Work at Someone Else's Startup First. I Learned This the Expensive Way.
I did this in the wrong order.
I started my first company at fifteen. Then I spent two years co-founding another one, building a product that reached real users, made real revenue, and taught me an enormous amount. Only after all of that did I join somebody else's early stage company as an engineer.
I expected the job to feel like a step sideways. Instead I spent the first month quietly realising how much I had been guessing about, for years, with total confidence.
The standard advice is that you should just start your own thing, because nobody learns to swim from the side of the pool, and every year spent working for someone else is a year of your own equity you gave away. I believed that completely. I now think it is right about the destination and badly wrong about the route.
The Problem With Learning Only From Your Own Company
When you run your own startup, you learn one company deeply. That sounds like a lot until you notice what it actually is, which is a sample size of one.
You experience one market, one set of customers, one funding situation, one way of making decisions. Every lesson you take from it is entangled with the specific circumstances you happened to be in, and you have no way to tell which lessons generalise. You cannot separate what is true about building companies from what was true about your company on that particular Tuesday.
Worse, you only see your own failures, and you see them from the inside, where they are hardest to interpret. When something went wrong at Existence, I could tell you exactly what happened, and I could give you four plausible explanations for why, and I had no way of knowing which one was correct. There was no control group. There was no second attempt with one variable changed.1
Joining someone else's company gives you a second data point, gathered at somebody else's expense, on a much shorter timeline.
What You Only See From the Inside of Someone Else's Company
The specific things that surprised me were not strategic. They were procedural, and they were things nobody writes down because everyone who knows them assumes they are obvious.
How a decision actually gets made when two senior people disagree and both have reasonable positions. How a deal moves from an introduction to a signature, and how many unremarkable follow ups sit in the middle of that. What a real customer conversation sounds like when the customer is not being polite, which is nothing like the conversations you have when you are the founder and everyone knows it.
I watched people who had done this several times before make decisions in ten minutes that I would have agonised over for a week. Not because they were smarter, but because they had seen the situation before and knew which parts of it did not matter. That is the actual content of experience, and it is almost impossible to transmit through writing. You have to be in the room.2
You also get to see the boring machinery that a working company runs on. How things get prioritised. What a functional hiring process looks like. How money is actually tracked. When you have only ever run your own thing, you invent all of this from first principles, badly, while also doing everything else.
The Cost Argument
Here is the framing that convinced me, and it is not about learning at all. It is about who pays.
When you make a mistake in your own company, you pay for it with your runway, your reputation, and often two years of your life. When you make the same mistake inside someone else's company, you learn the same lesson and they absorb the cost. You are being paid to acquire pattern recognition that would otherwise cost you an entire venture to obtain.
That is an extraordinarily good trade, and the people who reject it are usually rejecting it on grounds of status rather than economics. Being a founder sounds better than being an employee. That is genuinely the reason, even when nobody says it out loud.3
The counterargument is that you lose time. But you only lose time if the years are spent somewhere that teaches you nothing, which brings me to the part that actually matters.
Not Any Job Will Do
This entire argument collapses if you join the wrong company, and most companies are the wrong company for this purpose.
You want to be somewhere small enough that you can see the whole machine. At a company of fifteen people, you sit near the person doing sales, you hear the customer calls, you watch pricing get decided. At a company of five thousand, you get a well defined slice of a system nobody fully understands and you learn how to be a good engineer, which is valuable and completely different from what we are discussing.
You want to be somewhere that is actually trying to figure something out, not executing a plan that was settled two years ago. The uncertainty is the curriculum. If everything is already known, you are learning to operate rather than learning to decide.
And you want proximity to whoever is making the hard calls. Not access in the sense of a title, but the ordinary kind where you are in the room when things are unresolved and you get to watch how a decision is actually reached rather than hearing the summary afterwards.
If a job does not offer those three things, it is a job, and this argument does not apply to it.
And Do Not Stay Too Long
The failure mode on the other side is real and I have watched it happen to people.
A good early stage job is genuinely comfortable. The salary is fine, the problems are interesting, you are respected, and the risk belongs to somebody else. Every year you stay, the opportunity cost of leaving goes up, and the story you tell yourself about needing a bit more experience gets easier to believe. There is always one more thing to learn.
Two or three years is enough. It is enough to see a full cycle, to watch a strategy get chosen and then hit reality, and to build the pattern library you came for. Past that, you are mostly repeating the first year with better tools.
The point of the job was never the job. It is to arrive at your own company already knowing which mistakes are survivable, so you can spend your limited attempts on the ones nobody could have warned you about.
What I Would Tell Fifteen Year Old Me
Start things. Start early, start badly, ship them to strangers. I would not undo any of it, and the instinct to build your own thing is not something you should suppress until you feel qualified, because that feeling does not arrive.
But somewhere in there, go and work inside somebody else's early stage company for a couple of years. Not to get credentialed, and not because you are not ready. Go because it is the cheapest way to get a second data point, and because you cannot see the shape of the thing you are inside of while you are the one holding it up.
Learning on your own dime is admirable. Learning on someone else's is just faster.