Swappable was the fourth virtue of a dumb wire — a stranger can faithfully reimplement the interface, which is why R2 and StackIT could compete with S3 without asking anyone to rewrite a line of code. The same property has an institutional name, decades older than any of this series’ vocabulary, and it turns out to explain more than the wire metaphor alone ever could.

Exit Is the Cheap Signal

Albert Hirschman’s Exit, Voice, and Loyalty (1970) named the mechanism precisely: when Exit is available — when defecting to an alternative is cheap — it’s the sensitive signal that keeps quality honest. A firm losing customers doesn’t need to interpret complaints; the departure rate does the talking. When Exit is closed or expensive, the only channel left is Voice — complaints, regulators, PR pressure — slower, noisier, and considerably easier for an organisation to manage rather than be disciplined by.

This isn’t a moral claim. Quality doesn’t degrade because anyone decides to let it slide. It degrades because the one signal that reliably bypassed bureaucracy is gone, and everything left runs through channels the organisation itself controls.

Two Depths of Insulation

Not all failures of Exit are the same depth, and the difference predicts whether the problem can fix itself. Chrome’s dominance is economic insulation — switching cost is a dial, not a wall, and it moves with market share and rival browser quality. Germany’s Sunday-closing law, Ladenschluss, is constitutional insulation — traced to Article 140 of the Basic Law, immune to ordinary correction by design, because changing it requires clearing a bar set deliberately above market or even ordinary legislative pressure. Majority support for relaxed hours has existed in polling for years. The workaround is visible at the border, where Germans in Aachen drive into the Netherlands on Sundays. The law isn’t ignoring feedback because none exists. It’s structurally exempted from having to notice feedback that’s plainly there.

Economic insulation can self-correct if a credible rival re-emerges — the door is expensive, not welded shut. Constitutional insulation cannot self-correct by market pressure at all, no matter how loud Voice gets. The fix has to be imposed from outside, deliberately, as a rule that doesn’t ask permission.

The 22nd Amendment Is What the Fix Looks Like

For roughly 150 years, the US presidential two-term limit wasn’t law — it was Washington’s informal norm, held by convention, no enforcement behind it, working purely because nobody who could break it chose to. That’s Exit-by-custom: swappability by restraint, not by structure. It worked exactly until someone tested it. FDR won four elections, not because voters lacked Exit in principle but because the informal constraint turned out to have no teeth once someone was popular enough to ignore it. Only after that failure did the norm get codified — the 22nd Amendment, ratified 1951, unconditional, applying regardless of whether the incumbent is doing a good job.

Chrome is standing roughly where that norm stood before it was tested. The IE-era of browser competition was the Washington-norm phase — dominance held loosely in check by the threat of defection, nothing codified behind it. Manifest V3 was the FDR moment: someone tested whether the informal constraint still bound, and it mostly didn’t, because switching costs had risen enough that real discipline was gone. What hasn’t happened yet is Chrome’s 22nd Amendment — a structural, unconditional Exit guarantee that doesn’t depend on the incumbent’s restraint. The EU’s Digital Markets Act, with its interoperability and default-choice mandates for designated “gatekeeper” platforms, is the closest running analogue: Def-Push imposed specifically to re-manufacture cheap Exit, because Use-Pull alone had stopped supplying it.

It’s worth being precise about what the Amendment actually guarantees, because it isn’t what elections usually get credited with. Choosing between candidates on a ballot is Voice — a preference expressed within a bounded set, degradable the same way any Voice channel is, by gerrymandering, media capture, or fundraising asymmetry, without anyone’s vote technically being taken away. The 22nd Amendment doesn’t touch Voice at all. It guarantees Exit from one specific incumbent, unconditionally, even if most voters would happily have kept them. The real democratic Exit signal was never “choosing the winner.” It’s not re-electing the loser — the structural floor that operates whether or not Voice ever gets exercised, which is exactly why a district that returns the same incumbent for fifteen straight elections can mean genuine satisfaction, or can mean Exit was never actually live, and the vote count alone can’t tell you which.

Monopoly, Cartel, and Public Office Kill It Three Different Ways

Monopoly, cartel, and public office aren’t three examples of the same failure. They’re three distinct mechanisms for killing the identical feedback pressure, and the difference is what makes the diagnosis useful rather than just descriptive.

Monopoly kills Exit directly — no rival to defect to. Cartel is subtler and arguably worse, because Exit looks alive: several firms, apparent choice, while Voice’s actual discipline is gone by collusion. Four telecom operators raising prices in lockstep aren’t a monopoly by structure, but they’ve manufactured monopoly’s outcome by agreement — the Exit option exists on paper and does nothing, which is worse than its visible absence, because it hides the failure from the people relying on it. Public office kills Exit by design, not failure — there was never supposed to be a competing townhall. Three different causes. Identical output: nothing left to correct the correction loop.

CEOs Are the Same Curve, Measured

Corporate boards ran the natural experiment on this without framing it in these terms. A widely cited 2012 study, tracking 365 US companies over a decade, found the statistical bend from a CEO adding value to a CEO becoming waste sits at roughly 4.8 years — the point where the search for outside information tends to stop, the same yes-people accumulate, and the firm-customer relationship, having strengthened for a few years, begins to erode even as the firm-employee relationship keeps improving. The average CEO tenure over the same period ran closer to 7.2 to 7.6 years. That gap between the optimal point and the actual average is the same diagnosis as an unamended two-term norm: nothing structural forces Exit at the point the value curve turns, so incumbency drifts two or three years past it by default, held in place by informal restraint rather than a rule.

A board is Voice with teeth, in principle — it can remove a CEO at any time, unlike voters facing a fixed term. In practice, boards are frequently populated or influenced by the incumbent they’re meant to be checking, the same entanglement that makes a captured regulator worse than an absent one. Term limits for CEOs would do for corporate leadership exactly what the 22nd Amendment did for the presidency: replace Voice’s hope that the board notices the bend in time with a structural guarantee that Exit happens regardless of whether anyone in the room is willing to say so out loud.

The Telemetry Keeps Running After Exit Dies

Every one of these keeps producing numbers long after the numbers stopped meaning anything. A cartel can still watch market share. A monopoly utility can still run satisfaction surveys. A public office can still collect complaint forms. The instrument doesn’t go quiet when Exit dies — it keeps generating a chart, and that chart gets cited as evidence of responsiveness precisely because it’s been quietly decoupled from the one thing that made it trustworthy: the possibility that measured dissatisfaction could ever cost the incumbent something.

Voice without a live Exit threat behind it isn’t feedback. It’s theater that happens to generate a chart.

The actual test for whether Swappable still holds, sharper than “is there technically an alternative”: does dissatisfaction, expressed through the cheapest available channel, cost the incumbent anything real? Not whether a complaint can be filed — whether defection can happen at a price low enough that the incumbent has to care. It’s the same question the four virtues were quietly answering from the start, aimed now at institutions instead of wires: not is this honest, but does it cost anything to stop being honest.

This post is part of the Gutenberg/Semantic series. Related: In Praise of Dumb Wires on Swappable as the fourth virtue this post extends, and S3 Out-Evolved HDFS by Refusing to Be Smart on the technical case cheap Exit was originally built from.

Sources: Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States, Albert O. Hirschman, Harvard University Press, 1970; Luo, X., Kanuri, V.K., and Andrews, M., “How Does CEO Tenure Matter? The Mediating Role of Firm-Employee and Firm-Customer Relationships,” Strategic Management Journal, 2012.