Writing
AI Coding Part 1: Why I Stopped Feeling Guilty
On legitimacy, assistance, and the difference between building software and performing purity
For a while, using AI to build software felt faintly illegitimate - as if the work only counted when it looked difficult in familiar ways. This essay is about that discomfort, where it comes from, and why I eventually stopped performing anxiety about tooling and started asking a better question: what kind of product work am I actually trying to do?
The first reaction I had to AI-assisted coding was not excitement. It was embarrassment.
I stopped asking whether AI use made the work legit, and started asking whether I still owned the decisions.
Not because the output was bad. Often it was useful. The discomfort came from somewhere else: a feeling that if the work became easier, maybe it counted less.
When people talk about AI coding, they usually argue at the level of capability: speed, quality, reliability, risk. All of that matters. But there is a social layer underneath it that we pretend is not there.
Tool choice has always carried status signals in software culture. How hard something looked, how much you did manually, how visibly craft a workflow appeared - these were often treated as proxies for seriousness.
AI unsettles those proxies. That can feel like liberation. It can also feel like loss. The early guilt, at least for me, was less about ethics and more about identity.
I stopped asking whether AI use made the work legit, and started asking whether I still owned the decisions. If I can still explain the architecture choices, defend the trade-offs, understand the failure modes, and carry responsibility for outcomes, then assistance is exactly that: assistance.
This shift changed how I build. AI did not just make existing work faster; it changed how quickly I could test options before overcommitting. That became part of how products like PLOT, Restormel, and SOPHIA could be explored in practice.
Part 2 is about the cost side of this shift. Part 3 is about what we may lose if fluent generation replaces careful thinking.