Coase's Law Breaks: When Coordination Costs More Than Building

This piece was sparked by The New Era of Jobs: Organizational Singularity with Peter Diamandis and Salim Ismail. Ismail’s framing was blunt: in an AI-native world, “Coase’s law no longer applies,” because you can now find yourself “building the feature cheaper than having the meeting about the feature.”

That’s a big claim leaning on an old, load-bearing idea. It’s worth checking whether the foundation actually cracks — or just shifts.

What Coase actually said #

In 1937, Ronald Coase asked a question economists had mostly skipped: if markets are so efficient, why do firms exist at all? His answer was that using the market isn’t free. Search, negotiation, contracting — what he called “marketing costs” and we now call transaction costs — are real. A firm forms when organizing an activity inside a hierarchy is cheaper than buying it on the open market, and it grows until the cost of one more internal transaction rises to meet the market price at the margin (Econlib). The idea was foundational enough to win the 1991 Nobel Prize (Nobel Prize), and Oliver Williamson extended it into the “markets versus hierarchies” framework that earned the 2009 Nobel (Nobel Prize).

So the boundary of a company is, in Coase’s telling, a cost comparison. That’s precisely why AI is interesting here: it attacks both sides of the comparison at once.

Why the equation tilts #

The podcast’s tweet-sized version — building beats meeting — points at something real. When an agent can stand up a working feature in minutes, the cost of execution falls toward zero. But the cost of coordination — the meeting, the spec review, the sign-off chain — is sticky, because it’s made of human calendars. When execution gets cheap faster than coordination does, the rational move flips: just build the thing rather than convene about it.

You can see the downstream effect in revenue-per-head. WhatsApp, at its 2014 acquisition, ran a service for about 450 million monthly users with roughly 55 employees (around 32 engineers) and sold for about $19 billion (NBC News; Sequoia) — years before LLMs, by deliberately minimizing managerial overhead. The AI-era versions push further: Telegram reported roughly $1.4 billion in revenue and its first annual profit in 2024 serving over a billion users with a core team often cited around 30 (TechCrunch); Anysphere (Cursor) reportedly ran from $100M to $500M+ ARR inside 2025 (TechCrunch); Midjourney built a $200M+ self-funded business with a small team (WebProNews). Sam Altman now openly floats the one-person billion-dollar company as newly plausible (WIRED).

A caveat worth stating plainly: several of those headcount figures are self-reported, mix in contractors, and the one-person unicorn remains a prediction, not a documented event. The direction is well-evidenced; the most viral specific numbers deserve a raised eyebrow.

Why “breaks” is the wrong word #

Here’s where the builders’ framing overshoots. Coordination cost doesn’t vanish — it relocates. The most useful counterweight comes from a 2025 NBER chapter literally titled “The Coasean Singularity?” Its authors find that AI agents do lower search, communication, and contracting costs — but they simultaneously “introduce frictions such as congestion and price obfuscation,” leaving the net welfare effect “an empirical question” (NBER). Costs get reshuffled, not deleted.

And orchestrating a swarm of agents is itself coordination. UC Berkeley’s California Management Review argues that unmanaged agent proliferation creates “organizational entropy” and platform lock-in that can undermine the very coordinating function a firm exists to provide (Berkeley CMR). Anyone who has watched ten agents step on each other knows the meeting didn’t disappear; it moved inside the system prompt.

There’s a deeper reason firms persist that a pure cost argument misses entirely — and, to their credit, the podcast names it elsewhere as the “fiduciary wedge.” Companies are also legal containers, liability sinks, fiduciary holders, brand-trust vessels, and capital-raising vehicles. Cheap execution dissolves none of those. So the honest reading isn’t that Coase’s law breaks. It’s that the equilibrium re-balances at a new margin: smaller firms, far thinner middle layers, more is-it-cheaper-to-just-build-it decisions resolving toward build.

What to do with this #

For operators, the practical takeaways survive the correction intact:

• Treat coordination as a cost center. If a feature is cheaper to build than to meet about, that’s a signal your approval chain is now the expensive part. Audit where decisions queue.
• Right-size from first principles. Don’t assume the headcount your industry considers normal. The competitive frontier is revenue-per-employee, and it’s moving fast.
• Budget for the new coordination. The cost you saved on human meetings reappears as agent orchestration, verification, and governance. Plan for it instead of being surprised by it.
• Keep the parts cost can’t explain. Liability, trust, and capital still want a firm around them. Shrink the coordination layer; don’t dissolve the container.

Coase isn’t wrong, and he isn’t obsolete. He’s the reason the shape of the change is predictable: lower the transaction costs, and the boundary of the firm moves. AI just moved it a long way, very quickly — and the companies that win will be the ones who redraw the boundary on purpose rather than defend the old one out of habit.