Capitalism is not the enemy. Extraction is. Capitalism rewards initiative, risk, competence, and ownership. It is the only system that has ever lifted large numbers of people out of poverty at scale. The problem is not that some people get rich. The problem is how. There are two ways to make money: build something people are better off for having, or find a way to profit from making people weaker. Both are legal. Only one deserves respect.
Value creation looks like this: a product that saves a customer time, a service that fixes a real problem, a company that pays for a skill the market needed. The customer is better off after the transaction than before it. Extraction looks different: profit that depends on the customer staying confused, dependent, or unable to leave. The transaction makes the customer worse off, but the exchange still clears, because the seller has engineered it that way.
The clearest proof of this line is what happens when a company gets caught doing it deliberately. Purdue Pharma marketed OxyContin as having a low risk of addiction while internal sales data showed the opposite; company documents released in litigation showed executives tracking which doctors were prescribing in volumes consistent with abuse, and continuing to target them. The Sackler family extracted an estimated 10 billion dollars from the company between 2008 and 2017, years that overlapped with the sharpest rise in opioid overdose deaths in American history. That is not a business model failure. It is value extraction working exactly as designed: profit that rises precisely as the customer’s condition worsens.
Social media offers a cleaner, more universal case, because almost everyone has felt it firsthand. In 2017, Sean Parker, Facebook’s founding president, said publicly that the platform was built to exploit “a vulnerability in human psychology,” engineering a small dopamine hit from likes and comments to keep users coming back. Tristan Harris, a former Google design ethicist, has documented how notification design, infinite scroll, and variable reward schedules borrow directly from slot machine psychology: unpredictable reward is more addictive than predictable reward, a finding B.F. Skinner demonstrated with pigeons in the 1950s and that app designers now apply to humans at a scale of billions. The user is not the customer in this model. The user is the product, and attention is the extraction.
Naval Ravikant’s frame for this is precise: wealth is created, money is earned by trade, and status is granted by loyalty and relationships. Confusing these categories is where extraction hides. A company can look wealthy while it is only extracting money, because money can move from someone else’s pocket without anything new being built. Real wealth creation shows up as something that exists that did not exist before: a cure, a tool, a service, knowledge, capability. Extraction shows up as a balance sheet moving while nothing in the world got better. The two can look identical in a quarterly earnings report. They are opposite in what they cost the world.
This is not a call to regulate capitalism out of existence, and it is not an argument that profit is suspect. Profit earned by building something someone freely chooses, understanding what they are choosing, is the cleanest signal a market produces. The test is simple and does not require a committee: does the customer’s understanding and freedom to walk away increase or decrease the longer the relationship continues? A gym that gets you fitter passes. A slot machine app that gets you more compulsive fails. A loan that funds a business passes. A payday loan structured so the fees exceed the principal within weeks fails. The mechanism, not the industry, is what to judge.
The reason this distinction matters beyond ethics: institutions built around extraction do not stay contained to one industry. An institution funded to solve a problem develops a structural interest in that problem never fully going away, because solving it ends the funding. This is close to what Robert Michels called the iron law of oligarchy and what public choice economists later formalized: organizations optimize for their own persistence, and persistence is easiest to justify when the problem they exist to solve stays unsolved. Nobody has to be a villain for this to run. Ordinary self-interest, keeping the job, growing the budget, does most of the work on its own. It runs alongside, not instead of, the cases where someone is a deliberate villain. Purdue’s sales force was not confused about what they were doing.
The fix is not to distrust markets. It is to demand the one thing extraction always avoids: legibility. Extraction depends on the customer not fully understanding the deal, whether that is a 40-page loan agreement, a terms of service nobody reads, or a feed engineered to feel spontaneous while being fully deliberate. Value creation survives disclosure, because a customer who fully understands the exchange still wants it. That is the whole test. If a business could not survive its customers understanding exactly what it is doing to them, it has already told you which kind of business it is.
I write about AI, systems, markets, and the incentives hiding underneath them. Get the next essay by email.
Unsubscribe any time. Privacy.