When I was an AI noob

A few years back, I wrote a blog post about things I used to do when I was a noob. Let’s be honest: at the time, I thought I wasn’t a noob anymore, and I was trying to be funny while sharing my journey with junior developers.

These days, with the advent of AI, I feel like a noob again, so I decided to share a few things that I was — or still am — doing as an AI noob.

What did I do when I was an AI noob ?

• When I was an AI noob, I used to think AI was just a fancy auto-completer, and that I could probably write code faster than it could make suggestions.
• When I was an AI noob, I used to think AI suggestions were useless. Am I really an engineer if I use AI-generated code?
• When I was an AI noob, I was afraid of disclosing that I’d used AI to generate some code. What would people think of me if they knew I didn’t come up with it myself? That I’m a noob?
• When I was an AI noob, I was afraid of letting my agent perform git operations. At the very least, I was going to write my commit messages myself. A man needs his dignity.
• When I was an AI noob, if the AI got something wrong I would fix it by hand, quietly, instead of telling it what was wrong — and then act surprised when it overwrote my fix five minutes later.
• When I was an AI noob, I ran a single AI session at a time. Somewhere, an idle agent was waiting for me to finish reading its output, wondering if I was the bottleneck. I was the bottleneck.
• When I was an AI noob, I thought code was precious — that I needed to review every line, I’d delay shipping a feature users were begging for because a variable name offended me. Twenty years of experience had to be good for something, right? right?
• When I was an AI noob, at some point I got over my AI fears. That’s when the real noob era started.
• When I was an AI noob, I would type “try again” fifteen times in a row, hoping the sixteenth attempt would be different. Reading the code myself was never on the table.
• When I was an AI noob, I spent forty-five minutes prompting, re-prompting, and refining instructions to automate a task that would have taken me four minutes. But it’s about the principle. After that, I spent 40 more minutes creating a skill based of my ingenious prompting. Those accumulated one hundred skills are more important than actually shipping.
• When I was an AI noob, I wasn’t reading a damn thing: AI-generated descriptions, AI-generated commits, AI-generated code. Shipping is what matters, right? Somewhere, a dashboard was counting my PRs, and the dashboard needed to know I was a 100x engineer. Definitely not a noob anymore.
• When I was an AI noob, AI gave me so much confidence that I’d jump into big, unknown repositories to address my own needs, ignoring all the context and history of the project. I thought I knew better, because look — I can implement these big features!
• When I was an AI noob, AI gave me so much confidence that I’d parachute into repositories I’d never read, PR in hand, solving problems the maintainers didn’t know they had, ignoring ten years of project history because Claudio and I had figured it out in an afternoon. I forgot that everyone had their own Claudio.
• When I was an AI noob, I’d open that huge PR without understanding or bothering about a line, expecting the other maintainers to review it and land it for me, and when a reviewer ask the question, I’d relay it to my agent and reply with: “Claudio said that this line is load bearing, so let’s keep it”
• When I was an AI noob, I was shipping feature after feature, until at some point I lost track of how many features we even had. How could I expect to market them properly, or users to even know they existed?
• When I was an AI noob, I tried to set up what folks call a “software factory”: shipping sites and applications at the speed of thought, sometimes even letting the AI come up with the ideas. What I forgot is that software has no value if no one uses it. I took pride in the act of shipping, but forgot why I was shipping in the first place.
• When I was an AI noob, I wrote a post about the things I did when I was an AI noob, believing — for the second time — that I was no longer a noob.

And you ? did you learn how to use AI properly ? or still fumbling in the dark like me.