Alt Binaries

Man Stabs Roommate Over Thermostat; Ponzi Scheme Exposed After Seventeen Years — April 2026

In September 2025 a man in Santa Clara, California stabbed his roommate over the thermostat. The two had been arguing about the setting for months, tensions compounding alongside an eviction dispute, until one morning the man kicked open a door and started stabbing. The roommate survived only because a police officer arrived in time to shoot the attacker dead. The responding officer later said that if he’d been a few seconds later, the roommate would not have made it. A few seconds. A thermostat. A life and a death, separated by the amount of time it takes to put on your shoes.

The thermostat killing is useful not because it is unusual but because it is the reductio of something extremely common, which is the crime of not being able to stop yourself. The man did not plan the stabbing. He did not spend weeks considering the optimal thermostat-defence strategy, sourcing weapons, establishing alibis. He was angry. There was a knife. Two facts converged in a single bad second and a person was nearly dead. You could replay that second a thousand times with a thousand slight variations (the knife is in a drawer instead of on the counter, the roommate says “fine, whatever” instead of escalating, the man had eaten lunch and his blood sugar was half a point higher) and in most of those replays nobody gets hurt. The attack was, in every sense that matters, an accident of timing grafted onto a failure of self-regulation, and the combination is responsible for a quantity of human suffering so vast that if you could visualise it as a physical structure it would be architecturally implausible.

Start with the numbers, because the numbers are startling and because starting with the numbers earns the right to say something less quantifiable later.

The Bureau of Justice Statistics has been surveying prison inmates in the United States for decades, and one of the things it consistently finds is that a remarkable proportion of violent offenders were drunk or on drugs at the time of the offence. The figures vary by study and by substance, but the range is something like forty to sixty percent for alcohol alone, higher when you add other substances. This does not mean alcohol causes violence. It means alcohol degrades the specific cognitive function (inhibitory control, the prefrontal cortex’s ability to override an impulse with a plan) that stands between “I want to hit this person” and “I am hitting this person.” The desire was already there. The capacity to not act on it was removed. This is a pharmacological fact about ethanol, not a moral judgment about drinkers, and the distinction matters because it tells you something crucial about the architecture of the act: the intention to harm was not the bottleneck. The bottleneck was the brake. Damage the brake (with alcohol, with a traumatic brain injury, with any of the clinical conditions that psychiatry keeps renaming every generation as children age into adults) and the vehicle does not stop accelerating. It was always accelerating. You just couldn’t tell because the brake was working.

And then there are the crimes of sheer pettiness, which are impulsive crimes’ purest expression because the stakes are so low that no rational calculus could possibly justify the risk. A man in Houston shot another man dead over a parking space at a Walmart in 2022. There is an entire literature on homicides provoked by what criminologists delicately call “trivial altercations,” which is their way of saying someone died because someone else looked at someone funny, or jumped a queue, or played music too loud, or (as in a case from November 2017 in Carlisle, Arkansas, that I find myself thinking about more than is probably healthy) changed the channel during a football match. The channel. Not even the thermostat, which at least has a physiological argument. The channel.

These are not evil people, or more precisely, their evil is not what matters about them. What matters is that between the moment the channel changed and the moment the violence began, there was a window (maybe two seconds, maybe five) in which a different neurological configuration would have produced a different outcome. The desire to do harm existed in that window. It exists in everyone’s window, or nearly everyone’s, in the same way the desire to eat the entire cake exists in the window between seeing the cake and remembering that you are forty and your trousers already don’t fit. The difference between the person who eats the cake and the person who doesn’t is not that the first person wants the cake more. It’s that the second person has a functioning brake. And the difference between the person who stabs someone over the channel and the person who mutters “for fuck’s sake” and goes to another room is structurally identical. The impulse is shared. The brake is not.

Here is where it gets uncomfortable, because if you take this framework seriously (if you accept that the impulsive offender is a person whose brakes failed rather than a person whose engine is evil) then several things follow that you may not want to accept. First: deterrence, in the classical sense of “announce a punishment and people will calculate the costs,” is nearly useless against impulsive crime. You cannot deter a neurological failure with a threat. The threat assumes a decision-making process that is, by definition, offline at the moment it is needed. This is why the death penalty’s effect on impulsive homicide rates is, according to the best available evidence, approximately zero, which is a number that should embarrass a lot of confident people on both sides of that debate but somehow never does. Second: you can spot impulsive offenders relatively easily, because they can’t help it. They are the ones with the disciplinary records, the short fuses, the history of escalations that everyone around them saw coming. The thermostat attacker’s other roommates had been watching the tension build for months. They are never surprised.

Third, and most uncomfortably: when they are not in the grip of the impulse, these people are as full of good intentions as anyone else. The man who stabbed his roommate over the thermostat presumably had, at other moments, passed the salt, held a door, felt guilt. The remorse is real because the self that feels the remorse is real. It is the same self. It is not a different, better self that emerges to apologise for the bad one. It is the same person operating with the brakes engaged, looking back at what happened when they weren’t, and the horror is genuine. This does not mean you don’t remove them from situations where they can hurt people. You do. But you do it the way you take the car keys from someone with seizures, not the way you punish a strategist for a strategy.

Now.

The other thing.

In 2008, Bernard Madoff’s Ponzi scheme collapsed, revealing that approximately sixty-five billion dollars in reported client balances did not exist and had never existed. Madoff had been running the scheme for at least seventeen years, possibly longer, producing fabricated account statements with the consistency of a man who has never once in his life stabbed anyone over a thermostat. He was calm. He was patient. He showed up to work, generated fake returns, recruited new investors to pay old investors, and went home. He did this every day, for decades, with the disciplined focus of a medieval monk copying manuscripts, except instead of manuscripts he was copying the future out of other people’s retirement accounts.

Madoff is the archetype of the second thing, and the second thing is this: some people do bad things not because they can’t stop themselves but because they’ve decided to. Calmly. With forethought. Over breakfast.

The calculated offender is not impulsive. The calculated offender has a perfectly functional prefrontal cortex, magnificent inhibitory control, an executive function suite that would make a cognitive psychologist weep with professional admiration. The brakes work beautifully. They are simply not applied to the project in question, because the project in question is the point. The calculated offender has examined the costs, the benefits, the probabilities of detection, the likely sentence if detected, the quality of available legal representation, the specific vulnerabilities of the victim, and has concluded that the expected value of the crime exceeds the expected value of not committing it. This is, if you squint, rational actor theory as applied to felony. The Chicago school of economics would be proud, if the Chicago school of economics were willing to acknowledge that its models describe criminals as accurately as they describe entrepreneurs, which it mostly isn’t, because that would be embarrassing.

Elizabeth Holmes did not wake up one morning and impulsively lie about blood-testing technology. She did it over years, adjusting the lie as circumstances demanded, recruiting allies, neutralising critics, with a consistency that suggests either genuine sociopathy or a deeply considered decision that the performance was worth the risk. The crime was not a failure of self-regulation. It was a success of self-regulation applied to an antisocial project.

And this changes everything about how you respond to it. Because unlike the impulsive offender, the calculated offender responds to incentives. Deterrence works on people who are calculating, precisely because they are calculating. If you make the expected cost of fraud higher than the expected benefit (through detection, enforcement, sentencing, asset forfeiture) you change the equation. You do not need to change the person. You do not need to fix their neurological architecture or provide them with anger management or hope that this time the brakes will hold. The brakes are fine. You need to change the terrain so that a person with perfectly functional brakes chooses to apply them.

But follow that logic one step further, because it goes somewhere you might not want it to go.

It is tempting to say that the calculated offender has an evil engine. The impulsive offender has a good engine with bad brakes; the calculated offender has perfectly functional brakes attached to a machine that wants to run people over. Neat. Satisfying. And almost certainly wrong, or at least wrong about the ratio.

Because if deterrence works on the calculated offender, what you are really saying is that the calculated offender would not commit the crime if the costs were high enough. Which means the crime is not an expression of some deep wickedness in their character. It is an expression of opportunity. They saw a gap where the rewards exceeded the risks, and they walked through it. Remove the gap and they go back to being a law-abiding citizen, not because their soul has changed but because the arithmetic no longer works.

And now ask yourself: how many people are in that same position, with the same arithmetic, who simply have never encountered a gap wide enough to walk through? Your colleague who would never steal from the company, but who has also never had the combination of access, desperation, and plausible deniability that would make stealing from the company a rational bet. Your neighbour who would never commit fraud, but who has also never been offered the specific confluence of information asymmetry and institutional negligence that makes fraud easy. How many of the honest people you know are honest because they are good, and how many are honest because they have never been presented with a sufficiently attractive alternative?

This is, if you think about it for more than a few seconds, the entire premise of criminal law. We do not build elaborate systems of detection, prosecution, and punishment because we believe people are naturally good and the few bad apples need sorting. We build them because we believe (correctly, if the recidivism data and the cross-cultural crime statistics and the entire history of what happens when institutions collapse are any guide) that a meaningful proportion of the population will do harmful things if the consequences are removed. The system exists because we have already answered the question about humanity’s moral baseline, and the answer was not flattering. Every lock on every door is a small confession about what we expect from each other.

The reciprocal is also true and also uncomfortable: the calculated offender is much harder to identify before the fact. The impulsive offender advertises. The calculated offender camouflages. Madoff was a former chairman of the NASDAQ. Holmes sat on the boards of prestigious organisations. The entire premise of calculated crime is that the criminal presents a surface indistinguishable from legitimate behaviour, because maintaining that surface is part of the plan. You cannot spot them by their affect. You cannot spot them by their history. You can only spot them by their accounting, and checking the accounting requires suspecting the accounting, and suspecting the accounting of someone who seems perfectly reputable requires a kind of institutional paranoia that most social structures are not built to sustain.

These two categories of wrongdoing (the impulsive and the calculated) are not just different in degree. They are different in kind, and the difference runs all the way down to what you think a person is.

The impulsive model says: a person is a set of desires managed by a set of controls, and when the controls fail the desires express themselves destructively. The person is not their worst moment. They are the system that, most of the time, prevents the worst moment. To help them is to strengthen the controls (medication, therapy, structure, removal from triggering environments). The project is engineering.

The calculated model says: a person is a set of preferences pursued through a set of strategies, and when the strategies include harming others the harm is intentional. The person is exactly their worst moment, because the worst moment is the one they planned. To deter them is to change the incentive structure (enforcement, transparency, consequences). The project is architecture.

And the discomfort (the real discomfort, the one that keeps criminal justice systems oscillating between rehabilitation and retribution every thirty years like a pendulum that will never reach equilibrium) is that both of these are true, they apply to different people doing superficially similar things, and we have no reliable method of telling which one we’re looking at until after the damage is done.

The ancient Greeks had a word for it, or rather two words that they used in ways that should have told us everything but didn’t because we weren’t listening in the right language. Akrasia (weakness of will, acting against your own better judgment) was Aristotle’s term for the problem of the impulsive offender, and he considered it genuinely puzzling, because if you know the good and still do the bad, what exactly is happening in the space between the knowing and the doing? Socrates thought akrasia was impossible. If you truly knew the good, you would do it. That you didn’t meant you didn’t really know. Aristotle disagreed. He thought you could know and still fail, that knowledge and action were connected by a mechanism that could jam, and that the jam was where the philosophy lived. He was, it turns out, describing the prefrontal cortex two thousand three hundred years before anyone had a word for it.

The calculated offender, by contrast, was not akratic. The calculated offender knew perfectly well what they were doing and did it anyway, not because a mechanism jammed but because the mechanism was working as intended and the intention was bad. The Greeks called this kakia (vice, wickedness) and treated it as a settled character trait rather than a momentary failure. The akratic man was redeemable because his better judgment still existed, just temporarily overwhelmed. The vicious man was a different problem entirely, because his judgment had incorporated the harm into its calculus and found the calculus acceptable.

What neither Aristotle nor the modern criminal justice system has adequately addressed is that you can be looking at the same person and not know which one you are seeing. The man who embezzled from his employer: was he a person whose financial desperation overwhelmed his moral brakes, a momentary akratic who, given stability, would never offend again? Or was he a person who examined his options, identified embezzlement as the best available strategy, and executed it with the calm precision of someone ordering lunch? The difference between these two possibilities determines everything (whether rehabilitation will work, whether deterrence will work, whether incarceration is containment or punishment, whether forgiveness is wisdom or naivety) and in most cases you cannot tell. The act looks the same from the outside. The inner architecture is invisible.

There are, of course, other reasons people do terrible things. Systemic ones. The guard at the camp who was following orders. The middle manager who signed off on the cost-benefit analysis that valued a human life at less than a recall would cost. The true believer whose delusions of grandeur, or of divine mission, or of historical destiny, have elevated the harm past the point where “impulse” or “self-interest” can describe what is happening. Hannah Arendt wrote a whole book about how the most catastrophic evil of the twentieth century was perpetrated by people who were neither impulsive nor (in the ordinary sense) selfish, but simply obedient, and the banality she identified is real and important and deserves its own accounting.

But those are not what you are looking at when you are looking at a single person who has done a bad thing and asking: why? The systemic failures produce their own grammar. The delusions produce theirs. What akrasia and kakia describe is the territory of individual moral failure, the kind that the person standing before you is responsible for in a way that the cog in the machine and the zealot with the vision are (perhaps) not. These are the two failure modes that we, as individuals, and the systems we build (courts, prisons, parole boards, HR departments, the whole trembling apparatus of institutional judgment) are actually trying to stamp out. The impulsive act we want to prevent. The calculated act we want to deter. And the premise of every system of justice ever devised is that if you can get the prevention and the deterrence right, you have addressed most of what one person can do to another person on purpose or through failure.

There is no single mystery of why people do bad things. There are many reasons and most of them are boring and most of them are old. But these two (the broken brake and the open ledger) are the ones that matter most when you are standing in front of a single person who has done a single bad thing and trying to decide what to do about it, because they are the ones that live inside individual will, the ones the system is actually built to address. And even between these two you cannot always tell. The whole wretched insight of the lock on the door is that the second mechanism (the calculated one, the rational one, the one that responds to incentives) may be running quietly in almost everyone, waiting for conditions that never quite arrive, which means the person who made a plan may not be a different kind of creature from you. The man before you has done a terrible thing. Was he overtaken, or was he decided? Was the act a wave that crashed over him, or a house he built? You are going to have to respond. Your response will shape his life and the lives of everyone affected. And you are standing in the gap between two entirely different kinds of wrongdoing, and the gap is dark, and neither your compassion nor your severity will illuminate it, because the light you need is the light that would let you see the inside of another person’s will, and that light does not exist and has never existed and will not exist, and every system of justice ever devised is a monument to the attempt to act without it.

So you do what everyone has always done. You look at the man. You look at what he did. You make your best guess. You get it wrong about as often as you get it right, and you build the system anyway, because the alternative is not building one, and you’ve seen what that looks like too.

The thermostat. The spreadsheet. The gap between them. We keep trying.

Toothpaste

GALIT: You cannot seriously be telling me you comparison-shop for toothpaste.

BARTUK: I can seriously be telling you that, yes.

GALIT: Toothpaste. The thing you use for two minutes and spit into a drain.

BARTUK: The thing I use twice a day, every day, for the rest of my life. That is seven hundred and thirty uses per year. At roughly four grams per use, a standard tube lasts about thirty-four uses, which means I go through twenty-one tubes annually. Over the next fifty years, conservatively, that is a thousand tubes. The price differential between the brand you grab without looking and the brand I select after reading three ingredient panels is about two stelmarks per tube. Two thousand stelmarks. You’re telling me you wouldn’t spend four minutes to save two thousand stelmarks.

GALIT: I am telling you that the four minutes you spend reading ingredient panels is four minutes you are not spending being alive.

BARTUK: Being alive is expensive. That’s why I read the panels.

GALIT: You know what I did last week? I walked into a shop, I saw a toothpaste that had a drawing of a mint leaf on it, and I thought, “I like mint.” I bought it. I have been brushing my teeth with it. It’s fine. I have spent zero seconds thinking about it since, which means I am already winning.

BARTUK: Winning what, exactly?

GALIT: The time war. Every second you spend optimising something that doesn’t matter is a second you’ve donated to the void. You’re not saving money, Bartuk. You’re spending your life in four-minute increments and getting two stelmarks back each time. That’s a terrible exchange rate.

BARTUK: And every stelmark you waste because you couldn’t be bothered to look at a shelf for ten seconds is a stelmark someone else earned by looking. You haven’t escaped the system. You’ve just decided to be bad at it.

GALIT: I haven’t decided to be bad at it. I’ve decided it’s not worth being good at. There’s a difference.

BARTUK: Is there? Because from where I’m standing, someone who refuses to engage with a system they’re trapped in anyway is not making a philosophical statement. They’re just losing.

GALIT: I bought the wrong bread last Tuesday.

BARTUK: What?

GALIT: I bought the wrong bread. I wanted the dark rye, I grabbed the light rye, I didn’t notice until I got home. You know what I did? I ate the light rye. It was fine. I didn’t go back to the shop. I didn’t recalculate my bread strategy. I just ate the bread and moved on with my life, and the amount of suffering this caused me was zero.

BARTUK: The amount of suffering it caused you was one meal of bread you didn’t want.

GALIT: Which I didn’t notice! Because it turns out the gap between “the bread I wanted” and “the bread I got” is vanishingly small when you are not the kind of person who catalogues bread outcomes.

BARTUK: You are describing carelessness and calling it freedom.

GALIT: You are describing obsession and calling it competence.

BARTUK: Galit, I have watched you punch a wall because a Council vote went the wrong way. You threw a chair at Aleksander last month because he used the word “incremental.” You are not a person who doesn’t care about outcomes. You are a person who cares enormously about outcomes, but only when the outcome involves something that makes you feel righteous. The toothpaste doesn’t make you feel anything, so you ignore it. The Council vote makes you furious, so you break furniture. That’s not philosophy. That’s just impulse with a dress code.

GALIT: That is completely unfair.

BARTUK: It is completely accurate.

GALIT: The chair was already broken.

BARTUK: The chair was not already broken. I paid for that chair. Fourteen stelmarks. Which I know because I checked the price before I bought it, because that is what a person does when they are spending resources they earned.

GALIT: Fine. Fine. You want to know the real difference between us? You would never punch a wall. You would never throw a chair. You would sit in that chair, calmly, while the Council voted to do something monstrous, and you would calculate the cost of objecting versus the cost of compliance, and you would comply, and you would go home, and you would brush your teeth with your optimally selected toothpaste, and you would sleep perfectly well. And in the morning, you would tell yourself you were being practical.

BARTUK: Yes.

GALIT: That doesn’t bother you?

BARTUK: What bothers me is waste. What you’re describing is efficiency. I don’t break things I can’t fix. I don’t spend capital (emotional, political, financial) on gestures that accomplish nothing. When I object, I object in ways that have a chance of working, which means I object rarely, and I prepare first, and I pick my moment. You object constantly, at maximum volume, with no preparation, and then you are confused when nothing changes except the furniture budget.

GALIT: The wall I punched was in Hadria’s quarters. She didn’t mind.

BARTUK: Hadria minded.

GALIT: Hadria said she didn’t mind.

BARTUK: Hadria says a lot of things. I know what the repair cost. Want to guess?

GALIT: I absolutely do not want to guess. I want you to understand that there are things more important than what they cost. That sometimes the right response to an injustice is immediate, is loud, is physical, and the fact that it breaks something is not a bug, it’s the whole point. The breaking is the message. “This matters enough to break things over.” You never send that message. You send invoices.

BARTUK: Invoices get paid. Messages get ignored.

GALIT: Not always.

BARTUK: Often enough.

GALIT: (long pause) You are genuinely the most depressing person I know, and I know Abbas.

BARTUK: Abbas is depressing for theological reasons. I am depressing because I’m right. There’s a difference. You claimed earlier there was a difference between deciding to be bad at something and not caring. I’ll extend you the same courtesy. There is a difference between being depressing and being correct. I am both. You are neither. But your toothpaste is fine.

GALIT: My toothpaste is great, actually. It has a mint leaf on it.

BARTUK: It has sodium lauryl sulphate and a mint leaf on it. Mine doesn’t. Mine cost less. But you’re happy, and I have two thousand stelmarks. Shall we call it even?

GALIT: We will never be even. That’s the whole problem with people like you. You think everything resolves into a ledger.

BARTUK: Everything does resolve into a ledger. The question is whether you’re the one holding the pen.