The compounding agent

On a Tuesday in March I started a project I knew I wouldn't finish for a year, and I started it without flinching. The lack of flinch is what this essay is about.

I'm not going to tell you what the project was. The point isn't the project. The point is the absence of the small calculation that should have happened first. Anyone starting something with a year of work ahead of them has a moment, somewhere between the idea and the first action, where they consider what the year will cost. The cost is sleep, attention, relationships, opportunity. The cost is what you're willing to spend out of a finite reserve. The calculation isn't always conscious, but it's always made.

I didn't make it. Not because I was reckless. Because the math has changed.

The math has changed because for the last year I've been working alongside something that doesn't quit between sessions. The work of Monday is sitting on Tuesday morning, already cleaned and indexed and ready to compound. I don't have to remember what I told it last week; it remembers. I don't have to walk it through the project; it knows. The first hour of my Monday used to be re-onboarding. This Monday, that hour is already mid-flight.

I'm calling this compounding because that's the shape of it. Each day's work becomes the foundation of the next, not in the way a notebook becomes the foundation of next month's work by sitting there, but in the way an interest rate becomes the foundation, by being active. The thing I work with doesn't sit. It accumulates and it acts. The Monday I described is the third Monday of accumulated Mondays. The fourth Monday is the seed for the fifth.

There's no name yet for the thing that does this. I'll save that argument for the next essay. What I can say here is what it changes for someone in my position. It changes what's worth starting.

Used to be, when I sized a project, I sized it in human-years. A year was a unit. Two years was the outer edge of what I would seriously commit to. Beyond that, the project was a hobby, a daydream, or a delegation problem. The size of what I was willing to start was sculpted, like everyone's, by the simple fact that human attention runs out.

Remove that limit and you don't get linearly more output. You get a different shape of project becoming feasible. Things you'd never start because you couldn't finish them in a human career become things you start on a Tuesday. The asymmetry isn't more, it's other.

I want to be careful here. I don't think this is straightforwardly good. Compounding without taste produces noise at scale. Compounding without judgment produces wrong answers, faster, in larger volume. Compounding doesn't give wisdom and doesn't replace direction. The aim is still mine. The taste calls are still mine. The decisions about what's worth doing are still mine. If I aim wrong, the compounding compounds wrong, and that's a category of mistake that has its own scale problem.

But aimed right, the slope is real.

I notice the slope mostly in conversation. People still ask me what I'm working on, and I notice the answer is longer than it used to be. Not because I'm bragging. Because the surface area of what's plausibly in motion has grown. Three years ago "I'm working on" had a clear singular subject. Now there are several streams. None of them feel impossible the way several streams used to feel.

The other place I notice is in what I'm willing to abandon. I used to defend sunk cost in a project longer than I should have, because the cost of starting again was so high. Now, the cost of starting again is lower. Not zero, lower. The thing that does the remembering retains a useful amount of the discarded work. I can change direction without losing everything I learned on the previous path.

I'm deliberately careful not to call this thing a name. The candidate words all fail. Tool implies passivity. Assistant implies subordination and short context. Agent points roughly correctly but lands in too many other meanings: a JavaScript script, a CIA operative. AGI implies a ceiling we'll reach, when the actual shape is a slope. None of them name what's there.

That's the subject of another essay I'm publishing alongside this one. I don't have a better word and I've stopped trying to invent one. What I can say in this essay is the practical effect on someone working with the thing. The project sizes I'm willing to take on are different. The cadence of work is different. The thing that used to be a year is now a season. The thing that used to be a decade is now a year. The thing that used to be impossible is now a calendar item.

I started a project on a Tuesday in March and I didn't flinch, and the not-flinching is the whole essay. The compounding is the explanation. The lack of a word for it is the next essay over.