Unfamiliar Stops Being Unfamiliar
For the first two weeks of this build, most of what came back to me was words I did not know. Not the concepts. The vocabulary. That density is why most people who could build something never start.
For the first two weeks of this build, most of what came back to me was words I did not know.
Not the concepts, which were mostly reasonable once explained. The vocabulary. A wall of acronyms and product names and terms of art, arriving at a rate where I could not tell which ones were important, which ones were incidental, and which ones were the same thing wearing a different name. Every sentence contained two things I would have to look up, and looking one up produced three more.
That experience is the reason most people who could build something never start. It is not difficulty. It is density, and density feels like difficulty for about three weeks.
Here is what I would tell somebody who is considering building an application.
Almost none of it is hard. All of it is unfamiliar, and those are completely different problems. A webhook is one system calling another when something happens. A migration is a written instruction for changing the shape of a database. An environment variable is a setting kept outside the code so it can differ between your laptop and production. None of those are difficult ideas. They are simple ideas with names that tell you nothing, arriving faster than you can absorb them.
Once you understand that the difficulty is a naming problem rather than a comprehension problem, the wall stops looking like a wall and starts looking like a list.
The learning method is saturation, not study. I did not sit down and read documentation front to back, and I would not recommend trying. What worked was constant immersion: encountering the same term in six different contexts over two weeks until it stopped being a term and became something I had an opinion about. That is not a technique. It is exactly how you learned your own industry.
Think about how a person learns freight. Nobody reads a book about it. They hear detention and accessorial and tender and drop trailer forty times in a month, in situations where the meaning is constrained by what is happening around them, and one day they realize they have known what those words mean for a while without being able to say when it happened. Software vocabulary works identically. The mechanism you already have is the right one.
The variable that decides how fast this goes is your willingness to say you do not know.
That is the whole thing, and it is where most people fail, because it is an ego problem rather than a capability problem. You are in a conversation, a term goes by, you have a rough sense of it, and you nod. Now you have made every subsequent conversation harder, because you cannot ask later without admitting you did not ask earlier, and the gap compounds. I have watched senior people in my own industry do this for years about their own systems.
The cheapest thing available to a builder is the question "what is that, actually." The cost is three seconds and the alternative is a hole in your understanding that you will route around forever.
Which brings me to something about AI that I think is underrated, and it is not the code.
It removes the social cost of not knowing. I have asked the same question nine different ways in a row. I have asked things I would have been embarrassed to ask a person. I have gone back to something explained to me an hour earlier and asked for it again, differently, because the first explanation did not land. No sighing, no impatience, no sense that I was using up somebody's goodwill.
For a non-technical person building something, that matters more than the code generation does. The bottleneck was never that I could not learn this material. The bottleneck was that learning it would have required a patient expert on call for months, and nobody has that. The embarrassment tax on ignorance is what actually stops people, and it just went to zero.
The failure mode to watch for is working around instead of working through.
You will develop a piece of the system you always route around. A thing you copy from somewhere else without knowing what it does. A configuration you do not touch because it works and you are not sure why. That is the tell, and it is worth treating as an alarm rather than an inconvenience, because those places do not stay small. Every future change near them costs more, and eventually one of them is where your problem is and you have no ability to reason about it.
Going through it costs an afternoon. Going around it costs a little bit forever.
And the curve drops far faster than it feels like it will.
The first time you do an unfamiliar task, it takes all day, and most of the day is not the task. It is figuring out what the task even is. The second time takes a fraction of that. By the fifth it is unremarkable, and by the tenth you are not aware you are doing it.
You pay for unfamiliar once, per task, and only if you go through it. What makes the early weeks feel endless is that you are paying that bill on twenty things at the same time, and every one of them is at its most expensive. That does not last, and while you are inside it there is no way to tell that it does not last, which is why somebody should say so out loud.
There is a second curve running underneath that one, and it took me far longer to notice.
At the start I could hold exactly one task. Not one at a time as a matter of discipline, one because that was the entire capacity. Everything else fell out of my head while I worked, and finishing meant reconstructing where I had been before I could start the next thing.
Then the same tasks got faster, which is the curve in the chart above.
Then something changed that was not speed at all. I started seeing the consequences of a change before making it. Not all of them, but enough that I would stop mid-sentence and say that this is going to break something over there, so check that first. That is a different capability from doing the same work quickly, and it arrived without my noticing it had.
After that I could carry more than one thread. Work moving on one front while I was thinking through a second, holding both without losing either.
And most recently, the part I did not anticipate: I rely on the answers less and challenge them more. Early on, a proposed solution was information I did not have, so I took it. Now a proposed solution is an opening position, and my first instinct is to ask what it does not handle, what it quietly assumes, and whether there is a version that holds up better under load. That is not skepticism for its own sake. It is that I finally know enough to hold an opinion worth defending.
Every one of those stages felt like the ceiling while I was standing on it.
None of that would have happened on somebody else's schedule. The learning went at whatever pace I happened to be moving that week, deep on the parts that mattered to me and shallow on the parts that did not, revisiting anything as many times as it took. Setting your own pace is, to my mind, among the most valuable things this technology offers, and it has almost nothing to do with what it can produce.
The test I use for whether I actually understand something is whether I can explain it to somebody else without using the term. If I can only define a webhook by saying webhook, I do not have it yet. If I can say it is one system calling another when something happens, I do. That test is quick, it is honest, and it will tell you which parts of your understanding are real and which are vocabulary you are wearing.
None of this requires being technical. It requires being willing to be a beginner in public, at an age where most people have stopped doing that, and doing it repeatedly for several months.
That willingness is the entire qualification. Everything else is available.
Selecting supply chain technology?
PreShiftIQ matches buyers to vendors on measured fit across TMS, dock scheduling, ELD, carrier vetting, and fleet management. Vendors pay for outcomes, never for influence, and the buyer match is free.
See how matching works at preshiftiq.com/how-it-works.
Next in the series: what AI actually does, and what people think it does.
