The Loops You Run

Think of the last thing you meant to do and didn't.

Now think of the one before that. And the one before that.

If you're honest about the sequence — not the outcome, the sequence — you'll usually find the same handful of moves in the same order. An idea lands. It's exciting. You start reading around it, because reading feels like progress and is free of risk. Somewhere in the reading you find someone who has already done it better, and the excitement curdles into doubt. Doubt is uncomfortable, so you go looking for more information to resolve it. More reading. And the days pass.

That isn't a discipline failure. It's a loop — a small circuit of mental states you re-enter without noticing, that terminates in the same place every time.

---

Why loops are invisible from the inside

You never experience the loop. You experience one lap.

Each lap has a story that makes complete sense on its own. This one really did need more research. That one really was bad timing. This week was genuinely unusual. Every excuse is individually true, which is exactly why the pattern survives — the pattern lives at a level of zoom you never look at, and the story lives at the level you always do.

To see a loop you need two things a person, in the middle of their own life, structurally does not have: the sequence, and the count. Not "I procrastinated," but "in the last four months, from doubt I went back to researching eleven times and forward to starting twice."

That is not introspection. That's bookkeeping — and it's the sort of thing software is embarrassingly good at and humans are hopeless at.

---

Treating a mind like a graph

So we treat it like one. As you talk to Fracty, it extracts the recurring states you actually move through — anxious, overthinking, researching, doubting, relieved, shipping — and, more importantly, the transitions between them. Which state follows which, and how often.

Do that for long enough and you have a graph of your own mind: nodes for states, weighted edges for how reliably one leads to another. Normalize the weights into probabilities and it becomes a Markov chain — a structure with a useful property. You can ask it questions:

• Where does most of my time actually go, as opposed to where I think it goes?
• Which state, once I enter it, do I struggle to leave?
• What's the most likely next move after doubt?

And it answers with sentences like this one:

When you're in research, 71% of the time your next move is doubt.

Nobody enjoys reading that about themselves. In fairness, we've never shown someone a line like it and had them disagree.

---

The point isn't the diagnosis

It's tempting to stop here, because a sharp diagnosis feels like an ending. It isn't. Naming a loop changes nothing on its own — plenty of people can recite their pattern fluently and run it anyway that afternoon.

What a loop gives you that a personality label doesn't is an edge to cut.

A trait is a wall: "I'm a perfectionist" offers nothing to act on. A loop is a circuit with specific, identifiable transitions — and every one of them is a moment in time where something else could happen. If doubt → more research is the edge that eats your months, then that transition, and only that transition, is the intervention. Not "be more disciplined." Something absurdly small and specific: when doubt shows up, the next action has to be one that produces an artifact, however bad. Ship the ugly version. Send the half-formed message.

You aren't fixing yourself. You're re-routing one edge on a graph.

That's a tractable problem, and it's the only kind of self-improvement advice we've found that survives contact with a real week.

---

The fractal part

Here's the thing about loops that makes them worth this much attention: you don't have many.

The circuit you run on a side project is, at a different scale and with different vocabulary, the circuit you run on the career move and the conversation you've been avoiding. Same shape, different size. Which means the small loop you can actually catch — the one visible in a fortnight of ordinary messages — is a scale model of the one costing you years.

Catch the small one. It's the same shape as the big one.

Fractal is live at shinyfractal.ai. It watches for the sequence, not the excuse.