Ten engineers. Ten salaries. The output of 4.6 people.
54% of the work goes into talking about work.
This isn't bad leadership. It's math. And it changes everything about how we should organize in the AI era.
The ProblemThe Coordination Tax
The assumption was obvious: more people, more output.
Fred Brooks disproved it in 1975. Every new hire adds a communication channel to everyone already there. Two people share one channel. Ten people maintain forty-five. Fifty people navigate 1,225.
Headcount scales linearly. Coordination costs scale quadratically. Output? It's what's left over.
| Team Size | Channels | Actual Output | Lost to Overhead |
|---|---|---|---|
| 1 + AI | 0 | 1.55 units | 0% |
| 5 + AI | 10 | 4.8 units | 22% |
| 10 + AI | 45 | 7.2 units | 54% |
| 20 + AI | 190 | 9.8 units | 68% |
Here's the twist that changes everything: AI doesn't add communication channels.
A solo developer with Claude gets a 55% productivity boost with zero coordination overhead. A team of ten using the same tools gets a similar per-person boost — but loses more than half of it to alignment costs, context-switching, and meeting debt.
The solo gets 100%. The team gets what's left after the tax.
The OptionsThree Configurations Emerge
AI doesn't eliminate the coordination tax. It makes it visible. Three organizational forms emerge:
Solo + AI
Pure Leverage
100% of AI boost flows to output. Zero channels.
Team + AI
Taxed Leverage
46-78% of boost survives coordination costs.
The Swarm
Network Leverage
N × 100% — solo efficiency at network scale.
The third option is the strange one. And potentially the most powerful.
The ShiftWhy Vision Becomes the Bottleneck
Execution used to be the constraint. Code, builds, deploys — these consumed the work.
AI made execution abundant. Now the bottleneck moves upstream.
Where Work Actually Goes
Execution
Code, deploys
Vision
Clarity, intent
Coordination
Alignment, sync
Ask ten employees what the strategy is. You'll get eleven answers. Only 5% can articulate it clearly. This was always expensive. AI makes it the only expense that matters.
One clear mind with AI beats ten confused engineers.
The Efficiency Cliff
The AlternativeThe Swarm Model
There's a third path. It looks strange at first.
Ten solos. One protocol. Zero meetings.
Definition
The Swarm model: Multiple solo developers, each with their own AI stack, aligned through shared standards and protocols — not synchronized through meetings. Ten intentions, one protocol, zero coordination overhead.
Each node operates independently. Each maintains full efficiency. They don't coordinate through process. They coordinate through protocol — shared APIs, standards, principles that let them ship without synchronizing.
Network, not hierarchy. Each node keeps full efficiency. Scale horizontally, not vertically.
This isn't theoretical. Linux has thousands of contributors who never meet. They coordinate through the code itself — protocol, not process. Open source proved the model works. AI makes it viable at smaller scales.
You never hit Brooks' Law because you never add the 10th person to a 9-person team. You start a new 1-person team.
The organization becomes a network.
The ImplicationThe Uncomfortable Truth
Follow this logic and you arrive somewhere uncomfortable.
Companies that need 100 engineers today might need 20 tomorrow. Not because AI replaced the engineers directly. Because a new organizational form made 80 of them unnecessary coordination overhead.
This is what happened to factory workers a century ago. Not displaced by machines directly — displaced by the organizational structures machines made possible.
The Question
The question isn't whether this is happening.
Are you building a team, or a network?
Are you optimizing for headcount, or clarity?
The future favors the coherent.
4.6 people — or ten, working alone. Together.