After going the full hype cycle with Claude Code, I'm not convinced that it's a productivity win for "real" code

A developer, after several months of using AI tools like Claude Code, specifically Opus 4.8 and Opus 5, has concluded that these tools are not a productivity win for "real" or important production code. Initially enthusiastic, the developer's opinion shifted, finding Claude Code less useful than first thought, and even noting that Opus 5 was worse than Opus 4.8.

This report offers a developer's long-term perspective on Claude Code's practical utility, unlike initial hype, and specifically compares Opus 4.8 and Opus 5 versions.