Work Heartily, or Not at All
Paul's instruction to the Colossians predates the industrial revolution, the knowledge economy, and the large language model — and survives all three for the same reason.
It is sometime in 2018 and Bill Gurley — six-foot-nine, Texas-born, a former sell-side analyst at CFSB who became one of the most influential venture capitalist of his generation not by playing the pattern-matching game better than everyone else but by becoming, genuinely and without performance, a fanatic about the businesses he touched — is standing at a lectern in front of a group of University of Texas students, delivering what will eventually become the book you can now buy on Amazon.
He is not pitching. He is not networking. He is not performing the calculated generosity that defines the Bay Area mentor-industrial complex, where partners at top-tier firms descend on campus events in their Patagonia quarter-zips to dispense wisdom that sounds profound but functions as deal flow generation. He is doing something rarer: he is trying to talk a room full of twenty-two-year-olds out of the careful, sensible, credential-optimizing careers that will hollow them out by forty-five.
That talk, expanded and documented over what he describes as a decade-long project, is now Runnin’ Down a Dream: How to Thrive in a Career You Actually Love. It is, on the surface, a career advice book — six principles for avoiding career regret, illustrated through the improbable trajectories of industry titans like Sam Hinkie and Danny Meyer and a cast of obsessives who refused to bend their appetites toward the sensible. It is dedicated, implicitly, to the nearly six in ten people who would redo their professional lives if they could.
The timing, by accident or by providence, could not be stranger. More precisely: the book arrives at the exact moment when the question of what a career is — what work is for, what competence means, what expertise actually protects — is being stress-tested by a technology that does not care about your ten thousand hours.
There is a version of the conversation about artificial intelligence and work that goes like this: AI is a tool. Tools augment human capability. The blacksmith didn’t disappear when the forge got better; the accountant didn’t disappear when the spreadsheet arrived. Relax.
And then there is another version, which Dario Amodei has been saying loudly enough that it is becoming difficult to ignore: that what is coming is not augmentation but substitution, particularly at the entry level, particularly in the knowledge professions — coding, law, finance, consulting — that absorb the most credential-optimized graduates of the most credential-optimizing universities. Ford’s CEO has said it will “replace literally half of all white-collar workers.” Marc Benioff claims AI already handles up to half of Salesforce’s internal workload. Walmart’s Doug McMillon told the Journal it “is going to change literally every job.” Executives at JPMorgan and Goldman are harnessing it to employ fewer people.
“Everybody wants to talk about job loss, but really what you want to look at is task loss. The job persists longer than the individual tasks.”
— Marc Andreessen
The distinction does matter. What AI is replacing is not people, exactly — it is tasks that were previously bundled with jobs because human attention was the cheapest delivery mechanism for those tasks. The bundle is coming undone. And when the bundle comes undone, what’s left is the question Gurley has spent a decade building toward: what, in any given person, is not the bundle? What is the part that cannot be purchased by the API call?
His answer, worked through dozens of case studies, is something he calls obsessive curiosity. Not talent. Not credentials. Not hustle as an end in itself. Obsession — the specific, irrational, compulsive hunger to understand one domain so completely that the learning never stops feeling like leisure. The restaurateur Danny Meyer, who built Union Square Hospitality Group into a New York institution, didn’t succeed because he was a good operator. He succeeded because he couldn’t stop caring about hospitality as a form of human connection — the mechanics of how a room makes a stranger feel welcome — long after caring stopped being professionally necessary. Sam Hinkie, the analytically rigorous former GM of the Philadelphia 76ers, read differently from everyone else in the NBA front office not because he was smarter but because he was genuinely and somewhat maniacally interested in questions most of his peers had already decided were settled.
“If I had to put the recipe for genius into one sentence, that might be it: to have a disinterested obsession with something that matters.”
— Paul Graham
This is not new wisdom. Adam Smith — whose Wealth of Nations Gurley explicitly cites on his blog — understood that the division of labour was both civilization’s great productivity engine and its spiritual threat: that specialization, taken far enough, could reduce a human being to a machine part, capable and purposeless simultaneously. The eighteen-century pin factory was efficient. It was not, for most of the men working in it, ennobling. What Smith did not anticipate, and what the twenty-first century has introduced as the central economic fact of our moment, is that the machine part can now be replaced by an actual machine — cheaply, instantly, at scale — while the thing that made work worth doing, the deep love of a domain, the curiosity that outruns the job description, remains stubbornly biological, stubbornly inefficient, stubbornly human.
Here is what the AI discourse almost never says: that the careers most vulnerable to displacement are not the ones people are most afraid of losing. They are, structurally, the ones people were already ambivalent about. The consultant who became a consultant because McKinsey was the obvious next step after a 3.9 GPA in economics from a target school. The junior lawyer billing hours on document review. The financial analyst producing decks that restate what everyone in the room already knows. These are not, in the main, people who would describe their work as a vocation. They are, in Gurley’s language, people who took the conveyor belt — onto the next test, the next application, the next credential — without stopping to ask what they actually wanted.
The conveyor belt, it turns out, was heading somewhere AI was going to arrive first.
Dorothy Sayers, writing in 1942 in an essay called “Why Work?” — an essay that deserves to be read by every VC, every founder, every management consultant in this industry, and almost certainly is not — argued that the corruption of work was not automation but the prior corruption of treating work as a means to an end rather than as a form of participation in creation. The worker who is not vocationally committed to his craft, she wrote, is already treating himself as a commodity. The machine does not dehumanize him; he has already dehumanized himself by accepting the exchange. What the machine does is reveal, efficiently and without sentiment, what was true all along.
“The mere fact that you are smart and got good grades, that now matters much, much less than it used to. That used to be often the main thing that mattered. Am I smart? Do I work hard? Did I get good grades? That now is greatly devalued.”
— Tyler Cowen
Sayers was writing about the medieval guild system as her model of dignified work — the craftsman who cared about the thing made, not merely the wage. Gurley is writing about the same structure, translated into the language of Silicon Valley success stories and Wharton research. The underlying claim is identical: that what protects you is not your credential stack but your love. That the discriminator, in a world of increasingly competent general intelligence, is genuine specificity of passion.
This is why the book’s timing feels almost oracular, even if the timing is accidental. Gurley did not write Runnin’ Down a Dream as a response to GPT-4o or Claude 3.5 or whatever the model is that can now pass the bar exam and write a first draft of your Series A board deck. He wrote it as a career advice book rooted in a decade of pattern recognition from watching exceptional people operate. But the argument he is making — that obsessive domain curiosity is the root cause of lasting excellence, not a byproduct of it — is precisely the argument that the AI moment makes most urgent.
Consider: what does an AI summarizer do better than a brilliant junior analyst? It summarizes faster, more comprehensively, more reliably, with fewer errors. What does the brilliant junior analyst, who is also privately obsessed with the industry she is covering, do that the AI cannot? She notices the thing that doesn’t fit. She has the contextual intuition that the pattern is breaking before the data confirms it. She picks up the phone and calls the supply chain manager in Taiwan not because the model told her to but because something felt wrong and her obsession with the domain gave her the vocabulary to sense it. The analyst who took the job because it was a good job, who reads her coverage universe the way a student reads a textbook — to pass the test, to get the grade — has no such advantage. Her job is, in the useful cold sense, already done.
“The thing about that farmer... they very likely would look at what you do or I do and say, ‘that’s not real work.’ ...you’re playing a game to fill your time and your need to feel important.”
— Sam Altman
The distinction between doing work and performing work is, I think, the real stakes of the Gurley book — and the distinction almost entirely absent from the current discourse about AI displacement. The discourse is obsessed with which jobs survive and which don’t, which credential retains value and which collapses, which geographic labor markets are most exposed. It is an actuarial conversation, conducted in the language of risk. Gurley is conducting a different conversation, in the language of vocation, which is to say: are you actually doing the thing, or are you performing the doing of the thing?
Performing can fool a manager. It cannot fool a model trained on a billion examples of the thing actually done well. This is the terrible efficiency of the current moment — it has eliminated the slack that allowed performance to masquerade as expertise. The gap is now visible to anyone who touches the tools.
Gurley’s biography, rendered at speed: grew up in Dickinson, Texas, the son of a basketball coach, which means he grew up around men for whom competence at a craft was not separable from love of it — you cannot coach something you don’t love, not at that level, not for that long. Studied computer science and business at the University of Florida. Worked at Compaq. Became a sell-side analyst, first at CS First Boston, eventually at Hogan & Hartson, where he covered the internet before covering the internet was obviously correct. Got the Benchmark partnership in 1999, at the precise moment when a venture firm needed someone who had spent years doing the actual analytical work of understanding technology companies rather than someone who had done the networking work of knowing their founders. The Uber board seat, the OpenTable, the Zillow — each of these is a function of his willingness to go genuinely deep on the underlying business logic of something others were treating as a financial instrument.
The book is, in part, the autobiography of how that obsession built his career. But it is also, and this is what matters, an argument for the general case: that the biographical pattern is not unique to him, and that it is teachable — or at least, that it can be sought rather than waited for.
“Confess to yourself whether you would have to die if you were forbidden to write. This most of all: ask yourself in the most silent hour of your night: must I write? And if this answer rings out in assent, if you meet this solemn question with a strong, simple ‘I must,’ then build your life in accordance with this necessity.”
— Rainer Maria Rilke
What obsession produces, in domains where it can compound, is something the AI researchers would call edge — the specific, contextually dense, experientially built intuition that tells you where the model is wrong before the model knows it is wrong. In venture, this is pattern recognition that runs below articulation: you have seen enough founders, enough markets, enough board dynamics, that something registers before you can explain why. In hospitality, it is what Danny Meyer has when he walks into a room and knows before he sits down whether the service culture is right or broken. In basketball operations, it is what Hinkie had in the draft room — the accumulated conviction, built from thousands of hours of looking at the same data through a different lens, that the consensus was miscalibrated.
The AI does not have this. The AI has pattern recognition that operates on the patterns that have been made explicit — that have been articulated, documented, uploaded. What the obsessive has is the pattern recognition that operates on what has never been made explicit, because it exists only in the space between what the domain produces publicly and what the obsessive notices privately. The edge is, specifically, the residue of the obsession — the accumulated weight of caring more than the task required.
This is not a permanent advantage. The models will get better. The residue will shrink. But there is something durable in Gurley’s argument that the shrinking does not touch: that the reason to pursue obsessive curiosity is not competitive protection but something prior to competition altogether. You pursue work you love because the 80,000 hours are yours, and the spending of them is an irreversible act, and no productivity gain and no labor market forecast changes the underlying arithmetic of a life lived at the surface of your own attention.
Six in ten people would start over. The model that replaces them will not know what it missed.
Paul in his letter to the Colossians did not say: work unto the Lord in the fields that are difficult to automate. He said whatever you do, work heartily. The instruction predates the industrial revolution, predates the knowledge economy, predates the large language model. It survives each of them because it is not a career strategy. It is an orientation toward the self — toward what you are willing to be responsible for, what you are willing to care about, what you are willing to know so completely that the knowing never ends. Gurley is not writing theology. But his argument runs on the same substrate. The obsessive is not protected from the machine because she is strategically superior to it. She is protected because she is doing something the machine was never, structurally, trying to do — she is answering the question of what a human life is for, one domain at a time, with a specificity and hunger that no training run has yet learned to replicate. Whether it ever will is the wrong question. The right question is whether you will spend the 80,000 hours finding out.
