Eat a Lot, Work Little: Cancer Cells and a Certain Kind of People
About 7 minutes to read
Key points
- Cancer isn't about "growing fast." It's about tearing up the contract: no longer accepting any constraint, while still living off the system's resources for free. Cancer cells are even inefficient: one unit of sugar yields less than a tenth of the energy a normal cell gets, and it hides this by grabbing wildly more. One of the ways hospitals find tumors is to look for the most wasteful spot.
- Academic fraud lines up almost point for point: the brake of peer review was never connected; fake papers get twice the citations of real ones, pulling funding and students' time toward a dead end; retracted papers keep getting cited and have seeped into the reviews doctors read; and the few people who check for fraud get legal threats.
- But the analogy stops there. Cancer is mainly caused by the environment, not the cell. Fraudsters and honest people use the same language. They aren't invaders; they're a product selected by the rules of this game. Calling someone a "malignant tumor" isn't too harsh. It's too soft: it downgrades a moral problem into a biological inevitability, which lets them off the hook.
- So what's really worth borrowing is the treatment approach: prevention (dismantling the paper-counting system) works far better than excision; screening relies on "abnormal metabolism"; the key to immunotherapy is releasing the brake — protecting the people who check for fraud, not adding more review.
The main text follows
An unflattering analogy
Are the people who commit fraud in cancer research the cancer of society?
It's an uncomfortable thing to say, but if you treat it as an analogy worth checking carefully, it gets a lot more interesting.
Cancer isn't "growing fast"
First, a possible misunderstanding to clear up: cancer is not defined by cells growing fast.
Human bone marrow makes two hundred billion blood cells a day, the gut lining replaces itself entirely every few days, hair keeps growing. These cells all grow faster than most tumors, but they aren't cancer.
The human body is a community of tens of trillions of cells, held together by an unwritten contract: each cell gives up the freedom to "grow as big as I want" in exchange for the benefits of division of labor. Liver cells don't grab the muscle's nutrients. Skin cells don't run a blood vessel to themselves without asking.
Cancer is one side tearing up that contract. It's not growing fast. It's no longer listening to anyone, while still living off the nutrients, oxygen, and space the system provides for free. The difference is like "a hardworking employee" versus "someone who occupies a desk, does no work, and files the biggest expenses": the difference isn't the workload, it's whether he still counts as a member of the organization.
The most counterintuitive thing about cancer cells: they're actually wasteful
This is the most counterintuitive part of the whole thing.
A normal cell processes one unit of sugar using a "refined" method that squeezes out about thirty units of energy. A cancer cell, with oxygen clearly available, uses a crude method instead, getting only two units of energy per unit of sugar, more than ten times less efficient.
So how does it survive? By grabbing. A cancer cell can absorb sugar tens of times faster than the normal cells around it. It uses "grabbing more" to cover up "using it badly."
A German named Warburg discovered this a hundred years ago, and one of the most common tumor tests in hospitals today is built on it: inject a radioactive-labeled glucose, then scan the whole body. Wherever it eats the most, the screen lights up brightest. The way we find a tumor is by finding the most wasteful spot.
"Absorbs energy, does nothing." That's literally it.
And academia?
In early 2026, a French-Australian team published a study in The BMJ: they trained a language model to identify articles whose textual features closely resembled the products of already-retracted "paper mills," then ran it over 2.64 million cancer research papers from 1999 to 2024. The result: 9.87% were flagged — about two hundred sixty thousand papers[1].
The trend is steeper still: around one in a hundred at the start of the century, approaching one in six by 2022. And these papers don't only appear in small journals nobody reads; the proportion is rising in the top 10% of journals by impact factor too.
The so-called "paper mill" is a real factory, a business. It mass-produces papers that look plausible, then sells authorship. One study counted nearly nineteen thousand such ads, with the price of an author slot ranging from tens of dollars to several thousand[2].
Meanwhile, the number of retracted papers worldwide rose from just over a thousand in 2013 to more than ten thousand in 2023[3], with a cumulative record now past sixty thousand.
Point by point
The brake is broken. Papers go through peer review before publication, and that's supposed to be the brake. But reviewers are unpaid, anonymous, and overloaded. All they can check is "does your logic hold up," not whether the raw data is real. This isn't the brake being broken through. It's the brake never being connected.
What happens when the brake isn't connected? Fake papers get in, and once they're in, they start taking things from others.
Stealing the blood supply. Tumors have a nasty trick: they induce new blood vessels to grow around them, diverting blood flow that should go to normal tissue toward themselves. Academia's "blood flow" is citations — citations decide journal rankings, decide who gets included in reviews, decide which hypothesis the next batch of PhD students spends three years on. And an analysis reported by Nature in 2026 found that suspicious paper-mill cancer papers receive twice the citations of real papers[4]. They also cite each other heavily, which conveniently lifts the rankings of the journals that published them.
So fake papers don't just waste their own share of resources. They actively pull the surrounding resources toward a dead end.
Metastasis. Retraction is supposed to be like cutting out the lesion. But studies find that retracted papers keep getting cited, and often the people citing them have no idea they were retracted. Worse, they may already have entered systematic reviews — the things doctors and policymakers actually read. The primary tumor is cut out, but the metastases keep growing.
Something like this happened in Alzheimer's research: a key 2006 paper in Nature, cited more than two thousand times, was found in 2022 by journalists and image experts to have problems with its figures, and that hypothesis had steered the whole field for over a decade[5].
The security guard gets sued. The people actually checking for fraud in this system are very few: Elisabeth Bik, who compares paper images by eye, or Jennifer Byrne, who checks whether gene sequences actually match up. What they often get after finding problems is legal threats and online harassment.
In immunology terms: this is immunosuppression, one of a tumor's most effective survival tactics.
But the analogy has to stop here
This analogy has reached its own limits too.
First, cancer cells have no intentions. People do.
This looks like it makes fraudsters guiltier. It does the opposite. No one puts a gene mutation on trial. It isn't evil, it just happened. Calling people "cancer" downgrades a moral problem into a biological inevitability, and that lets them off the hook. So I'm not arguing that "these people are a tumor on society." Not because it's too harsh, but because it's too soft.
Second, and most important: what gives people cancer is mostly the environment, not the cells themselves.
The genome of a tumor cell is almost identical to the normal cell right next to it. The differences sit in a handful of spots, and most of those changes come from the environment: tobacco, radiation, long-term inflammation. You wouldn't blame the cell and ignore the smoking.
Same logic here. Fraudsters and honest people use the same language, the same chart formats, the same stock phrases. That's the basic reason fraud is so hard to spot. They aren't invaders who broke in from outside. They're what this rule set selects for.
When a system uses paper counts to decide promotions, journal rankings to decide who gets money, and assessments every few years to decide whether a lab lives or dies, it is applying selection pressure. Where the pressure is strongest, the most extreme things grow.
In the BMJ study above, 36% of the flagged papers came from Chinese institutions. The authors themselves warned that the model may have learned certain writing habits rather than actual fraud features, and that the number should be read with care[1]. But even if it holds only partly, it points not to the character of any nation but to a system that turns academic output into currency: papers directly buy titles, degrees, and bonuses. That design isn't unique to China. Some places just push it further.
Third, "tumor on society" has a very dirty history.
Calling a group of people "a tumor that needs to be cut out" was standard rhetoric in several twentieth-century purges. This isn't about word taboos. It's a real warning: once this analogy slides from "analyzing the system" to "naming the enemy," it turns from a tool into a weapon.
We should use it to see the system clearly, not to convict people.
How a doctor would treat it
Since the analogy has held this far, cancer treatment offers some ideas we can borrow.
Prevention works best, and it's the easiest to skip. In the fight against lung cancer, tobacco control has saved more lives than every targeted drug combined. The academic version of tobacco control is dismantling the paper-counting system: look less at quantity, look more at a few representative works. Let "repeating someone else's experiment" and "I tried for three years and showed this path is blocked" count as real results, eligible for funding and worth putting on a CV. Technically this is not hard at all. The hard part is that it requires reviewers to actually spend time reading, and the current metrics system was invented precisely to save that time.
Screening should target the "abnormal metabolism." Finding a tumor relies on the very feature of "it eats aggressively." The academic equivalents already exist: image duplication detection, gene sequence consistency checks, statistical distribution screening, and now machine learning scans. The BMJ study was essentially a scan of the entire cancer research field.
Immunotherapy works by releasing the brakes, not adding weapons. One of the most important advances against cancer in recent decades wasn't a more toxic drug. It was the discovery that tumors put the brakes on immune cells, and that releasing those brakes is enough. The parallel is clear: give fraud investigators legal protection and institutional support. Turn academic fraud detection from "a personal hobby adventure" into a proper job with a position, a budget, and liability protection. Right now this immune system runs on the spare time of a few dozen people, against tens of thousands of retractions a year. That ratio says everything. (If you know Geng, what he does should be formally brought into the education system as a position, not left as one person's adventure.)
Finally, accept something uncomfortable. No multicellular organism can completely eliminate cancer, because cell replication inevitably makes errors, and errors inevitably land sometimes on growth-control genes. Likewise, wherever there is competition and reward, someone will game it. The goal isn't zero incidence. It's pushing it back to a manageable range, and not letting the patient die of the treatment first. Endless review, mountains of compliance forms, assuming every data point is fake until proven otherwise, these kill healthy tissue too. Chemotherapy has a narrow dose window. So does governance.
I'd rather treat this piece as a memo to myself than a verdict on others. The lesson from cancer research these years isn't "some people are bad." It's this:
When the way we keep score has been decoupled from real value for too long, the system grows its own tumor, and it grows where things look healthiest.
For a Buddhist angle, see Claiming Attainment Without Attainment.
References
- Scancar B, Byrne JA, Causeur D, Barnett AG. Machine learning based screening of potential paper mill publications in cancer research. BMJ. 2026. doi:10.1136/bmj-2025-087581
- Richardson RAK, Hong SS, Byrne JA, Stoeger T, Amaral LAN. The entities enabling scientific fraud at scale are large, resilient, and growing rapidly. PNAS. 2025;122(32):e2420092122. doi:10.1073/pnas.2420092122
- Van Noorden R. More than 10,000 research papers were retracted in 2023 — a new record. Nature. 2023;624:479–481. doi:10.1038/d41586-023-03974-8
- Basu M. Paper mill cancer studies get double the number of citations as genuine papers. Nature (News). 2026. doi:10.1038/d41586-026-01908-8
- Piller C. Blots on a field? Science. 2022;377(6604):358–363. doi:10.1126/science.add9993
For further reading
- The classic review of cancer's "hallmark capabilities," which this piece's comparison framework comes from: Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144(5):646–674.
- Geng's channel https://space.bilibili.com/1732848825/
- The Economist report https://www.economist.com/china/2026/06/08/a-dropout-turned-influencer-shakes-up-chinese-science