Essay · Economics & AI · 8 min read

The New Stable Orbits

How AI is rewriting which businesses can exist, and quietly opening the door to millions of new ones, from Brooklyn to Bangalore to Lagos.

In this essay
  • The thesis: viable business shapes widen at both ends
  • Why businesses behave like dissipative structures
  • What AI actually reduces: coordination overhead
  • Why the global opportunity is bigger than replacement stories
  • Why the middle gets squeezed
  • The new strategic question for builders and operators
The thesis. Flip the toggle.

Which businesses could actually exist, before and after AI?

Each dot is a business. Green means it can sustain itself. Grey means overhead eats the margin.

Minimum viable
$50K
20× lower than before
Solo ceiling
$20M
40× higher than before
Viable businesses
34
of 42 shown

A weird thing about smoke rings

Blow a smoke ring and watch it. It holds its shape for several seconds, a coherent spinning donut of air moving through other air. It has no walls. Nothing is holding it together except its own motion. And yet it persists.

Smoke rings, hurricanes, candle flames, and the convection cells in a pot of boiling water all share a strange property: they are structures made of flow. They exist only because energy is moving through them at the right rate.

Too little flow and they dissipate into nothing. Too much and they tear themselves apart. In between, there's a band where they're stable.

Physicists call these dissipative structures. They have a minimum sustaining condition. Below some threshold of energy throughput, the structure simply cannot exist. The math doesn't allow it.

Visual 1

The minimum sustaining flow
Move the slider. Watch when the ring can hold its shape.
Stable structures need a minimum throughput to exist, and a maximum before they tear apart.

Hold that picture in your head. We're going to leave physics now and never really come back. But the shape of the idea is going to keep showing up.

Businesses are dissipative structures too

A business, at its core, is a coordinated loop. Inputs come in (capital, labor, attention, information) and outputs go out, hopefully at higher value than the inputs. The difference is what keeps the loop spinning.

But here's the catch: a huge amount of what a business does isn't producing the output. It's coordinating itself. Bookkeeping. Scheduling. Customer service. Marketing operations. Light legal work. Compliance. Onboarding. Reporting. The endless overhead of just being a coordinated entity that exists in the world.

~30hrs
A typical week spent by a small-business operator on coordination overhead rather than the actual work that generates revenue.

Economists since Ronald Coase have understood that this coordination cost is what defines the boundary of a firm. Coordination eats margin. Below some threshold, it eats all the margin. The loop can't generate enough surplus to pay for the cost of its own existence.

That's the minimum sustaining condition. That's the smoke ring dissipating.

Visual 2

Two businesses. Same revenue. Different fates.
Each bar is a business. Revenue is the dashed purple line. Red = overhead. Yellow = direct costs. Green on top = profit.

What AI actually changes

It’s not the work. It’s the coordination.

Here's the thing AI does that almost nothing else has done: it doesn't reduce the work a business needs to do. It reduces the coordination overhead per unit of output.

The bookkeeping, scheduling, drafting, light analysis, customer triage, content production, and operational glue still have to happen. What changes is how much coordination labor the loop has to spend just to keep itself moving.

The 30 hours a week of coordination overhead collapses. Maybe to 5 hours. Maybe to 2. The output still gets produced, but the cost of keeping the business coherent falls through the floor.

Lower overhead, lower threshold
When coordination overhead falls, the minimum sustaining condition for a stable business loop falls with it.

Visual 3

Which businesses can exist?
Each dot is a real business. Toggle pre-AI vs. now to see which can sustain themselves.
8of 8 viable

A loop processing $80,000 a year (a niche newsletter, a specialized consulting practice, a tiny SaaS, a one-person telehealth clinic) can now find a stable orbit. Pre-AI, the overhead would have eaten it alive. Post-AI, it generates enough margin to sustain itself. It exists. It holds its shape.

The optimistic part nobody is talking about

The bigger story is global

Most of the AI-and-business conversation happens in San Francisco and reads like a story about replacing American knowledge workers. That story is real but it's also small.

The bigger story is what happens when the minimum sustaining condition for a business drops everywhere at once. A teacher in Nairobi running a tutoring practice. A tailor in Dhaka selling internationally. A pharmacist in Lima starting a telehealth clinic. A translator in Manila building a five-person agency.

None of these were viable as real, growing businesses ten years ago. The coordination overhead, the English-language paperwork, the foreign customer service, the tax compliance, and the marketing would have eaten them.

3.3B
Working-age people in developing countries who now have access to the same coordination layer as someone in San Francisco. The leverage gap is collapsing.
The same operating layer, everywhere

For the first time in history, the coordination tools of a global business are roughly as good and roughly as cheap whether you're in Brooklyn or Bangalore. The business loop doesn't care about your zip code. It cares about whether the overhead eats the margin.

That’s not a footnote. That’s possibly the largest expansion of economic agency in human history.

Visual 4

Same business, before and after AI
Same company, same revenue, but AI collapses overhead. Watch the red slab shrink and green appear on top.

The ceiling is rising too

If one person plus AI can do the coordination work that used to require a team of 20, the ceiling on a solo operation rises dramatically. We're already seeing businesses that would have been mid-market companies a decade ago, $5M, $10M, even $20M in revenue, being run by one or two people.

These aren't lifestyle businesses. They're real economic objects, just with an org chart that would have been impossible before.

Visual 5

The new shape of business sizes
How many viable businesses exist at each revenue level. The middle is hollowing out, but the tails are exploding.
The old distribution was one hump in the middle. The new one is two peaks at the extremes with a valley where most yesterday's businesses used to sit. The shape of the economy is literally different.

The squeezed middle

This is where the pressure lands

If you're running a $3M business with 12 employees today, you should be paying very close attention. You're in the most dangerous zone in the new distribution.

You have enough complexity to need real coordination structures: middle management, ops processes, formal handoffs, HR. All of that exists because, historically, coordinating 12 people to do $3M of work required it.

Pressure from below

Micro-operators with AI leverage are eating into your customer base. They serve niche slices of your market at lower cost because their overhead is near zero.

Pressure from above

The new mega-solos are eating your higher-value customers because they offer the responsiveness of a small operation with the capability of a much larger one.

The escape hatch

Some medium businesses do not get stuck in the squeeze. If they adopt AI aggressively, collapse coordination overhead fast, and rebuild around a much leaner operating loop, they can jump the valley and start behaving more like the new high-capability, low-headcount large operators.

This isn't a prediction that all mid-market businesses will die. Some will adapt by ruthlessly reducing headcount and rebuilding around AI leverage. Some will grow up and out of the valley. But the structural pressure on this middle zone is real, and it's coming from both directions at once.

Visual 6

Two worlds, side by side
The old business landscape vs. the new one. Same axis (size). Different shape entirely.
Same axis. Same scale. Completely different economies. The old world clustered around one stable middle. The new world thrives at the edges, and the middle is being squeezed from both directions.
The old world
One viable middleTiny ventures couldn't pay their own overhead. Big firms needed huge teams. Most businesses clustered in the middle.
One peak
The new world
Two thriving extremesA long tail of micro-loops at the bottom. A new class of mega-solos at the top. The middle gets squeezed from both sides.
Two peaks + valley

Why this matters

The real strategic question

The framing I'd offer is this: the set of businesses that can exist is a function of coordination cost. When that cost changes, the set changes. AI is changing it in a structural, persistent way. Not as a productivity boost on top of existing businesses, but as a shift in which business shapes are possible at all.

The strategic question isn't "how do I use AI in my existing business?" It's "what shape of business is now viable that wasn't before?"

Those are very different questions, and the second one is the one that matters. The smoke rings in the new economy look different. They're smaller and more numerous at one end, larger and stranger at the other, and the comfortable old shapes in the middle are starting to wobble.

Worth knowing which one you're inside of, and which one you could build.

The underlying idea borrows from dissipative structure theory in physics and Coasean firm theory in economics. The synthesis, that AI changes the minimum sustaining condition for business loops, is the part I find most useful for thinking about what to explore next.