Lately, the term 'Software Engineer' seems to be an ambiguous one. Personally, I still engineer software constantly so identifying with the role should be simple - but I have always associated the title with somebody who writes code - and I certainly do not do that anymore.
Two years ago, I would spend half my day trawling through error logs as I attempted to produce a bit of software. The loop process of 'write the code', 'check for errors', 'write the code', 'check for errors' until finally hitting the sweet satisfaction of a functional app has buried itself in the past in what feels like an instant.
2025 still had a bit of debugging for me, but any errors I encountered were mostly pasted into ChatGPT and solved quickly. Now, in 2026, with the fast rise of Claude Code, it has been a long time since I have seen the thick red lettering "ERROR:" grace the grainy text on my terminal screen. Seldom will the AI produce code that generates such a thing, and when it does, it will spare me the agony of seeing it, and quietly deal with it behind the scenes.
Part of me misses error logs and debugging. Laboriously tweaking the script and spitting out the same error log about 20 times while constructing an application can be compared to climbing the greasy rungs of a ladder. The harder the journey, the more satisfying the result. Battling with bugs hammers the memory of building something firmly in your mind. When you have a constant puzzle to solve you are connected intensely with what you are producing and your progress is much more visceral. Now that Claude is able to build everything for me easily, the way of working has changed. I can produce whatever application I want - but if I am not careful, I can produce too many applications, not be engaged with my output, and even forget which tasks have been completed.
Right now AI is cheap - and there is no need to be conservative about compute. In the future (when the top companies decide to start making a profit) all of this will change and we will all need to be a bit more mindful about what we build and how we build it. The new hurdle will not be debugging; it will be optimising token usage and careful product planning to save not on build time but build costs. A new and exciting challenge: the software engineer mindset will need to be ready for this!