There's a number you see quoted sometimes in management books, and most engineers dismiss it the first time they read it: in a team of eight, there are 28 possible communication paths. In a team of sixteen, 120. The formula is n(n-1)/2 and it's just geometry, but geometry is the thing that runs your day. Every one of those paths, on an active project, is a message you triage, a Slack reply you owe, a meeting that could have been async but isn't. The paths don't cost much individually. They compound ruthlessly.
What Six Years of the Tally Actually Showed
I spent six years as an engineer inside Kyiv's product industry. Good companies. Real products. Smart people. Somewhere in year three I started keeping an informal tally: how long did each feature actually take, and how much of that time was I writing code? The ratio got worse every year. By the end, for a typical ticket, my time with a compiler open was maybe fifteen percent of the calendar span. The rest was alignment. Meetings about the meeting. Review cycles where the reviewer was waiting on a decision from someone who was waiting on a decision from someone else.
The numbers are public if you care to look. U.S. workers average 35 hours of meetings per month. The load of unproductive meetings on individual contributors has roughly doubled since 2019, from 1.7 hours a week to 3.7. Seventy percent of managers describe their meetings as costly and unproductive, and they are the ones calling them.
A Scene Worth Being Specific About
Abstractions let everyone off the hook, so here is a specific one. On one project we had forty microservices and eight engineers. Deploys took three of us half a day. We had optimized for a scale we didn't have, and we paid for it every Tuesday. Nobody had designed the dysfunction. It had accreted. Someone good had made a reasonable call in 2019 about future-proofing, someone else had reasonably extended it, and two years later we were a small team pretending to be a large one, carrying the operational overhead of a company ten times our size. There was no villain. That was the unnerving part.
The Thesis: Coordination Is a Physics Problem
Here is the thesis I landed on, and the one I'd like people to push back on:
Coordination is not a management problem. It is a physics problem, and it has a price curve that organizations systematically underestimate.
A team of four adds a fifth person and gets roughly 25% more output. They also get 40% more communication paths. By eight people the paths are growing faster than the headcount, and by sixteen the second derivative has taken over. You can flatten this curve with good tools, clear owners, tight docs. You cannot eliminate it. Past a certain size the paths dominate, and the marginal engineer produces less than the coordination cost of adding them.
The Accounting Exercise That Tipped Me
In one column I listed what I could build in a weekend if I owned the full stack and made every call myself. In the other, what the same feature would cost to ship at work, measured in calendar time from idea to production. The ratio was never less than 20x. Often 60x. Not because my coworkers were slow. They were excellent. It was because the system they were operating inside exacts a tax that scales with its size, and no individual inside it can opt out.
I left in late 2025 and shipped Flowly in thirty days.
The Hero Was the Absence of the Tax
I am not going to pretend AI was the hero of that story, even though it helped. The hero was the absence of the coordination tax. One mental model. No alignment meetings. When I decided on a data shape at 10 a.m. I could ship it by lunch, because the only person who needed to sign off was me.
Pieter Levels is at roughly $3M a year with zero employees, and Photo AI alone was clearing about $138K a month last quarter. He is an outlier in magnitude, not in kind. The infrastructure for small, opinionated, one-brain products has quietly become unreasonably good.
Flowly is my negative-space answer to the years I spent inside the tax. One workspace for tasks, timers, and analytics, aimed at freelancers and small teams who run four apps to answer one question: where did my week go?
A Question for People Who've Made the Jump
If you've made this jump from a team to solo, I'd like to hear the actual delta in concrete terms. Not "it feels faster" but: what's your time from idea to production now compared to before? How many meetings did you attend in your last week at a company versus now? The coordination tax is something most people feel in their bones but rarely put a number on. If you have the number, I'd like to hear it.
FAQ
What is the coordination tax in software teams?
The coordination tax is the overhead cost of communication and alignment that grows non-linearly as team size increases. In a team of 8 there are 28 possible communication paths (n(n-1)/2). By 16, there are 120. Beyond a certain size, the coordination cost of adding an engineer can exceed their productive contribution. DORA's 2025 research confirms that AI productivity gains often stall when organizations lack healthy value streams, because the bottleneck is the overhead, not the coding.
Is a solo developer really faster than a product team?
For well-defined, opinionated tools, often yes, and by a wide margin. In the corporate environment described here, the calendar-time ratio was 20x to 60x slower than solo development. The advantage is not raw coding speed but zero coordination overhead: decisions, deploys, and direction changes happen at the pace of one brain instead of a committee schedule.
What is Flowly?
Flowly is a task manager with built-in time tracking, calendar sync, and analytics for freelancers and remote workers. One workspace instead of four separate tools. Free tier available, 14-day Pro trial with no card required. flowly.run
This post first appeared on the Flowly blog.