Research
Bootstrapped a healthcare business to eight figures on no outside money. The lessons that survived: ship over polish, make risk cheap, and outlast a dopamine-seeking brain.
I bootstrapped a healthcare business to eight figures in revenue with no outside money. That sentence is the whole origin story, and I have learned to say it exactly that flatly, because everything else people want to add to it — the drama, the grind narrative, the founder-mythology gloss — is decoration on top of a much simpler fact: I built something real, on my own money, and it worked.
The beliefs that survived are the same three I still run everything on. Ship over polish. Make risk cheap to execute, not smaller. Keep every door open. None of them are original, and I do not pretend they are. What makes them mine is that I paid for every one of them in cash, in real time, with a business that either made payroll or did not.
Ship over polish is the one people misread first. It does not mean ship broken things and shrug. It means the version of a thing that exists and is imperfect beats the version that is perfect and does not exist yet, every single time, because only the first one can be corrected. A healthcare business does not get to workshop its offer for six months before finding out whether customers want it. You ship the imperfect version, you watch what breaks, and you fix the parts that actually mattered instead of the parts you were most anxious about in advance.
Making risk cheap to execute is the sharper version of the same idea, and it took me longer to be able to say it in one sentence: an entrepreneur should not be in the business of minimizing risk. The business is making risk cheaper and faster to execute. Minimizing risk is what you do when you are protecting something you already have. Making risk cheap is what you do when you are still building something, and it is the only posture that actually compounds, because it means you get to take the same bet ten times for the cost most people spend taking it once.
Keeping every door open sounds like indecision until you have watched what the alternative costs. A bootstrapped business with no outside money has exactly one strategic asset that a funded competitor does not: nobody can force you to close a door you are not ready to close. No board seat, no term sheet, no investor timeline deciding for you when the experiment ends. That asset is invisible on a balance sheet and it is the reason I could eventually walk away from a healthcare business at eight figures and rebuild my entire working life around agents, on my own schedule, with nobody's permission required.
None of that would matter if the same brain that built the first business had not almost sabotaged the second act. I have a dopamine-seeking brain, and I do not say that as a diagnosis, I say it as an operating constraint I have to design around. Everything new looks shiny. Every fresh repository, every new product idea, every clever agent workflow produces the exact same hit a slot machine produces, and the hit arrives whether or not anything actually shipped. You have to finish stuff. Otherwise all of it is just noise dressed up as productivity.
I wrote down what a year of coding almost exclusively with agents actually taught me, and the dopamine problem is the one that opens the list, ahead of anything technical. Then comes the second lesson, which sounds like a throwaway line and is not: code is temporary. Every three to six months the tooling changes enough that you want to throw away what you built anyway, so the discipline is building small things that help you now instead of big things you never launch. The two lessons are the same lesson from two directions — stop chasing the shiny new start, and stop being precious about the thing you already started, because neither extreme finishes anything.
The unglamorous fourth lesson from that same list is the one that actually keeps agent-built products alive past week one: refeed the AI the problem, over and over. When a build feels stuck or buggy, the instinct is to abandon it for a cleaner idea. The move that works instead is boring on purpose — say continue, say fix the bugs and the bad logic, and let the model grind on the same problem again instead of you grinding on a new one. Counterintuitive, but it is the mechanism that turns a dopamine-seeking brain's worst habit into something survivable: you get to feel the thrill of a new prompt every few minutes while the underlying project stays exactly one project, not fifty abandoned ones.
Architecture and planning matter more than obsessing over every small detail, and that lesson has nothing to do with agents specifically. It is the same lesson a healthcare business teaches anyone running it on their own money: the founder who spends the week perfecting a single form field while the actual system around it is unplanned is optimizing the part that feels productive instead of the part that determines whether the business survives. Agents just made the gap more visible, because a model will happily polish a detail forever if you let it and never once ask whether the surrounding structure makes sense.
There is a heart-versus-hustle version of the same tension I have had to sit with honestly. Providing genuine value to others is only one of the two pillars a business actually needs. The other is following what actually energizes you enough to keep going once the initial rush fades, and most people, myself included in the early years, obsess over the first pillar and ignore the second completely. You can build something valuable to others and still burn out hard, because the dopamine from quick wins always fades on a schedule shorter than the business needs you to last. The healthcare business taught me the first pillar. Building in the open with agents, oddly, is what finally taught me the second.
Learning to be calm did not make the list by accident either. The instinct at scale — more agents, more machines, more parallel goals — is to speed up further, and speeding up further is exactly how you end up tired instead of finished, especially once you are context-switching across a dozen tmux panes instead of one desk. A healthy fleet is not the one running the most agents at once. It is the one whose owner can still think clearly enough to notice when three of those agents are working on the same mistake.
Alumia is the proof that the discipline actually holds under real pressure, because I have rebuilt that product from scratch four times. The honest way I put it to the team going into round four was not a pep talk — it's going to be beautiful because we're going to use everything we learned to actually build something that works. A dopamine-seeking brain treats a failed attempt as permission to go chase a different shiny idea. The discipline I am actually proud of is treating a failed attempt as attempt data and starting the fifth version instead of the first version of something else.
The addiction never really left, it just changed shape. Coding with agents feels like playing a slot machine, and I have said that about my own daily habits more than once, half joking and half not. Every agent pane is a lever. You pull it, something interesting comes back, and the brain that once needed a new business idea every quarter now gets the same hit from a new prompt every few minutes. Recognizing the addiction did not cure it. It just gave me something more useful to point it at than a rotating cast of unfinished side projects.
The last lesson on that list is the shortest and the one that actually resolves the whole tension: if you want to feel satisfied, ship it. Not perfect it, not plan it further, not start the next idea before the current one is out the door. Satisfaction was never going to come from the dopamine hit of starting. It only ever came from the much quieter feeling of something existing in the world that did not exist yesterday, which is the exact feeling a healthcare business gives you the first time a real customer pays for something you built alone, and the exact feeling a shipped product gives you now, whether the hands that typed it were mine or a fleet's.
Why pivot the whole operation to AI instead of running a second version of the same kind of business. Because the actual lesson from the first company was never the industry I happened to build it in. It was that one person, with no outside money and a willingness to make risk cheap to execute, can build something at real scale if the leverage is right. Agents are the largest leverage shift I have seen in my working life, larger than anything available to me when I built the first business by hand. Not using it would have meant deliberately picking the smaller lever on purpose, out of loyalty to a business I had already proven I could build.
Over 170 billion tokens spent coding with agents this year is not a vanity number to me. It is the same instinct that built the first business, pointed at a completely different substrate: put in the reps, ship the imperfect version, refeed the same problem instead of chasing a new one, and let the volume of attempts do what caution never could. The healthcare business took years of cash-funded trial and error to reach eight figures. The agent fleet compresses that same trial-and-error loop into a pace no unfunded founder working alone could have touched a decade ago.
People ask if I miss running a traditional company. I do not, and the honest reason is not romantic. The old version of me had one body, one set of hours, and a business that could only grow as fast as I personally could execute or hire. The current version has the same appetite for risk and the same refusal to let a board slow me down, aimed at a fleet that does not sleep and does not need a headcount plan to scale. The lesson survived the industry change completely intact. Only the tool changed.
Ship the imperfect version. Make the risk cheap. Refeed the same problem instead of starting a new one. Finish what you started before the next shiny thing pulls you off it. That was true running a healthcare business on no outside money, and it is still true running a fleet of agents today. The brain that chases dopamine never goes away. You just have to build systems that outlast it.