The Codex dashboard shows 92% of my usage quota left. That’s enough to ask for almost any software project I can imagine: a registration system, an Elixir app, a full prototype. And yet, when I open the computer to think about the next project, I almost never know what to ask for.
Programming has always been a hobby for me. Since I was a kid, it was a form of play: building websites, trying different programming languages, starting projects that didn’t need to prove their usefulness before existing. After work, that impulse kept going. I came back to the computer out of a desire to discover what I could make.
That habit gave birth to several small projects. In 2022, I created Meu Santinho. It was also a national election year, and I started exploring the TSE Candidates API. In a few days, I built a proof of concept: a user could select the candidates they would vote for, and we generated a “santinho” image to share on Instagram. Then I called a few friends to finish the app and we launched it on the web.
Seeing something you created being used is always a great pleasure. But today, when I can write a page describing that same software and it appears in front of me, I realize I miss the process of creation. Modeling each part, discovering the problems that showed up, and programming line by line was a pleasure too.
It’s still possible to program by hand like before — nothing stops me. But it’s not the same. When I know an agent can walk the same path for me in moments, deliberately choosing something less efficient is hard. Especially when the alternative is one sentence away.
On one hand, I feel an ineffable power in being able to program technologies I’m not proficient in (e.g. native iOS apps, data pipelines with Airflow, realtime systems with Erlang). On the other, the moment I realize the challenges will be solved by agents before I even see them, I get a little discouraged.
There was a time when I was very curious about learning iOS programming with Swift. I watched some courses, found the most interesting YouTuber talking about the subject, and started making small apps in Xcode. It took me a while to build a login and a list screen, but that process was deeply rewarding.
It was rewarding for two reasons. The first was simply that I could do it with my own hands. The second was that it felt like the prelude to something bigger. I felt like I was in preparation. Today, I see it as entirely plausible to build a complete iOS app without needing to go through the login-and-list app. Why am I not happy about that?
Curiosity works when there is a mysterious but reachable gap in knowledge. We engage when we feel progress toward it. When the answer seems available before the question, curiosity loses its urgency. I have the impression that this is what happened to my relationship with building software: it lost some of its mystery.
That’s why I look at Codex and don’t know what to ask for: the answer already seems to be there.
