How to Do Great Work Max Effgen, September 17, 2026September 10, 2026 Paul Graham’s important essay from 2023 assumes you can choose your field, your hours, and your risks. Most people can’t. Here’s what still works when you don’t have any of that. Paul Graham is a programmer, essayist, and co-founder of Y Combinator, and the essays he’s published since the early 2000s are among the most widely read working documents in technology — short, concrete, and unusually willing to say the uncomfortable part of a true idea. His essay “How to Do Great Work” (July 2023) is the late-career version of that habit: a long attempt to extract what ambitious work has in common across fields. You don’t have to want to found a company, or agree with every claim, to get value from it — the useful parts are the mechanisms: how curiosity selects problems, how attention compounds, how fashion and fear distort taste. Graham’s thesis is simple: great work comes from following excited curiosity far enough to reach a real gap in a field you’re suited for. Curiosity is both engine and rudder. The rest of the essay is a set of working habits — guess and start, learn to the frontier, notice the cracks, explore the strange ones, protect long stretches of attention, finish, and stay earnest. That recipe is good. It’s also incomplete for most people who will read it. Graham assumes high ambition and unusual autonomy, and the recipe shrinks once your constraints are a job, a team, or a regulated field. Worth saying once — then worth asking what still applies. What Graham Is Actually Saying The essay has no chapter headings, but it has a sequence. First, choose the work by doing, not by planning. Young people especially can’t know the flavor of a field until they’ve worked in it. Guess. Try. Optimize for interestingness — importance often arrives later. Strange taste is an advantage, because conventional taste puts you where the crowd already is. Second, learn until the edge of the field stops looking smooth. From a distance, knowledge looks finished; up close it’s fractal and full of gaps. Great work often has a tincture of strangeness — an idea other people discounted, a question treated as settled, a rule nobody checked. Third, work in a way that compounds. Starting is harder than continuing, and interruptions are expensive. There are two kinds of procrastination: the daily kind, which you can trick your way through, and the yearly kind, which disguises itself as busyness on the wrong project. Consistency looks modest and isn’t — a page a day is a book. Leave room for undirected thought once the problem is loaded into the mind. Finish. The best work often shows up in the pass you thought was the last one. Fourth, stay earnest. Trying to be the best is often easier than trying to be merely good. Affectation, fashion, fear, and other people’s wishes distort the signal — make what you yourself would want. Find colleagues who raise the quality of the questions. Treat morale as infrastructure: optimism, health, a real audience, and distance from people who quietly drain the will to continue. The unglamorous formula at the end is ability, interest, effort, and luck. Curiosity is the part you can steer. The Caveat That Changes the Recipe Graham writes as if the main constraint is courage. For founders, independent researchers, writers, and anyone who can change fields without asking permission, that’s often true. For everyone else, the constraint set is different. A job means someone else owns the calendar and a large share of the problem selection. A team means your best idea has to survive other people’s priorities, risk tolerance, and definition of done. A regulated field means the interesting gap may be real and still unworkable, because the evidence bar, the approval path, or the liability isn’t shaped like a startup experiment. Import the essay without that filter and you get a familiar kind of private despair — the sense that you’re failing at great work because you haven’t abandoned the job, the team, or the license that pays for your life. That’s a misreading. It’s a high-autonomy recipe. The useful move is to keep the parts that survive reduced autonomy and drop the parts that require a blank page. What Still Transfers Protect the loop, even if you can’t protect the field. You may not get to choose the industry. You can still run Graham’s four steps inside the domain you already occupy — learn to the local frontier, notice the gaps your institution pretends are settled, explore the ones you can touch without asking for a new title. Most organizations are full of unowned problems sitting just next to the official work. Those are the realistic version of Graham’s “fractal buds.” Optimize for interestingness inside the constraint, not instead of it. “Follow your curiosity” doesn’t require quitting. It requires using whatever discretion you have to take the more interesting of two acceptable tasks. Over a year, that compounds. People who wait for a perfectly aligned mission statement never start; people who take the most live problem available inside the current job often discover the next field from the inside. Name the two calendars. Graham’s advice about long, uninterrupted blocks is more valuable in a job than in a cabin — the job is where the blocks get destroyed. No one does great work in thirty-minute fragments, whether the work is code, analysis, writing, or clinical reasoning. The transferable practice isn’t “reject meetings.” It’s treating one or two protected blocks a week as non-negotiable production time, and being honest about the cost when they disappear. Separate daily procrastination from yearly procrastination. Inside a company, yearly procrastination looks respectable — alignment decks, stakeholder reviews, projects that are moving without being the project you believe matters. Graham’s test still works: are you working on the thing you most want to work on, or on a substitute that’s easier to justify? You won’t always get to switch. You should still know which one you’re doing. Iterate in public enough to get a signal. “Start small and make successive versions” survives regulation and hierarchy better than “bet the company on a strange idea.” Version one can be a memo, a prototype, a pilot with one team, a measurement nobody asked for. The point is to get the idea out of your head and into contact with reality — institutions punish grand proposals and tend to tolerate small working artifacts. Keep the earnestness. Drop the lone-genius posture. Affectation is as common inside companies as it is in scenes — status language, process theater, and borrowed frameworks are the corporate version of pretending. The transferable standard is intellectual honesty: say what you actually think is true, at the highest resolution the room can use. Colleagues still matter more than most people admit; in a team, the equivalent of Graham’s “find people who raise the quality of the questions” is the most important hiring and alliance decision you control. Treat morale as a production input. This part of the essay is underrated and unusually portable. Sleep, health, a small audience that cares, and distance from draining people aren’t lifestyle accessories — they determine whether the next interesting problem gets any of your remaining attention. In constrained environments, the residue of the day is smaller, which makes the morale infrastructure more important, not less. What Does Not Transfer Cleanly Don’t expect to switch fields the week a new one becomes more exciting. Some people can; most can’t without a runway. Don’t expect planning to be optional. Graham is right that you can’t plan a discovery any more than natural selection can plan a species — but you can and must plan a clinical trial, a product launch inside a licensed category, or a budget cycle. Use planning for the parts of the work that are describable, and curiosity for the parts that aren’t. Don’t confuse “try to be the best” with “do it alone.” In a team, being the best often means making other people more effective at the same problem — a different skill from writing an essay or proving a theorem, and one Graham’s essay doesn’t spend much time on. Don’t treat luck as a personality trait. Graham says to make yourself a target for luck by reading, meeting people, and trying things. That still works — it just works inside a smaller radius when your days are already allocated. A Reduced Recipe Strip the essay down to what survives a job, a team, or a regulated field, and it looks like this: Pick the most interesting live problem you’re allowed to stay with. Get closer to the edge of it than the people who only visit. Look for the crack that official process has stopped seeing. Protect enough unbroken time to do something about it. Ship small versions until reality talks back. Stay earnest. Avoid theater. Keep the few people who make the work sharper. Guard the physical and psychological conditions that let you continue. That’s less romantic than “do great work.” It’s also closer to how most good work actually gets done. Graham’s essay is still worth reading whole. The curiosity test is better than most career advice. The warning about fashion and fear is durable. The compounding math of consistency holds in any setting that rewards finished artifacts. The mistake is reading a high-autonomy manual as a verdict on people who don’t have a blank calendar. Ability, interest, effort, and luck still decide the outcome. Autonomy changes how much of the interest you get to aim. Use the part of the recipe that still aims. Avanti. Measure what matters. ...