding is prediction.

But prediction can only work with possibilities we have already imagined. This essay is about recovering the imagination that has to come before verification. I call the method Cyclic Rationality. I’ll describe the problem that led me to it, trace where I think the problem came from, and lay out the method’s principles.

1. The Problem: Closed Minds

The idea came to me while reading Bernoulli’s Fallacy by Aubrey Clayton. Clayton shows that probability is only as good as the hypotheses and background information you bring to it. If you haven’t imagined the right possibility, no amount of careful updating will find it. I had started to see a problem in how the rationalist community thinks. Effective Altruism, LessWrong, AI safety researchers, and the existential- and suffering-risk communities were making grand projections, especially about AI, that didn’t add up.

Closer to home, when I shared exploratory ideas with peers, I got blank faces and narrow, closed questions. Why couldn’t they see what I was seeing?

So I set out to understand the difference between how we were thinking. Several ideas came out of that search: Jonathan Haidt’s elephant and rider, which echoes Jung’s picture of the conscious and unconscious mind; mind viruses, which describe how ideologies spread and take hold by exploiting cognitive weaknesses; and Cyclic Rationality. To see why this narrowness feels so suffocating, it helps to trace how modern reason developed.

2. How Reason Split

Before the Scientific Revolution, authority, not evidence, decided which ideas survived. It allowed some wild speculation and forbade much that was sound. Much of what it allowed was what I’ll call delusion: ideas held without any serious attempt to test them. Then came Galileo and his contemporaries, and a new rule took over: beliefs should be tested against evidence, not settled by authority. Among educated elites, delusion lost its standing. People stopped seriously entertaining wild stories.

That was real progress, but it had a side effect: ideas that can’t yet be tested now tend to be dismissed rather than explored. Stephen Wolfram’s A New Kind of Science is a modern example. Its claims are vast but hard to verify, because breaking a system into parts, the usual way we check a theory, fails when the behavior only shows up in the whole. They remain unconfirmed. Even so, I think it’s a live possibility that reality is, at bottom, computational.

The same era also split the sciences from the humanities. Once subjectivity became suspect, imagination came to look like an indulgence rather than an instrument of discovery. In The Two Cultures, C. P. Snow argued that the growing gulf between the two was a major handicap in solving the world’s problems.

The split was never absolute, but I think it helps explain what I saw in my peers. They struggled to speculate or hold a thought experiment for long. Industrial-era education, specialization, and the explosion of formal models all played a part. Logic stopped being a practical tool and became a formal one. And since the world was ablaze with progress, surely nothing was wrong?

3. What We Lost

Looking at the arc from authority to testing, I ask what we lost. My answer is that we lost much of our practice at creative thinking. We can crunch numbers with the best of them, but how good are our dreams? If you want inspiration, you may do better with a science fiction or fantasy novel than with a degree from MIT.

For most of human history, we dreamed. We told stories and made up unbelievable things, but at least we dreamed. The dreaming was real; it was authority, not testing, that decided which dreams survived. When testing took over, creativity didn’t die, but it was blunted and discouraged, and with it our ability to explore the space of possibilities.

This matters because discovery needs both halves: imagination to widen the space of possibilities, and measurement to narrow it. A culture that trains only the measuring half starves the source of its own discoveries.

Cyclic Rationality is my attempt to recover that ancestral capacity to dream without abandoning what science taught us. It doesn’t dismiss measurement. It puts it in a more helpful place: after imagination rather than in front of it.

4. The Principles of Cyclic Rationality

Reality is complex and emergent

The universe organizes itself through local interactions that produce global behavior you couldn’t predict from the parts. The whole can’t be reduced to its parts without losing something, as cellular automata, chaos theory, and computational irreducibility show.

Reality is only partly predictable

Some things are predictable and others aren’t, and mistakes happen when we draw the line in the wrong place. Predictions come from idealizing a problem: simplifying it while keeping the parts that seem to matter. Every idealization leaves out interactions with the rest of the world. Forget that, and we confuse the map with the territory.

All models are incomplete

Every description of reality is a reduction. Gödel showed that any consistent formal system rich enough for arithmetic contains truths it cannot prove. By analogy, no model can fully capture its territory. Taking systems apart and studying the pieces is powerful, but it works from the bottom up. Complex systems such as ant colonies, brains, and immune systems also need a top-down view that keeps the whole in sight.

Understanding is learned prediction

Human understanding is the result of learning predictive patterns. Predictability feels safe because it enables control. So we try to forecast the future. Yet no one has found a pattern underlying all of reality, and Shannon entropy reminds us that surprise can be measured but not eliminated. The right posture is epistemic humility.

There are two rationalities: qualitative and quantitative

Qualitative rationality explores possibility space, the set of all the ways information could be combined. Quantitative rationality measures and verifies, grounding ideas in evidence and shielding us from delusion. Each fails alone. Qualitative without quantitative produces delusional dreamers. Quantitative without qualitative produces smart idiots, trapped in sterile reductionism.

Qualitative should come before quantitative

Insight begins with imaginative synthesis, a brave step into the unknown. If verification comes first, we only consider ideas we already know how to measure, and exploration shrinks to fit our instruments. We have to stay open to ideas that outrun today’s measurements, because pursuing them is how new methods of measurement get discovered.

Cyclic Rationality unites the two

Cyclic Rationality is the disciplined alternation between imaginative expansion and empirical contraction, which I think is the rhythm of discovery itself. We dream up possibilities, then try to verify them, knowing we won’t always succeed. What survives testing becomes the ground for the next round of dreaming.

Conclusion: Dream, Then Test

I began by asking why my peers couldn’t see what I was seeing. My answer is that they had been trained to verify before they imagined. That order shrinks the world to what can already be measured. Cyclic Rationality restores the order: dream first, then test, then dream again from firmer ground.

Two practices follow. When you meet someone you disagree with, perhaps someone who seems a bit woo-woo, exercise your imagination before you judge. In my experience I often disagree with their methods, but I find real insight in their conclusions surprisingly often.

And when you think about the future, use your imagination to design inspiring visions, not only to quantify likely ones. I find it more inspiring to be a pioneer than a mere mapmaker: to go where no map exists yet, try to make things better there, and come back to tell people what I found. We really can do both.

If understanding is prediction, then imagination is the courage to predict the impossible.