---
title: "Cancrity: Growth That Severs Its Feedback"
url: "https://toddpaulbrownjr.com/corrigibility/cancrity/"
author: "Todd Paul Brown Jr."
description: "Some systems thrive by using up what they depend on. How to tell that from ordinary competition, and a three-line test that keeps the diagnosis honest."
kind: "guide-chapter"
updated: "2026-09-26T02:50:16+00:00"
---

# Cancrity: Growth That Severs Its Feedback

In the late 2000s, a major department store chain tried an experiment. According to one widely reported account, the company split itself into roughly thirty business units (apparel, tools, appliances, and more) and had each one operate largely as its own business, with its own profit-and-loss statement, dealing with the other units almost as if they were outside vendors. The stated idea was that internal competition would show which parts of the business were strongest and steer money toward them. Let the units compete. Let an internal market decide.

Picture what a structure like that invites. The appliance unit wants the marketing budget spent on appliances. The apparel unit wants it spent on apparel. Neither wants to pay for anything that mostly helps the other. Shared functions, the website, the store upkeep, the logistics that move goods from warehouse to shelf, belong to no one's profit-and-loss statement, so no unit has much reason to invest in them. A shopper walking into a store run this way is walking into thirty small businesses that happen to share a roof, each one optimizing itself and none of them optimizing the store.

Here is the strange part. By the units' own numbers, much of this can look rational. A unit head can point to a quarter where costs were down and margins up. Individually, plenty of units can be "winning" while the whole gets sicker. Customers notice the decline before the units' spreadsheets do, because the spreadsheets measure the unit, not the store, and the store is what every unit depends on to have customers at all.

The chain eventually filed for bankruptcy. Its decline had many causes, among them online competition, heavy debt, and years of underinvestment in its stores, and no honest account pins it on one. But reporting at the time linked the unit structure to the chain's troubles, and the shape it illustrates is worth isolating on its own: parts that thrive by their own measures while the whole they share wears away beneath them.

That gap, thriving parts and a dying whole, is the subject of this chapter.

## Naming the Pattern

There is a name for a process that sustains itself by degrading the very thing it depends on to survive: *cancrity*. A cancritic system is one whose persistence comes at the cost of the substrate that persistence requires. The cost keeps compounding, because nothing in the system's feedback registers it as a cost.

The name comes from cancer, and it's worth saying exactly what that borrowing does and doesn't claim. A cancer cell thrives locally. It divides, grows, and outcompetes the healthy tissue around it, while destroying the body it cannot survive without. That is the source model, not the evidence. A company's business units are not tumors. Nothing about corporate structure resembles a cell's biochemistry. What repeats is the shape: local success, bought by degrading a shared substrate, with no mechanism catching the trade before it's too late. Same failure shape, in entirely different machinery. Every time this chapter reaches for cancer as an illustration, that's the deal: a shape borrowed, not a diagnosis of literal disease.

It would be easy to hear "cancrity" and think it means aggressive or ruthless behavior, a fancy word for playing hardball. It doesn't. The units in that picture wouldn't be cancritic because they fought each other for budget. Fighting for budget is what business units do. They'd be cancritic because the fighting drew down something none of them could survive without (the shared store, the shared brand, the shared customer relationship) faster than anyone replenished it, and no unit's numbers ever flagged the draw-down as a problem.

Contrast that with a relationship most people understand without a business degree: wolves and caribou. Wolves kill caribou. That's predation, and there is nothing gentle about it. But the two populations are coupled. When wolves are numerous and hunting is good, caribou numbers fall. With less prey, fewer wolf pups survive, and the wolf population shrinks. With fewer wolves, the herd recovers, and in time the wolves do too. Too many wolves, fewer caribou, fewer wolves; too few wolves, more caribou, more wolves. Real ecosystems are messier than that loop, with weather, disease, and other predators and prey all in play, but the core coupling holds. Nobody designed that correction. The wolves' success is answerable to the state of the herd, whether the wolves "know" it or not.

Now put kudzu next to it. Kudzu is a vine from East Asia, introduced to the United States at the 1876 Centennial Exposition in Philadelphia as an ornamental, and later promoted by the federal Soil Conservation Service in the 1930s to control erosion on damaged farmland. In its native range it lives inside a web of insects, diseases, and competing plants that keeps it in check. In the American South, without most of that web, it did something wolves don't do to caribou. It didn't cull what it grew on. It could smother it, climbing over shrubs and trees and shading them out, with little pushing back. The federal government stopped recommending it in the 1950s and later classified it as a weed. That's not predation with a corrective loop. That's growth with no coupling to what it grows over.

Wolves and kudzu give the chapter its dividing line, and the line is the whole point. Cancrity is not competition, and it is not aggression. It is the specific case where a system's success has come structurally uncoupled from the persistence of what it depends on. Call that the coevolution criterion. A system is in a healthy relationship with what it depends on when its success stays answerable to that thing's health: when damage to the substrate shows up, sooner or later, as a cost to the system itself, and the system adjusts. The caribou don't need wolves to thrive, but the wolves' fortunes track the caribou's, and that tracking is what keeps the relationship going. A relationship turns cancritic when one side's success no longer registers what is happening to the thing it feeds on.

Notice that the coevolution criterion is a claim about a relationship, not about either party alone. A wolf pack examined with no caribou in the picture tells you nothing about whether it's cancritic. The question only has an answer once you ask what the pack's success is coupled to. That's why the diagnosis in this chapter is never "this system is cancritic," full stop. It's always "this system's relationship to this substrate is cancritic," with the substrate named. Drop the substrate and the sentence stops meaning anything.

## Not a Switch but a Trend

It would be convenient if cancrity were a yes-or-no diagnosis. It isn't, and pretending otherwise gets the tool used badly. Almost no real system is purely coupled or purely decoupled at any moment. What matters is which way it's moving.

The workable question is directional: *is this system's success becoming more or less answerable to the health of what it depends on?*

Take two hypothetical video platforms, each sharing ad revenue with the creators who make its content. One quietly renegotiates that share downward year after year while creator output keeps lifting its engagement numbers. The other responds to a wave of creator burnout by raising the share, or by building tools that help creators see how their work is doing before it flops. The first is moving one direction and the second the other, even if neither is close to a "fair" split by any absolute standard. Or take a fishery. One that adopts catch limits tied to actual stock counts, revised each season by outside biologists, is tightening the coupling between the fleet's fishing and the fish's survival. One that keeps limits flat while stock counts fall is loosening it.

None of these trends announce themselves in a single quarter. That's why the directional form matters more than a snapshot. A system caught at one moment in a rough patch, cutting costs and tightening belts, can look identical to a system in real decline unless you can see which way the line is bending.

It's also why a single bad year proves nothing on its own. A company that slashes its training budget once, during a real downturn, and restores it the next year is not on a cancritic path. The line dipped and came back. A company that slashes it once and then makes the smaller number the new baseline, year after year, calling each fresh cut "temporary," is. You can't see the difference in any single year's balance sheet. You see it only in the shape the years make when you line them up, which is why a healthy system keeps that line-up where anyone can look at it, and a cancritic one tends not to.

The next section gives that judgment a shape sharp enough to use.

## Naming the Substrate

A cancrity diagnosis is not a vibe. It has a form, and the form has three parts, each doing real work.

First, name the specific substrate the system depends on. Not "the ecosystem," not "society," but the actual thing, specific enough that someone could go check on it. For the department store's units, the substrate wasn't "the company" in the abstract. It was the shared store experience, the brand, and the logistics that every unit needed and none of them owned.

Second, show that substrate degrading faster than it's being replenished. Not just present: declining, at a rate that outpaces repair.

Third, show the decoupling itself: that the system's own success no longer registers the damage. The unit's numbers go up while the store goes down, and nothing in the unit's numbers says so.

Skip any one of the three and you don't have a diagnosis. You have a complaint. This is the discipline the concept lives or dies on, because "cancrity" is dangerously easy to reach for whenever a system does something you dislike, and the three-part form is what keeps it honest.

The form also forbids a tempting shortcut. A system consuming its own substrate is different from a system taking resources from someone else. If an executive steers vendor contracts to friends, that is favoritism, or corruption in the ordinary sense. It is not automatically cancrity, because what's being drained (a rival vendor's share of the business) is not a substrate the executive's company depends on to exist. Cancrity names a system eating what keeps *it* alive. "Redirects resources toward insiders" is a different failure than "extracts from its own members." Bundling the two, treating every case of a system favoring insiders as if it were consuming its own life-support, empties the word of the work it's supposed to do. Keep them separate on the page every time.

Here is the form done cleanly, on a case with no drama attached. An employer depends on its labor pool: the specific people trained to do the specific work, plus the pipeline that trains their replacements. That substrate can be shown degrading: a rising turnover rate, a thinning pipeline of trained replacements, exit interviews that used to surface real problems and now get skipped or ignored. And the decoupling can be shown directly: management's bonus targets are hit whether or not turnover rises, because the targets are tied to this quarter's output, not to whether there will be enough trained people to hit next year's. Substrate named. Degradation shown. Decoupling shown. Three lines, three facts, and the diagnosis holds weight because each line can be checked.

Notice what the form doesn't require. It doesn't require reading anyone's mind or proving intent. Nobody at that employer needs to be plotting the labor pool's collapse for the diagnosis to hold. A cancritic pattern usually runs on ordinary incentives followed faithfully. What the form does require is that all three blanks point at the same substrate. A diagnosis that names the labor pool in blank one and then offers evidence about customer complaints in blank two hasn't shown anything. It has shown two different problems and called them one. Precision in the blanks is what separates an audit from an accusation.

## What This Is Not

Because the word is easy to misuse, it's worth spending as much time on where it doesn't apply as on where it does.

The first exclusion is creative destruction with the coupling intact. A new company enters a market and wrecks the incumbents: puts them out of business, empties their offices, ends careers. That can look exactly like malignant growth. Something is thriving; something else is dying. But watch what the newcomer stays coupled to. An entrant whose success stays tied to serving its customers well and paying its suppliers fairly is not cancritic just because its success is somebody else's funeral. Incumbent death is not substrate death. The incumbents were never the entrant's substrate. Its customers and suppliers were, and if those relationships stay healthy while the market reshuffles, the reshuffling is competition doing what competition does.

A useful check: does the newcomer's growth still depend on treating its customers and suppliers well, or has it started drifting too? Imagine a hypothetical delivery company that wins by paying its drivers well and keeping its partner restaurants solvent. That is creative destruction for whoever it replaces. If the same company later starts cutting driver pay and raising restaurant fees because it no longer has to compete for either, it has started a second, separate story. That story gets checked with the three-line form like anything else, not waved through because the company was once the healthy disruptor.

The second exclusion is telon conflict. Picture an environmental regulator and a development agency working in the same region. The regulator's mandate is to protect a watershed. The development agency's mandate is to build the housing the region badly needs. Their goals collide, directly and often: a proposed development sits on wetland the regulator is chartered to protect. Neither agency is off course by the terms of what it's for. Neither is degrading a substrate it depends on. Each is pursuing a mandate that runs into the other's. This is goal conflict between two intact systems, not malignancy in either one. Cancrity describes a system's relationship to what sustains it, never rivalry, however sharp. This exclusion matters, because rivalry is exactly where the word gets reached for wrongly. In any fight it is tempting to describe your opponent's success as a disease. Two systems can be in real, unresolved conflict and both be healthy in the sense this chapter cares about.

Run the three-line form on the regulator and it fails at the first blank. The development agency is not a substrate the regulator depends on to exist, and the regulator is not one for the developer. Each depends on its own funding, its own legal mandate, its own constituency, and there's no evidence either is degrading those. What's happening is two systems with legitimate, colliding jobs, arguing about the same parcel of land. Calling that cancrity doesn't sharpen the disagreement. It just adds an uncheckable diagnosis to a conflict that was already honest about what it was.

The third exclusion is hard times, visibly handled. An organization drawing down its reserves in a real crisis (spending savings to survive a bad year, cutting a program to keep the lights on, running a deficit with an announced plan to close it) is not cancritic. The coupling is intact and speaking: the drawdown is visible, explained, and comes with a stated path back. Cancrity is what a drawdown looks like when nobody's tracking it, nobody's explaining it, and nothing in the system's own success measures ever has to notice it happened. The tell is usually whether the explanation survives the next good year. A drawdown that is rebuilt when conditions recover was crisis management. A drawdown repeated every year, good or bad, each one called a one-time exception, has become the operating model, and it owes the same three-line form as everything else here.

## Turning the Tool on What You Like

Every diagnostic tool can quietly become a weapon that only points outward. If the checklist above only lights up on institutions the reader already distrusts, it isn't measuring structure. It's confirming a prior opinion and calling the confirmation an analysis. The only way to know which is happening is to run the checklist on something the reader is fond of, in the same breath and by the same standard, as something the reader is more likely to suspect.

Start with something many readers of this book are fond of: a popular open-source project, the kind of free, volunteer-built tool that thousands of companies and developers rely on every day. It depends, above all, on its maintainers: the small group of volunteers who review contributions, triage bug reports, fix security problems, and cut releases. Name the substrate: the maintainers' time and willingness to keep doing that work. Show the degradation: an issue queue growing faster than anyone can clear it, review times stretching from days to months, long-time maintainers stepping back one by one with nobody trained to replace them. Show the decoupling: the project's visible measures of success, its download counts, stars, and number of projects that depend on it, can keep climbing while the maintainer pool thins, because none of those numbers registers who is doing the work or how tired they are. Nothing in that picture requires anyone to behave badly. It requires only that the project's success is counted in a way that can't see its own substrate.

Now take something many readers already regard with more suspicion: a large company whose products are built on top of open-source libraries it didn't write and doesn't pay for. It depends on those same maintainers, whether or not it thinks of them as part of its supply chain. Name the substrate: the maintainer pool behind the libraries its products need. Show the degradation: the same thinning pool, the same aging queue, security fixes arriving late or not at all. Show the decoupling: the company's revenue and shipping speed can rise for years while its critical dependencies are kept alive by fewer and more exhausted volunteers, because no line on its dashboard tracks the health of code it didn't write.

Notice what just happened. The same three-line form, applied without changing a rule, produced a live question about a community many readers belong to and a company many readers might already distrust. Both runs named the same kind of substrate, one each system genuinely depends on, rather than a cost that merely lands on someone else. That is what the tool is for. If it only ever fires in one direction, it isn't a diagnostic. It's a mirror, and a mirror won't tell you anything you didn't already believe walking in. A diagnostic has to be willing to return an uncomfortable answer about something you like, or it was never measuring anything.

It's also worth saying what the tool does not deliver here: a verdict. Nothing in the three-line form hands down a judgment about the project or the company. What it hands down is a structural finding: a claim about which way a coupling is trending, checkable against evidence, revisable if the evidence changes. What anyone does with that finding is a separate question, and this book keeps declining to answer it for the reader.

The symmetry test isn't a one-time exercise, either. It's a habit worth building into every use of the checklist. Before running the three-line form on an institution you distrust, run it once on an institution you belong to, work inside, or send your own children to. If the second run comes back clean, good: the tool just earned some credibility. If it doesn't, that's the more useful outcome, because you've found something worth fixing before it found you.

## Cancrity and Its Neighbor

The previous chapter named a different disease: a system optimizing a proxy until the proxy and the real end come apart, and the system keeps chasing the proxy anyway. That disease is about *direction*: the system is aimed at the wrong thing. Cancrity is about *relation*: the system's success has come uncoupled from what keeps it alive, whatever it's aimed at. They are different failures, and they often show up together.

The bank from Chapter 4 is an example of both at once. Its cross-sell number had drifted from the relationship depth it was meant to track, which is the proxy problem. And the pursuit of that number was spending down customer trust, a substrate the bank depended on and was not replenishing, which is the substrate problem. Two diagnoses, two different questions: what was the system optimizing, and what was that optimizing costing it? A system can have either disease alone. Telling them apart tells you which repair you need: fix the metric, or restore the coupling, or both.

## The Instrument

Anyone can use the three-line form on any system they're wondering about. It reads like this, filled in for whatever case is in front of you:

*This system depends on ________.*
*That substrate is being degraded faster than it's replenished, as shown by ________.*
*The system's success stopped registering that damage when ________.*

If any of the three blanks can't be filled with something specific and checkable, the honest answer is: not yet. Keep watching, keep asking, but don't call it cancrity on a hunch. The form is strict on purpose. It's what stands between a real diagnosis and a complaint dressed up in a scarier word, and it's the same form whether you're running it on a company, a movement, a government office, an open-source project, or your own household.

It also works as a standing question, not just a one-time test. A board meeting, a staff review, a family budget conversation: any of them can carry a version of the three lines as a recurring item. What are we depending on? Is it holding up? Would we notice from our own numbers if it weren't? Most institutions have plenty of meetings about output and almost none about substrate. Adding the three lines to an existing conversation costs almost nothing and catches the drift early, while there's still something left to redirect.

Run it on a rival and it will feel sharp and satisfying. Run it on your own institution, your own team, your own side, and it becomes something more useful than satisfying: an early warning, arriving while there's still a substrate left to save.

## Close

A system consuming what it depends on has a problem heading toward it, whether or not anyone inside can see it yet. But substrates rarely go quietly. They complain. Turnover spikes. Customers leave, and say why on the way out. Maintainers post that they're stepping down. Regulators write reports. Someone inside the organization notices the drift and says so out loud. The complaint is a form of correction trying to reach the system that needs it.

The truly durable pathologies are not the ones that consume their substrate fastest. They're the ones that find a way to make the complaining stop: not by fixing what the complaint is about, but by silencing the channel it travels through. The employer hears from the exit interview, right up until nobody bothers filling it out, or nobody reads it, or filling it out honestly becomes the kind of thing that gets a person noticed for the wrong reasons.

That's a different mechanism from anything in this chapter, and a harder one, because a system that has learned to do it can keep eating its substrate for a long time without anyone, including itself, finding out. Cancrity, left alone, eventually gets loud enough to diagnose. A system that has also disabled its capacity to hear the diagnosis is a different problem. Chapter 6 takes it up.

## Sources

- Alabama Cooperative Extension System, "History and Use of Kudzu in the Southeastern United States" (1876 Centennial Exposition introduction; Soil Conservation Service promotion in the 1930s; federal recommendation withdrawn 1953; USDA weed designation 1972). https://www.aces.edu/blog/topics/forestry-wildlife/the-history-and-use-of-kudzu-in-the-southeastern-united-states/
- USDA Forest Service / Alabama Cooperative Extension, "Kudzu: History, Uses, and Control" (ANR-65). https://research.fs.usda.gov/download/treesearch/2341.pdf
- The Nature Conservancy, "Kudzu: The Invasive Vine that Ate the South." https://www.nature.org/en-us/about-us/where-we-work/united-states/indiana/stories-in-indiana/kudzu-invasive-species/