The Leverage Stack of a Company of One
One person now compounds outsized output by stacking layers of leverage: skills, code, media, agents, and distribution. Name the layers and you can build the stack on purpose.
Two people put in the same hour of work. One ends the hour with an hour of output: a task done, a thing handled, and nothing that outlives it. The other ends the hour with something that keeps producing after they have walked away: code that runs while they sleep, a piece of writing that reaches ten thousand people they will never meet, an agent that does the next hundred versions of the task on its own.
Same hour. Wildly different output. The gap between them is not talent and it is not effort. It is leverage. And the reason a single person can now out-produce a whole team is that the number of usable forms of leverage just went up, and they stack.
Leverage is output per unit of your time
Strip away the mystique and leverage is one ratio: how much output you get per unit of your own time and attention. A person with no leverage trades an hour for an hour’s worth of results, and when they stop, the results stop. A person with high leverage puts in an hour that keeps paying out long after, or that commands the effort of many other units of work without commanding proportionally more of their time.
The classic forms have been understood for a long time. Labor is other people working on your behalf: powerful, but it comes with the coordination cost of hiring, managing, and aligning humans. Capital is money deployed to make more money: powerful, but it requires having capital first, which most people building something do not. Then two newer forms changed the game for individuals. Code and media are permissionless leverage: you do not need anyone’s money to deploy them and you do not need to convince anyone to work for you. You write them once and they work for you a million times over, at close to zero marginal cost, while you sleep.
And now there is one more, the one that reorganizes the whole picture: agents. Agents are labor without the coordination cost. That single sentence is the reason a company of one is suddenly a real category and not a slogan.
“Agents are the first form of labor leverage that does not come with the coordination cost of humans. You get the output of a team without the overhead of managing one. That is the change everything else follows from.
”
The five layers, and how each one multiplies
The useful move is to stop treating leverage as one vague thing and see it as a stack of distinct layers, each with its own properties. Named, they become buildable. Here is the stack, bottom to top, with what each layer actually does for you.
layers, each multiplying the ones below it
of them (code, media) are permissionless and scale without you
marginal cost to serve one more user once built
Skills and taste: the base that aims everything
At the bottom sits your own skill and judgment, and it is at the bottom for a reason. It is the lowest-leverage layer measured in raw output, because your skill produces only what you personally can do in the time you have. But it is the foundation the whole stack multiplies, because every layer above it amplifies the direction you point it in. This is the load-bearing part most people skip past on the way to the shiny layers, and skipping it is the classic mistake. More on why in a moment.
Code: work done a million times for free
Code is the first great multiplier. You solve a problem once, encode the solution, and it runs for every user, every request, forever, at near-zero marginal cost. This is why software has produced so many outsized individual outcomes: the same hour of building serves one person or ten million with no extra hour from you. Code is permissionless (nobody has to approve it) and it scales without your presence (it runs while you sleep). For a company of one, code is how one person’s solution reaches a market that used to require a whole engineering department.
Media: your judgment, copied without you
Media is code’s twin. Write an essay, record a talk, publish a piece of thinking, and it reaches people continuously without you being in the room. Like code, it is permissionless and it scales without your time. What it multiplies is different: code multiplies your ability to do a thing, media multiplies your ability to be known for thinking about it. It builds the audience and the trust that every other layer cashes in on. A person known for clear thinking on a subject has a distribution advantage that money often cannot buy.
Agents: a team without the management
Now the new layer. For most of history, if you wanted more done than your own two hands could do, you needed labor, and labor meant humans, and humans meant coordination: hiring, onboarding, managing, aligning, the meetings about the meetings. Coordination cost is why small teams are fast and large ones are slow, and why scaling a company usually means scaling its overhead faster than its output.
Agents break that trade. An agent is labor you can direct without the coordination tax. You do not interview it, you do not run its one-on-ones, you do not wait for it to be free. You spin up ten of them on ten tasks and the overhead of adding the eleventh is near zero. This is the layer that lets one person operate at the output scale of a team, and it is exactly why so much now depends on being able to verify what all that cheap labor produces, because the production stopped being the constraint the moment agents made it cheap.
Distribution: the multiplier on the whole stack
At the top sits distribution: the audience, the channels, the reach that decides how many people your output actually touches. It goes on top because it multiplies everything beneath it. The best code with no distribution reaches nobody. Mediocre code with great distribution reaches millions. And distribution compounds: an audience that trusts you brings you the next audience, every published thing feeds the channel that carries the next one, and reach earned is reach that keeps earning. Distribution is the layer that turns good work into results, and its absence is why so much good work sinks without a trace.
The order to build the stack
Naming the layers is only useful if it tells you what to do, and it does: it tells you the order. Most people build the stack in exactly the wrong sequence, chasing the flashy top layers before the base can aim them.
Sharpen the base first: skill and taste
Before anything multiplies, there has to be something worth multiplying. Get genuinely good at the thing and develop the judgment to know what good looks like. This feels slow and unglamorous next to spinning up agents, and it is the step that makes every later step worth taking. Leverage on top of weak judgment just produces more of the wrong thing, faster.
Build the code layer: encode your solution once
Take the thing you are good at and encode it so it runs without you. This is where one person’s expertise stops being an hourly trade and starts being an asset that serves a market. You are converting skill, which is bounded by your time, into code, which is not.
Build the media layer in parallel: get known for the thinking
Do not wait until the product is perfect to start publishing your thinking. Media compounds slowly, so it has to start early. The audience you build now is the distribution you will need later, and the trust it earns is the thing money cannot shortcut. Code and media are twins, so build them together.
Add agents: turn one operator into a team
Once you have skill aiming a real product and a growing audience, agents multiply your throughput across all of it: the building, the operations, the repetitive reps. This is where a company of one starts producing at the scale that used to need a payroll. But it comes after the base, not before, because an agent pointed by weak judgment just industrializes the mistake.
Compound distribution: let reach feed reach
With real output and a real audience, invest in the channels that carry it. Every piece that lands widens the channel for the next one. This is the layer that turns the whole stack into results and, done right, the one that keeps working hardest while you do the least.
Taste is the gate on the whole thing
Here is the catch that makes or breaks everything above, and it is worth stating as bluntly as possible: leverage multiplies direction, and it does not care whether the direction is good.
That is the whole risk of the stack. Every layer is an amplifier, and an amplifier is indifferent to what it amplifies. Point it at something valuable and it produces enormous value. Point it at something wrong, off, mediocre, or misjudged, and it produces enormous amounts of that. Code ships your bad decision to a million users instead of one. Media broadcasts your shallow take to ten thousand people instead of keeping it in your head. Agents execute your flawed plan a hundred times before you have noticed it was flawed. High leverage on bad judgment is not neutral, it is actively worse than no leverage at all, because it removes the friction that used to slow a mistake down and give you time to catch it.
“Leverage is an amplifier, and an amplifier does not check whether the signal is good. It multiplies direction. Which is why taste, the judgment about what is worth doing at all, is not one layer among five. It is the gate on all of them.
”
This is why the base layer is the base layer. Taste is the judgment about what deserves to be built, said, automated, and amplified in the first place, and it is precisely the thing that does not get automated. The models get more capable, the agents get cheaper, the reach gets wider, and every one of those advances raises the premium on the human who knows what is worth pointing all of it at. When production is expensive, mediocre taste is survivable because you cannot afford to produce much anyway. When production is free, taste is the only thing standing between you and an infinite firehose of well-made, widely-distributed, confidently-wrong output. I have argued elsewhere that taste is the actual job now, and the leverage stack is the reason: it is the one input the whole stack cannot supply for itself, and the one that decides whether all that multiplication was worth anything.
Build your stack
So here is the concrete version, the thing to actually do. Look at where your own time goes this week and sort it by leverage. How much of it is the base-layer trade of an hour for an hour, and how much is building something that keeps producing after you stop? Most people are shocked at how little of their week touches a multiplying layer at all.
Then pick the next layer up from where you are. If you are all raw skill and no code, encode one thing you do repeatedly so it runs without you. If you build but nobody knows, start publishing the thinking behind the building. If you have skill and audience but you are still the only pair of hands, put agents on the reps. And through all of it, keep sharpening the base, because it is the only layer that aims the rest, and the whole point of stacking leverage is that a company of one, built on good judgment, can now do what used to take a building full of people. The stack is available. Whether it produces something worth producing is still on you.
