My AI Coding Journey
It's almost six months since my last post on this blog. In that time, my daily work changed rapidly and completely. Until November 2025, I thought AI was not useful for my work in complex enterprise Angular projects, where code quality has to be very high.
But then Opus 4.5 was released to the public on November 24, 2025. I started to experiment with it and quickly found that it could help with real work: code generation, code review, documentation, modernization, refactoring, and a lot more.
The quality surprised me first. Some generated Angular code looked close to how I would have written it myself. The speed surprised me next. Naturally, I felt a bit obsolete, but it was and still is pretty exciting.
The biggest change is not that the latest models write better code. The bigger shift is that using them has become much easier: I no longer need to craft a perfect prompt upfront and hope that the result matches my intent. Instead, the interaction feels much closer to talking to a senior developer: I can describe the problem, add constraints, answer questions, clarify the spec, and let the model help discover what we actually want to build.
