In Praise of the Dilettante

The word means “one who delights”

The word dilettante comes from the Italian dilettare, “to delight.” For the first century of its English usage the term was a compliment.^[1] It described someone who pursued a subject for the sheer pleasure of pursuing it, free of the obligations and conformities that mark the professional. To call someone a dilettante in eighteenth-century Europe was to honor their willingness to follow curiosity wherever it led.

Then the credential economy arrived. Universities professionalized. Disciplines specialized. By the twentieth century, dilettante meant amateur in the worst sense — a person who skimmed without mastery, who collected interests like trophies, who never committed. The pejorative has stuck so firmly that we now use the word almost exclusively as an insult.

The pejorative is not entirely wrong. The unreformed dilettante does fail in the ways the critics describe. But the pejorative has crowded out something we badly need to remember. Every polymath begins as a dilettante. Curiosity that ranges across domains, that hops from one delight to the next, that refuses the early invitation to specialize is the only soil in which the polymath grows.

For founders, this matters more than it does for anyone else. Building a venture requires the ability to recognize structural patterns from one domain that solve problems in another. The classic name for this move is bisociation, a term Arthur Koestler coined to describe the moment when a concept from one mental matrix collides productively with a concept from a different one.^[2] Most bisociations are empty. Some are real. The only way to find the real ones is to allow yourself to make many of the empty ones first. The dilettante is the person who has not yet lost permission to do this.

The dilettante phase of polymathic founders

The pattern is everywhere once you start looking.

Steve Jobs dropped out of Reed College and spent his time auditing a calligraphy class.^[3] The class had no obvious application to anything Jobs was going to do for a living. He attended because the typography fascinated him. A decade later, when the original Macintosh shipped with multiple typefaces and proportionally spaced fonts — a design choice that no other personal computer of the era had made — Jobs traced the decision directly to the calligraphy class. The dilettante interest had become the design instinct that distinguished the product.

Patrick Collison, the co-founder of Stripe, is widely known among the people who know him for reading habits that range absurdly far from payments infrastructure. Economic history. The history of science. Eighteenth-century political philosophy. Architectural theory. Materials science. He reads these subjects because he is interested in them, not because they obviously bear on his company. They have nonetheless bent his thinking about Stripe in ways that show up in the company's actual decisions — its long-term orientation, its bets on developing-country payments rails, its willingness to publish essays on topics like accelerated scientific progress that have no commercial payoff. Collison treats his curiosity as an asset to invest in, not a distraction to apologize for.

Charlie Munger, the longtime vice chair of Berkshire Hathaway, spent fifty years building what he called a “latticework of mental models” drawn from psychology, biology, physics, history, mathematics, and a dozen other disciplines.^[4] Munger's argument was simple: a person trained in only one discipline mistakes their tools for the world, while a person who carries models from many disciplines can route around the failures of any one of them. The investment record that resulted from Munger's partnership with Warren Buffett is one of the most extraordinary in capital-markets history. Munger never tired of pointing out that his polymathic reading habit was the actual source of the edge.

Demis Hassabis, the co-founder of DeepMind, began as a chess prodigy, became a professional video-game designer in his teens, returned to university to study cognitive neuroscience, completed a doctorate on hippocampal memory, and then founded the AI laboratory that would eventually solve protein folding with AlphaFold.^[5] Each of his earlier careers seemed like a detour. None of them was. The neuroscience informed the AI architecture. Game design taught him how to build evaluation systems. Chess sharpened the pattern-recognition habits at the heart of his engineering instincts. Hassabis's career reads as a sequence of dilettante delights that turned out to be a single coherent project the whole time.

Frances Arnold earned her undergraduate degree in mechanical and aerospace engineering at Princeton and her doctorate in chemical engineering at Berkeley.^[6] She might have spent a successful career within her field. Instead she imported evolutionary biology into chemical engineering and built a body of work — what she calls directed evolution — that uses the principles of natural selection to engineer new proteins for fuels, pharmaceuticals, and agriculture. The work won her the 2018 Nobel Prize in Chemistry. It also launched her into entrepreneurship: she co-founded Gevo to commercialize biofuels and Provivi to develop pheromone-based crop protection. Arnold's career demonstrates that the integrative dimension of polymathy can be domain-creating rather than just connection-finding. She built a new field at the intersection of evolutionary biology and chemical engineering.

Kiran Mazumdar-Shaw trained as a brewmaster in Australia in the mid-1970s — a degree no Indian brewery wanted to hire because she was a woman.^[7] Rather than retraining, she carried what she had learned about enzyme technology in fermentation across into a different domain entirely. In 1978, at age twenty-five, she founded Biocon in her garage in Bangalore with seed capital of ten thousand rupees (roughly a thousand dollars at the time), in joint venture with the Irish entrepreneur Leslie Auchincloss. The initial work applied brewing-science principles to industrial enzyme production for food and textile manufacturing. Over the decades that followed, the company traversed enzymes, then biopharmaceuticals, then biosimilars and insulin — becoming India's largest biotechnology company and a global manufacturer of oncology and diabetes drugs. Mazumdar-Shaw's path is a particularly clean example of bisociative entrepreneurship: the move that eventually connected enzyme techniques optimized for beer to insulin biosimilars for millions of patients began with a question that could only have occurred to someone whose training crossed two disciplines that did not yet talk to each other. The polymathic discipline here lives in the move itself — connecting a brewing degree to industrial biotechnology, a Bangalore garage to a global health-equity company.

Stewart Brand, who founded the Whole Earth Catalog in 1968 and has continued to instigate consequential institutions ever since (the Long Now Foundation, the Global Business Network, the Revive & Restore project), began with a degree in biology from Stanford and an apprenticeship in counterculture.^[8] He went on to write about computers, about climate, about ecological restoration, about cathedral-scale time horizons. His refusal to settle in any one domain is the reason his contributions span so many. Brand's career is a demonstration that the dilettante's restlessness, given a long enough horizon, becomes a body of work that no single specialist could have produced.

The list could be longer. Benjamin Franklin, who was a printer, a postmaster, a diplomat, an inventor, a scientist, and a statesman across overlapping decades. Hypatia of Alexandria, the late-antique mathematician and philosopher whose commentaries on conic sections shaped how the ideas of Apollonius were transmitted forward. Ibn Sina (Avicenna), the Persian physician, philosopher, astronomer, and poet whose Canon of Medicine shaped medical practice across centuries of both Islamic and European traditions. Zhang Heng, the Han-dynasty Chinese polymath who served as imperial astronomer, mapped the celestial sphere, invented the first known seismoscope, and wrote poetry of sufficient quality that scholars still translate it. Maria Sibylla Merian, the seventeenth-century naturalist and artist who traveled with her younger daughter to Suriname at the age of fifty-two — an unprecedented expedition for a woman without institutional patronage — and produced the illustrated volumes that helped found modern entomology.^[9] Hedy Lamarr, who acted in films at MGM and in her spare time co-invented (with composer George Antheil) the frequency-hopping spread-spectrum technology whose principles underlie modern wireless communications.^[10]

None of these people began with a clean specialty. They began as dilettantes — sometimes for years, sometimes for decades — and the dilettantism turned out to be the soil out of which their later achievements grew.

The pattern is consistent across the ages and across geographies. Cross-domain wandering, often without permission, becomes the soil out of which something the world later calls a polymathic career grows.

From delight to discipline: the polymathic orientation

Dilettantism is the beginning. It is not the end. The dilettante who never grows past the dilettante phase becomes the figure the pejorative imagines: shallow, distractable, perpetually in motion without arriving. The question is how the transformation happens — how curiosity-across-domains becomes capability-across-domains. How does someone become a polymath rather than remaining a hobbyist?

The most thoughtful contemporary answer to this question comes from Michael Araki, a researcher at UNSW Sydney whose work on polymathy has restored the concept's intellectual credibility after decades of neglect.^[11] Araki's framework distinguishes three dimensions that any polymathic development must engage:

Breadth. The horizontal dimension. The range of fields, contexts, and experiences a person has engaged with seriously enough to develop genuine acquaintance. Engaged contact rather than casual exposure. Breadth is what the dilettante phase contributes.

Depth. The vertical dimension. The sustained engagement with at least some of those fields to a level where the person has gone past the introductory texts and into the questions the experts argue about. Depth is what specialization, in its non-pathological form, is for. The polymath needs at least some depth somewhere; depth is what gives the breadth purchase.

Integration. The relational dimension. The capacity to draw connections across the fields the person has come to know — to see that a structure described in one domain has a counterpart in another, and to use the counterpart productively. Integration is what makes polymathy more than a collection. Without integration, breadth and depth remain in separate compartments and the person is a generalist with hobbies rather than a polymath in any meaningful sense.

Araki's argument is that these three dimensions are not interchangeable. They are not substitutes for one another. The polymath needs all three. A person with breadth and no depth is the hobbyist the pejorative imagines. A person with depth and no breadth is the conventional specialist, talented within their lane and helpless outside it. The most common failure mode, counterintuitively, is the person with breadth and depth but no integration — they have done the work but cannot exploit it. Integration is what makes the prior investments compound.

Araki has developed an instrument, the Polymathic Orientation Scale, that measures these dimensions as a personal disposition rather than as a credentialing artifact. The scale captures whether individuals value, prefer, and habitually pursue breadth, depth, and integration as a coordinated practice. The instrument's existence matters because it makes polymathy something more than a metaphor. It can be measured. It varies across people. It predicts outcomes that conventional measures of intelligence or expertise do not.

The framework also makes the dilettante-to-polymath progression operationally clear. The dilettante has the breadth dimension activated and the other two latent. The transition to polymathy is the development of depth in at least some of those domains, plus the cultivation of integrative habits that turn the breadth into a network of usable connections rather than a list of interests. The shape of the progression is a layering, not a sequence of stages. Breadth comes first because that is what curiosity produces when it has not been disciplined. Depth and integration come later, and they have to be cultivated deliberately because the institutional incentives do not generally reward them.

Mike Araki's body of emerging work is essential reading for anyone who wants to take polymathic development seriously. His scholarship — across articles, working papers, and his recent doctoral research — has restored polymathy to a place of serious empirical study, developing instruments that measure it, mapping the behaviors that constitute it, and articulating the developmental theory that explains how a polymathic disposition matures over a career. The lab's research program owes his work an intellectual debt that this essay cannot fully discharge.

The cyborg architecture and polymathic AI

The polymathic-development project has always been hard for one reason that has nothing to do with the framework: time. A single human life is short. The breadth-depth-integration triad is a budget problem before it is a discipline problem. You can read widely or you can read deeply, and most days you cannot do both. The polymaths who have managed it are unusual people in part because they have unusual amounts of cognitive bandwidth to spend.

Something is changing in 2026 that will eventually be recognized as one of the more consequential developments in the long history of how humans organize their cognitive labor. The arrival of high-capability AI partners changes the budget problem. The AI carries the breadth dimension at machine scale. The human partner gets to spend more of their cognitive budget on the dimensions where human contribution is irreplaceable: depth, and the integration that connects the depths to one another.

The arrival of high-capability AI partners changes the budget problem. The AI carries the breadth dimension at machine scale; the human partner gets to spend more cognitive budget on depth and the integration that connects the depths to one another.

The AI does not make the human polymathic on its own. The human-AI partnership forms a composite cognitive architecture in which the polymathic dimensions are distributed differently than they would be in a single human practitioner. The AI partner reads everything. The human partner directs the AI's attention toward the connections worth testing, evaluates which of those connections survive scrutiny, and develops the depth in the particular threads that matter. The integration happens at the boundary — in the dialogue between the human's directional instinct and the AI's recombinatory capacity.

This is what we mean by cyborg entrepreneurship in the literal sense. The cyborg founder is no futurist scenario. It is a description of how the work of building a venture is already changing for people who have begun to use AI partners seriously. The founder who reads voraciously and asks the AI to surface the structural connections across what they have read is doing polymathic work at a cadence no prior generation of founders could have matched. The founder who treats the AI as a search-engine replacement and stops there is missing the architecture.

The implication for the dilettante phase is surprising and good news. The cost of breadth has dropped. A curious founder in 2026 can engage seriously with many more domains than a curious founder in 2010 could have, because the AI partner makes the initial breadth-building radically faster and the integration-building substantially more tractable. The dilettantism phase is shorter and richer than it used to be. The polymathic-development trajectory is more accessible than it has ever been.

This does not mean depth is optional. Depth is the dimension the human partner contributes uniquely, and a founder who tries to skip depth in favor of AI-assisted breadth will end up building a venture that does not survive contact with anyone who has the depth. The polymathic discipline that Araki names — breadth and depth and integration together — still applies. The AI changes the mix. It does not change the requirement.

The AI partner does not solve any of the problems the dilettante's critics worried about. The unreformed dilettante can use AI to skim more shallowly, just as a serious polymath can use AI to read more deeply. The instrument is dual-use. What it does, when used well, is widen the productive range of the curious. The discipline that converts dilettantism into polymathy is the same one it has always been. The cyborg architecture just means the discipline buys more.

Trust the curiosity. The discipline comes later.

If you are a founder, or a researcher, or someone who is trying to build something that does not yet exist, the practical message is the simplest. Do not be afraid of the dilettante phase. Do not apologize for the cross-domain curiosity that draws you toward subjects that are not obviously related to what you are trying to do. The instincts that lead you to follow those threads are the same instincts that, given time and discipline, become the polymathic capabilities the most consequential founders have used to build their ventures.

The professional environment around you will discourage this type of work. The institutional incentives will reward specialization. The well-meaning advisors will tell you to focus. Some of that advice is right and some of it is not, and the only way to tell the difference is to develop your own judgment about which connections matter. That judgment cannot be developed by staying in your lane. It can only be developed by following enough cross-domain threads that you begin to see, in your own experience, which kinds of connections compound and which kinds dissipate.

The discipline comes later. It always comes later. You do not begin a polymathic career by deciding to be a polymath; you begin by being a dilettante who keeps showing up. Over time, the showing-up develops into depth in some of the threads. Over more time, the depth begins to connect with the other threads in ways that surprise you. Over a longer time still, the connections themselves become the work — the body of contributions that distinguishes a polymath from a generalist.

Mike Araki's framework is the map. The cyborg architecture is the new vehicle. The dilettante's delight is the engine.

Begin where the curiosity is. Trust the instinct that pulls you toward what does not yet make obvious sense. Follow it long enough to discover whether it leads anywhere. Some of it will not. Some of it will. The only way to find out which is which is to have the courage to dilettante in the first place.