Every new era brings its own set of pros and cons. Sometimes the balance between them really matters. What I want to talk about here is one of the discouraging effects AI has on us engineers today.
AI has brought major changes. With AI editors, we can finish boring tasks very quickly. Even people who do not know how to code can now show impressive demos online through vibe coding. This is all great. Everything is faster and more accessible. On paper, it feels like we have already won. But what have we lost?
If we think back to software development before AI, the workflow was different. We relied heavily on search engines. We did long research sessions for a feature or a bug fix. We jumped from site to site, collected hints, and followed trails. Not every problem had a simple answer. It felt like detective work.
I still think of software engineering as a kind of detective work. You gather as much material as possible, then you act. This is still true today, but the scale has changed. There was also another side to these practices: communication with other people. When we ran into a problem, we would ask questions on forums. Anyone remember Stack Overflow? We asked for help. Sometimes we were mocked for asking simple questions, but that was part of it. Through comments and discussions, you could directly interact with engineers who were actively working in the field. It gave us a chance to meet people who were really good at what they did. I cannot say this is completely gone today, but it is clearly fading. In a world that is becoming more individualistic, this has become even more so. I still remember the joy of finding the correct answer hidden between small comments on Stack Overflow.
When we worked in offices, if we had a problem, we would go to a colleague’s desk and think it through together. Now we prefer asking an LLM. Ask and move on. It is fast. It feels rewarding. A quick hit for the dopamine receptors.
I wonder what kinds of opportunities we are starting to miss. Meeting new people? The unexpected opportunities that might come from those meetings? I do not have an answer yet.
In May 2021, someone named David Dohan reached out to me on Twitter. At the time, he was an engineer at Google DeepMind. I had recently published a ClojureScript library called re-frame-flow. David liked the library a lot and asked me for ClojureScript consulting. He was working on a kind of Clojure and React flowchart-style application and needed extra help. We then spent nearly two hours on a consulting session and solved his problem. It was a great session. Helping such a smart engineer was a really nice experience. I suspect that today, if he needed help, he would just ask an LLM and move on, and this connection would never have happened. These kinds of encounters could have led to great beginnings, but AI may be quietly cutting off that path.
What I am saying is this: every new innovation will take some things away while bringing others. I just hope we do not start regretting what it takes from us.