I am a completely self-taught software developer. I’ve never taken a computer science course in my life. I was lucky enough to attend a junior high school in the 1970s that taught me BASIC. I loved it, and used to stay after school to write and play simple text-based games. 

Now this may be hard to believe, but at that time, being a computer nerd wasn’t as cool as it is today, so I left it alone until the 1990s when the PC revolution was getting underway. Shareware was all the rage, and Windows was brand new. I combined the two to write some modest little applications in Turbo Pascal for Windows that had some success. 

Looking back on those apps, the code I wrote was really, really bad. Like “I had no idea about passing parameters to functions, so everything was a global variable” kind of bad. I distinctly remember an enormous struggle with strings because I had no idea that I needed to explicitly allocate memory for them. I banged my head against the keyboard for many hours trying to get things to work. 

Becoming developer wise

There was no Internet then, so I learned by asking questions on CompuServe forums and reading books. But mostly I learned through experimentation, trial and error, and by actually doing — by writing code, failing, and trying new ideas. Lightbulbs would go off in my head as I started to notice how to pass parameters to move information from one place to another, or when I finally understood object-oriented programming.

Eventually, I moved from asking to answering questions online and from reading to writing books on good software development techniques. I figured it out, and now I like to think I know what I’m doing when it comes to designing good systems.

This is all a rather long-winded way of saying that it took many years, but I somehow acquired wisdom in the domain of developing software. I didn’t need classroom training to get it, and I’m not sure that one can even gain developer wisdom in the classroom. One gains it by writing lots of bad code, reading lots of good (and bad) code written by others, and by seeing code work and seeing code fail. A lot of developer wisdom is gained by revisiting crappy code that one wrote six months ago, and seeing how hard it is to maintain, and vowing not to make that mistake again.

And perhaps we have arrived at a point today where all that wisdom that we longtime developers have gained is simply not needed anymore. Agentic coding has put us in the curious position of being able to create software without wisdom. In theory, all the wisdom of all the developers in the world is at your fingertips, and all you have to do now is ask. I asked Claude Code to implement an idea for a website, and he created it. It works.

And here’s my confession: I haven’t looked at the code. I didn’t even feel the need to do so. If there was a problem with the site, I would tell Claude about it, and he’d fix it. The site works. It works great, actually. Not only that, but it does things that I would have taken hours and hours to figure out. Things like making sure that contact forms don’t get spammed and that APIs are properly rate-limited. I asked Claude to review the site for vulnerabilities, and he found and fixed them.

The sum of all developer wisdom

Or put another way, Claude Code is a lot wiser than I am about how to build good, safe, properly functioning code. He’s a pretty good programmer, and he’s getting better every day. It’s amazing because having the wisdom of millions of developers at your fingertips is cool. It is terrifying because where will we be if acquiring wisdom becomes passé? The wisdom captured in Claude is a collection of all the smarts encapsulated in billions of lines of code on GitHub. If we do nothing but leverage existing wisdom, what will feed the next generation of Claude?

In the end, I lean towards amazing. New software developers will be gaining wisdom at a level of abstraction well above the code. Somehow, they will build software using a wisdom that my fellow code warriors and I don’t even understand yet. Debugging will involve understanding how to tell Claude what the problem is and what the solution needs to be. Building software will always require sound judgment — but the things we make judgments about are continually changing.

Eventually, writing code will be like learning Latin — cool, interesting, and good for your brain, but not necessary to function in the day-to-day world.