Craft, Vision, Purpose

A framework for jobs, ventures, and the thing you would do unpaid.

Why I wrote this

I have been thinking about this for a while.

What is a job for, what is a venture for, and what does either of them have to do with the thing you would do even if nobody paid you. I found I could not think about any of these cleanly, because every time I tried to think about the next job I was also, without meaning to, trying to think about my whole life. Should the job be close to what I care about? Should I chase equity, or chase a big exit and do the real thing later, or take the stable thing and do the real thing on the side? Every question opened into three more.

Then I wrote it down, and once it was on paper most of it collapsed into something small. This is that small thing. It is long and a little raw, because I would rather have the whole of it in one place than a tidy version that leaves half of it out.

I am writing it as someone who loves technology, loves building companies, and has a purpose that is neither of those things. If that combination sounds like you, this might save you some time.

The three loves

What I had been tangling together turned out to be three separate things.

The first is love for the craft, which is the making itself. For me this comes in two flavours that I care about roughly equally: the craft of technology, where you write systems that actually work, and the craft of entrepreneurship, where you build a company that actually works. They are different skills, and a person can love both.

The second is love for the vision, which is a picture of a future you want to see built. Flying cars, rockets, self-driving cars, machines that think.² A vision is about the world, not about you, and you can love one deeply without it being what your life is for.

The third is purpose, which is what your life is actually for. Mine is finding answers to existential questions through the intersection of technology, science, and philosophy, and consciousness is the first question of them all.¹ I will keep doing it unpaid because it came to me; I did not choose it. That is the thing about purpose: you can choose one, or one can come to you. But that is for a different article.

The mistake I kept making was treating these three as rungs on a ladder, with craft at the bottom, vision above it, purpose at the top, and every career move expected to climb. It is not a ladder. They are three independent axes, and almost any combination can live inside one person, one job, or one company.

                 CRAFT          VISION         PURPOSE
                 (the making)   (the world)    (your life)

  Job            required       bonus          bonus
  Venture        required       required       optional
  Purpose work   instrument     incidental     required

Once I saw it as a grid rather than a ladder, most of the overthinking simply stopped. The rest of this article is me filling in the cells, and checking them against real people.

Real people in each cell

I wanted proof that these combinations exist in the wild before I trusted the grid, so I went looking, and the examples that convinced me are below.

Craft alone, inside a job

Anders Hejlsberg built Turbo Pascal at Borland and then C# and TypeScript at Microsoft, which is forty years of language design as an employee while the companies and their visions came and went around him. Ken Thompson did Unix at Bell Labs and Go at Google half a century apart, with the same craft carrying him through both. Jeff Dean has been doing systems at Google for a quarter of a century, and the products are almost incidental to the work.³

The entrepreneurial version of this cell surprised me, because I used to assume the craft of building companies required founding one. Sheryl Sandberg scaled Google's advertising business and then Facebook as its chief operating officer, and in both cases the vision belonged to someone else while the company-building craft was entirely hers. Tim Cook ran operations at Compaq, IBM, and Apple, and the vision was always Jobs's; Cook's love was for running the machine well. Eric Schmidt was brought in to turn Google into a company, which is a craft practised on someone else's dream.

And there is a cell where both crafts live in one job. Greg Brockman joined Stripe as its first engineer and became its CTO, loving the technology and the business-building in the same week, all of it on someone else's vision, years before he went and started his own thing.⁴ Werner Vogels has been Amazon's CTO for two decades, deep systems craft fused with commercial judgement, never a founder. This is the founding engineer's cell, and I will come back to it when I talk about jobs.

Vision and craft together, in a venture, with no purpose required

Jeff Bezos ran Amazon for twenty seven years, and he genuinely loved both the craft of company building and the vision of the everything store. But Amazon was never his purpose. His purpose was space, and had been since a high school speech,⁵ and in the end Amazon paid for Blue Origin. Reed Hastings loved the streaming future and had the strategic craft to build it, without ever claiming Netflix was what his life was for. Palmer Luckey had loved virtual reality since he was a kid; he built Oculus, sold it, and the vision moved on without him.

Purpose present

Geoffrey Hinton's purpose was to understand how the brain learns, his craft was neural networks, and he spent a decade employed at Google without ever holding the AGI vision. He left, in fact, to warn against it.⁶ That is purpose without vision-love, sitting inside an ordinary job.

Chris Olah's purpose is to understand what is inside neural networks, and his job is interpretability research at Anthropic, so his purpose and his salary happen to coincide. It is worth noticing why that is possible: his purpose lives in an industry that pays. Anthropic's founders more broadly build the very technology they are ambivalent about as a vision, because their purpose requires being close to it, which is purpose and vision pulling in different directions inside one company.

Yvon Chouinard's purpose was the environment and his craft was outdoor gear, and there was no dazzling technological vision anywhere in Patagonia. When it came time to exit he gave the company away rather than sell it,⁷ because purpose changes what an exit can even look like.

All three at once

Demis Hassabis has craft in AI research and games entrepreneurship, a vision of AGI, and a stated purpose of using intelligence to understand the nature of reality,⁸ which makes his purpose existential in the same way mine is. I keep him around as evidence that the triple is possible, and also as a reminder that he got there after a games chapter that was craft and vision only.

Steve Jobs had product craft, the product visions, and a purpose he described as building tools that amplify the mind. What I find most useful about him is a smaller story. In his early twenties he thought seriously about leaving for a monastery in Japan, and his Zen teacher told him he could pursue the practice through the company instead.⁹ He had asked, in effect, whether to choose purpose or career, and was told the two could be the same thing. For him they could, because his purpose was making, and a company is the natural home for making. That distinction matters later.

Musk at SpaceX is the clean case where the vision and the purpose are the same sentence.

Where vision and purpose split inside one life

Bezos had vision at Amazon and purpose at Blue Origin. Gates had vision at Microsoft and purpose at the Foundation. Richard Branson has vision-love without any purpose, moving from airlines to space to whatever is next, while Chouinard had purpose without any thrilling vision at all. Both built large companies, and only one of them could never sell.

The lesson I took from the whole list is that vision and purpose really are different things, and that most of the builders I admire spent years on visions that were not their purpose. Those years were not wasted. More often than not, they paid for the purpose.

What a job is for

After all of that, I think a job has to give you three things, and only three.

It has to give you money, meaning a real salary above your floor, cash that arrives every month rather than equity that might become money in seven years. If people depend on you, this is a hard gate, and a job that fails it fails no matter how exciting it is.

It has to give you freedom, meaning your evenings, reasonable hours and a contract that lets you work on your own things and publish. This is also a hard gate, because the purpose lives in the evenings, and a job that takes them takes the purpose with it however good the salary.

And it has to give you love for the craft, which I have come to think of as a threshold rather than a requirement for passion. You do not need to adore the daily work. You need it to be work you would do well without dread, with good problems and good people and some room to decide things, and you need to be able to respect it for a few years.

Everything beyond those three is a bonus. Vision is a bonus: if two jobs clear the gates and one is at a company whose future you find magnificent, take that one, because you will be happier and happiness at the day job is energy the purpose gets to spend at night. Purpose alignment is a bonus too, and a rare one. If a job happens to be your purpose, take it and do not look back. But it cannot be required, because most jobs are not anyone's purpose, and that is what the evenings are for. The only thing that can veto a job that clears the gates is shame: nothing you would be embarrassed to have built. Indifference is fine.

The example that made this concrete for me was imaginary. Picture a plastics manufacturer with a lean, good AI team and a good salary. I do not care about plastics at all, and under this test that turns out not to matter, because the craft is the same craft, the salary is real, and the evenings survive. T. S. Eliot did not care about banking and worked at Lloyds for eight years while writing The Waste Land, and he described the bank as stabilising.¹⁰ Wallace Stevens ran an insurance company and wrote poems on the walk to work.¹¹ Einstein had the patent office.¹² The meaning lived elsewhere, the job was a good patron, and none of them was working for the bank or the insurer or the patent office. They were working for the thing that mattered, and the job was paying for it.

Within that test there is a lot of room for different kinds of job, and they satisfy different loves. A large company gives you tech craft at scale and usually the cleanest version of the money and freedom gates. A research engineering role at a lab gives you craft and, if you are lucky, purpose in the same seat. Forward-deployed engineering fuses building with sitting next to the customer, which is a small dose of the entrepreneurial craft with a salary attached. And a founding engineer role,¹³ which is the one I know from the inside, is where both crafts live in one job: you are there before the thing exists, you decide what gets built and then you build it, you talk to the first customers because there is nobody else to, and you take ownership of the whole product because there is nobody to hand it to. If you love technology and have an entrepreneurial drive, it is one of the best roles that exists, and it asks for full ownership and full agency, which is not a warning so much as the whole point. It clears or fails the same three gates as any other job, and when it clears them it is worth taking seriously.

What a venture is for

Starting your own thing differs from a job in one important way. Love for the craft is guaranteed, since nobody starts a company otherwise, but love for the vision stops being a bonus and becomes necessary.

The reason is that a venture is years without reward, and craft alone does not survive that, because craft can be practised anywhere and the moment things get bad the craft-lover simply leaves. Vision-love is the glue that keeps a founder in the chair through the trough. Bezos had it, Gates had it, and every founder who lasted had it.

Purpose, on the other hand, stays optional, and this was the hardest thing for me to accept. It feels as though if you are going to spend ten years of your life on something, it had better be what your life is for. But the examples say otherwise. Amazon, Microsoft, Braintree: none of those was its founder's purpose, and every one of them paid for the purpose afterwards.

What purpose actually changes in a venture is two things. It changes the exit condition, because a venture without purpose alignment is a chapter that you eventually sell or step away from to go do the real thing, while a venture with purpose alignment is the life's work that you never leave. Chouinard could not sell Patagonia; Bezos could leave Amazon; both are legitimate, as long as you know which one you are in on the first day. And it changes how failure feels, because when an unaligned business dies, a business died, whereas when a purpose-aligned business dies it can feel like the purpose itself failed, and people leave whole fields after that. In this one respect the unaligned venture is safer for the purpose.

The real risk of an unaligned venture was never the years themselves. It is the venture never ending. Ventures are not two-year commitments but seven, ten, fifteen, and the failure mode is not Bezos but the founder who is forty five and still running the thing while the purpose quietly died on a shelf, or who changed so much inside the company that he no longer wants the purpose he once had. So the question I would ask of any venture that is not my purpose is not whether it is aligned, but whether it ends, and whether the purpose survives it.

If I ever start a flying car company, and I would love to, it will be because I love that vision and want that adventure, and because a life is allowed to contain more than one thing. I will know the exit condition before I start, and I will know what happens to the purpose while I am inside.

The purpose is a separate thing

This is the part I resisted the longest. I kept trying to make my career serve my purpose, and it made every career decision three times harder than it needed to be.

The purpose does not need the career. It needs time, honesty, and a patron, and for now the patron is the job. I have come to think of it in three phases, where the phases are about where the money comes from rather than what you spend your life on.

  PHASE 1                 PHASE 2               PHASE 3
  self-patronage          endowment             self-funding

  job pays you            savings pay you       purpose pays you
  purpose is unpaid       purpose is full time  purpose is the job
  fully separate          still separate        merged

  ─────────────────────────────────────────────────────────►
  the purpose gets stronger; money and purpose converge
  only once the purpose can carry money without bending

In the first phase the job pays for the person and the purpose is done in the evenings, unpaid and kept strictly apart from the money, because a one-person inquiry that has to earn immediately turns into a content business or a consultancy within a year. The separation is there to protect the purpose while it is small.

In the second phase your savings pay for the person. You leave the job when the number is right and the purpose becomes full time, still unpaid by itself. This is the familiar idea of leaving once you have enough, and it works, but only if "enough" is a number written down in advance rather than a goalpost that moves every time you approach it.

In the third phase the purpose pays for itself, through grants or patrons or products, and career and purpose finally merge. That merging is not a confusion; it is the goal state, and every serious research institution lives there. The merging that hurt me was trying to do it in the first phase.

This is also where the Jobs story comes back. His teacher could tell him the company was the practice because his purpose was making, and a company self-funds from the first sale, so he never had a first phase at all. A purpose that is inquiry rather than making has no such luck. The natural home for inquiry is a lab or an institute or a school, and none of those can pay for themselves on day one. That is the entire reason my career and my purpose have to start apart while his never did, and it is a difference in the shape of the purpose, not in the depth of commitment.

When the purpose does become an institution and someone says that a piece of its open research could be a product, the answer is structural rather than agonised: do not turn the institute into a product company, but let the institute own a product arm. Mozilla works this way, with a nonprofit foundation owning a for-profit corporation whose revenue flows back to the mission without the mission having to sell anything.¹⁴ The only question worth asking of any such money is whether taking it would change what we ask, and if it would, the answer is no however large the cheque.

One sorting rule keeps all of this clean. The institution owns only what descends from its research, and everything else is career. If a friend arrives with a startup idea, it is career, judged on money and hours. If a tool grows out of my own research and someone wants to pay for it, that belongs to the institution's product arm. The two do not blur, and once you hold that rule the question of where any new thing fits answers itself.

Two shapes of a life

Looking back over the examples, there seem to be two patterns.

In the first, everything aligns from the start: an aligned job, then an aligned venture, then the purpose. Ilya Sutskever went from Google Brain to OpenAI to Safe Superintelligence. Chris Olah went from Google Brain to Anthropic with the same purpose the whole way. Craig Venter went from the NIH to Celera to his own institute, and Andrew Ng from Stanford and Google Brain to Coursera and DeepLearning.AI. What they share is that their purpose was clear early and lived in an industry that could pay salaries and fund companies. Alignment was available to them because their inquiry had a labour market.¹⁵

In the second, the pieces stay separate: a craft job, then a venture built on craft and vision, then the purpose funded but never mixed. Bezos went from D. E. Shaw to Amazon to Blue Origin. Paul Allen went from a programming job at Honeywell to Microsoft to the Allen Institute for Brain Science. Bryan Johnson went from sales jobs to Braintree to Kernel and Blueprint. And there is a quieter version with no venture in it at all, which is Stevens and Eliot and Einstein, a job and a purpose on the side for as long as it took.

My purpose has almost no labour market. Consciousness inquiry does not hire the way deep learning hired Ilya in 2013, so the first pattern is not available to me today, which is a fact about the field rather than about me. The second pattern is my honest default: a craft job now, a vision venture if one ever clears the gates, and the purpose funded and never mixed.

There is one door that could move me into the first pattern later, and it is AI. It is the single corner of the world where my inquiry does have a labour market, because whether a machine can have a mind is itself one of the existential questions, and models are the closest thing to an instrument bench that consciousness research has ever had. AI is the one domain where craft, vision, and purpose coincide for me. That does not mean I should bend my career towards it; it means that if that door opens, it is the one job that does not require the separation. If it does, I become Olah. If it does not, I become Bezos with a much smaller Amazon and a much more interesting Blue Origin, and both of those are complete lives.

Open questions

I am not done with this. I do not know whether vision-love is really enough fuel for a venture that is not your purpose, or whether I am underestimating how much the trough hurts when the thing is not what your life is for. I do not know whether the question of whether a venture ends is genuinely answerable on the first day, or whether every founder believes theirs will end and then it does not. I do not have a good instrument for telling, from inside a job, whether I still respect the craft or have slowly started to dread it. And I do not know whether keeping the purpose strictly separate in the first phase slows it down more than I think; the examples say no, and my impatience says yes.

This is an ever-evolving process, and each of these questions will probably grow into its own article in time.