In 1995, the Christian Broadcasting Network hired me as a junior FoxPro programmer. CBN was the first place I had the Internet on my desktop. It didn't help my FoxPro work much, but I started learning HTML, server-side scripting, and eventually a new language called PHP.

It was obvious to anyone paying attention that desktop applications were dead. Developers who only knew the desktop got shunted onto side projects, and those were the lucky ones who still had jobs. If you didn't have a job, you had better be able to prove your Internet chops, or you had no chance.

Companies "refocused" and laid people off. But new companies sprang up too, little startups like Amazon, Yahoo, and RealNetworks (remember RealPlayer?), and they hired the developers who had made the jump. Those developers didn't throw away their careers and start over. This wasn't coal miners being callously told to "learn to code," like it's a weekend course at the adult learning annex. It was "learn the new tools, and bring what you know about software development with you."

We're in another shift now. Companies are refocusing again. Some of it is AI, and some of it is cleaning up COVID overhiring. I honestly don't know where the new jobs will come from this time. I think it's too early to tell. Nobody knew who the Amazon of the Internet would be in 1995 either. But I know who they'll hire: developers who can prove their AI chops. (I wrote about why developers are scared right now.)

Here's how you start building that proof.

Hand it something you keep putting off. Pick a task you dread, give it to AI, and say, "Interview me until you understand." The first time takes about fifteen minutes. Don't ask AI whether it can solve the problem. Tell it to solve the problem, then talk it through. Asking implies you're not sure it can. Telling says you know it can.

Tell it what, not how. Describe the outcome you want and let AI work out the implementation. Then review it like you'd review a pull request from a junior developer, because the review is where your years of experience pay off.

Build something you don't know how to build. In the last month alone, I've used AI to:

  • Build a workflow that edits and publishes a video and audio podcast, unattended
  • Hack a digital picture frame so I can share pictures with my mom
  • Build a personal assistant that runs on Claude

Each of these projects built on existing ideas and, more importantly, experience. I could build these things because I understood the outcome I wanted, even if I didn't understand the path to get to that outcome.

AI did not make me useful. I made AI useful.

In 1995, the proof was a website you'd built. Today it's something you built with AI that you couldn't have built without it. Go build it.

Originally posted on LinkedIn: "In 1995, the Proof Was a Website"