Surprise Debt

Programmers have a name for the shortcut that works today and charges you later. Ship the quick fix, skip the redesign, and the codebase still runs — but every future change now costs a little more, and the cost compounds, because the structure of the program has quietly fallen behind the structure of the problem. Ward Cunningham called it technical debt, and the metaphor stuck because it is exact: you are borrowing against your own future understanding, and understanding charges interest.

I want to move that ledger somewhere more personal. Not your codebase — your model of the world. Because the synthetic subconscious we have each been assembling — the autopay, the auto-route, the auto-reorder, the feed that knows what you want before you form the want — runs on exactly this kind of borrowing. Every routine the stack absorbs is a quick fix that works today. What it borrows against is your future ability to be surprised at a survivable size.

I have argued that attention is not a spotlight you aim but a fuse that blows: expectation carries the current of ordinary life, and attention arrives only when prediction fails — when the current gets too hot for the circuit of habit to carry. And I have argued that your remembered life is built from those failures: the retrospective clock runs on surprise, which is why smoothed years shrink in memory even as smoothed days feel long.

Put those two claims together and a third one follows, and it is the one I have been circling without landing on. When the stack absorbs a routine, it does not merely save you the minutes. It takes over the predicting. The grocery order predicts your appetite, the thermostat predicts your comfort, the route predicts the road, the calendar predicts your obligations. The fuse stops blowing — not because the world stopped changing, but because a machine now absorbs the prediction errors on your behalf and quietly re-models around them. Prices drifted, the road closed, your own tastes moved — and something metabolized each of those surprises so you never felt the heat.

A surprise you are never shown is not a surprise that never happened. It is a surprise you now owe.

That is the claim of this essay. Prediction error is not deleted by automation; it is transferred to the machine and deferred for you. The world’s drift away from your model of it continues at full speed. What stops is your metering of the drift. And un-metered drift behaves exactly like unpaid principal: it compounds, invisibly, in the gap between what the world is and what you last checked.

Walk the ledger slowly. On the day you hand a routine to the stack, your model of that corner of the world is current — you were, until yesterday, the one doing the predicting. From that day forward the world keeps changing and your model does not. The gap between them is the principal. The interest is subtler: it is the rising cost of re-modeling. Skills decay with disuse; reference points go stale; the small neighboring facts you used to absorb incidentally — what things cost, who to call, which street floods — stop arriving, because they rode along on a routine you no longer perform. Each season of automated smoothness makes the eventual reconciliation larger, and makes you less equipped to perform it.

Notice what the ledger does not say. It does not say the stack predicts badly. The stack predicts superbly; that is the whole problem. A bad automation fails often enough to keep you current — it blows your fuse regularly, in small denominations. A superb automation fails rarely, which means it accumulates your surprise in one large denomination and pays it out on a schedule you do not choose. This is the automation researcher’s oldest warning, stated by Lisanne Bainbridge four decades ago as an irony: the better the automatic system, the more atrophied the human left to handle the moment it cannot.

The debt gets called three ways, and only the first one is the one people plan for.

1. The outage The substrate fails — the service goes down, the account locks, the card expires, the region loses power. Everyone models this one, and everyone models it as an inconvenience, measuring the cost in lost minutes. The real cost is the reconciliation: every deferred surprise from that routine presents at once, to a person whose re-modeling muscles have been resting.
2. The drift Nothing fails visibly. The stack keeps predicting yesterday’s you — the old appetite, the old commute, the old convictions — while the actual you moves on. Because its errors are smoothed rather than shown, you can live inside a model of yourself that is quarters out of date, and the call comes as a dull, diffuse wrongness no dashboard reports: the life fits like someone else’s coat.
3. The exit You leave voluntarily — new city, new platform, new decade of life, or simply the decision to walk out of a system that no longer deserves you. Exits price in the debt. The switching cost everyone attributes to the platform’s lock-in is partly this: not the data you cannot take with you, but the world-model you never built because the platform was building it for you.

The third case is the one that interests me most, because it turns surprise debt into a sovereignty problem. A system does not need to hold your files hostage if it holds your calibration. The truest measure of dependence on a stack is not what leaves when you export your data — it is how much of the world you would have to re-learn on the day you stop being predicted for.

The answer is not refusal. I am not going to advise anyone to hand-navigate their city or memorize prices as a spiritual discipline; deferring surprise is precisely what made room for the work that matters, and I would make most of the trades again. Debt is an instrument, not a sin. The failure mode is not borrowing — it is borrowing unknowingly, at an interest rate no one quotes you, with the full balance callable at random.

So service it the way you would service any debt: deliberately, in small scheduled payments, on dates you choose. Do the absorbed routine manually once in a while — not for nostalgia, but as a reconciliation pass, the way an accountant closes the books. Walk the route without the map. Do the month’s numbers by hand once a quarter. Let the feed go dark for a weekend and notice what you reach for. Each of these is a small controlled surprise — an interest payment made in a denomination you can afford, at a moment you selected, which is everything the called balance is not. And each one deposits a marker in the retrospective clock, which is a compounding asset of its own: the reconciliation days are the ones you will remember, because they are the days the fuse actually blew.

Pay your surprise on purpose, in small denominations, or pay it on the world’s schedule, all at once.

Where am I wrong? Here is the honest failure condition. Humanity has always deferred surprise onto its systems — onto spouses who knew where things were, towns that held the knowledge, institutions that absorbed the shocks — and most of that debt was never called; we simply named the un-called balance civilization and lived comfortably on top of it. If the stack’s reliability continues to outrun every disruption that matters — if the outages stay short, the drift stays small, and the exits stay rare — then surprise debt is a balance nobody ever settles, and this essay is a farmer’s warning shouted at people who, correctly, never intend to farm. If a decade from now the people deepest inside the synthetic subconscious handle their outages, their drifts, and their exits no worse than anyone else, I was wrong, and the debt was free money after all. I do not think reliability compounds faster than dependence. But that is the claim to check me on.