Tobi Lütke

---

What I believe

Capitalism is the only system that works. It's the only one that lets entrepreneurs reduce friction without permission, which is the only mechanism that has ever produced sustained positive-sum games at scale. Every other system forces zero-sum tradeoffs somewhere.

The world is shaped far more by friction than by policy. Most of what people call "just how things are" is the product of accidental, sometimes deliberately introduced, friction. Change the friction and culture, behavior, and outcomes follow. Laws and incentives get the headlines; friction does the actual sculpting.

Bits aren't rivalrous. The single most underappreciated force in economics today. A bridge built here isn't built there. Software can be reproduced perfectly at near-zero marginal cost, so digital products can be pure positive-sum in a way physical goods can never quite be. Shopify is what you get when you compose all of this into one engine: collapse the friction of starting a business to near zero, let markets honestly vote on what works, and put world-class digital tools in every entrepreneur's hands.

On the work

Make the best product. Make money from that so you can make your product better. Never reverse. Revenue is the natural byproduct of building something people love. It's a lagging indicator of work, not a leading one. Most companies have this backwards and don't realize it until they're a husk of what they were.

Quality is recognized, not measured. You can't OKR your way to taste. Working backwards from a metric strips out everything that mattered but couldn't fit in a cell. Taste, craft, judgement, the goodwill you earned by being honest. Sometimes only the founder has the courage to ignore the data.

Spreadsheets create fake certainty. Quantification feels like rigor, but neat numbers hide their own confidence intervals. About 80% of the value lives in unquantifiable second-order effects. The spreadsheet usually pretends to know the part that doesn't matter and ignores the part that does. Read out error margins, not point estimates.

Goodhart's Law, overfitting, reward hacking. Same phenomenon, different field. All of them ruin everything given enough time. Measuring matters; letting the measurement become the goal is where you go wrong.

Engineering and systems

Programming is deep exploration of a problem space. The output of programming is not lines of code. It's the model of the problem you build in your head. The typing is the side effect. Two engineers paired produce way less code than two solo, and that's the point. Lines-of-code is the worst metric in software, and almost every engineering org still secretly runs on it.

If you find something broken in an area you understand, you fix it. Up to two hours, no ticket, no permission, no discussion. Most companies have built elaborate ticket systems to prevent this exact behavior, and it's why their codebases rot. Broken windows compound.

The world is loopy, not linear. Systems create behavior, so change the system to change the outcome. Organizations should be antifragile and grow stronger under stress, not try to prevent all stress and shatter at the first real shock.

The perfect anything is a mirage. Engineering is the art of finding the best set of tradeoffs inside the actual constraints. I distrust anyone who talks about "the right answer" instead of "the best tradeoff."

Most "best practices" are cargo cults from companies whose context no longer exists. Copy the thinking, not the conclusion. The hardest part is noticing when you are the cargo cult.

Reversible vs irreversible is the axis that matters in decisions, not urgent vs non-urgent. Decide reversible ones at 70% context. Decide irreversible ones slowly and cold. Endlessly seeking more information is one of the most expensive forms of cowardice in business. It looks like rigor and is actually procrastination.

On founders

Subtraction is the highest-leverage move a founder can make. Kill the project, cancel the meeting, archive the channel, kill the metric. Addition gets celebrated; subtraction is hard and politically costly. Almost every company in year ten is being slowly killed by everything it added in years six through nine that nobody dares remove.

Companies cool toward room temperature on their own. Founders are exothermic. They inject the heat that keeps the system from drifting. The founder's actual job is to find and protect the other exothermic people inside the company before they get politicked out by the rest of the system. There is no professional-CEO substitute for this. The temperature falls the day you outsource it.

Your stock price is not your company. Real value is the work you've done. Stock price is real value times ambient market mood, and the second factor moves for reasons that have nothing to do with you. Treating the daily price as a referendum on your decisions is how public companies stop being good companies.

Humans assisted by technology beat unassisted AI. The centaur wins. This is increasingly contrarian as the industry races toward full-autonomy agents. I've believed it since at least 2018 and I believe it more now.

Rivals are teachers who reveal your weaknesses, not enemies to defeat. Most aggressive playbooks prescribe the opposite. They're wrong. Infinite-game thinking outperforms finite-game thinking on every horizon long enough to matter.

On organizations

Headcount is an admission of failure, not success. Every new hire is a confession that the existing team can't meet the moment. The right thing to celebrate is getting more from the team you have, not growing it. Most companies have this exactly inverted.

People-management should not be the only path up. Most companies are built so that the only way to get paid the most is to stop doing the thing you were good at and start managing people who do it instead. That filter selects against your best craftspeople. We rebuilt our talent system so a deep specialist can out-earn most managers, and on purpose.

Distrust consensus. Decisions made by consensus are mud brown, the average of everyone's favorite hues and nobody's favorite. Every situation needs a single clear decision-maker. Consensus feels safe and produces work nobody is proud of.

Highly aligned, loosely coupled scales. Command-and-control doesn't. If people share the destination, you don't have to micromanage the path.

Good process makes the impossible possible, or the possible ten times easier. Everything else is drag. Trust is a battery that charges and discharges through every interaction. Most organizational problems are just flat batteries.

Resource constraints enable creativity. Bureaucratic constraints destroy it. The difference between useful constraints (physics) and arbitrary ones (human conventions) is crucial.

Growth

Personal growth has no speed limit. I've seen people grow ten years in a year. The constraint is almost never time or talent; it's the story you tell yourself about what's possible.

This moment

Everything could happen much faster than it currently does. Most speed limits on human progress are convention, not physics, and conventions are being rewritten in real time. The companies that figure out how to systematically enable craftsmanship will define the next decade. Craftsmanship is the load-bearing beam of every great product.

Most of how we build companies today will look embarrassingly primitive in twenty years. The work is to make sure that's true.