A seasoned developer explains why they've abandoned AI-powered coding assistants despite their popularity. After running out of free credits while using LLM-integrated IDEs, they found the subscription model absurd and returned to traditional tools like Emacs. Their resistance stems partly from generational experience with past automation hype cycles.

The author draws on Fred Brooks' classic distinction between accidental and essential complexity in software development. While modern tools—including AI—effectively reduce accidental complexity like boilerplate code and refactoring tasks, they cannot address essential complexity: designing elegant, maintainable abstractions. This requires deep system understanding that LLMs lack.

Drawing from The Mythical Man-Month, the piece argues that programming's core challenge remains creating simplified representations of messy reality. AI excels at predictable tasks but struggles with the nuanced, context-dependent decisions that define quality software architecture. The author concludes that essential complexity demands human judgment, not algorithmic shortcuts.

Their perspective reflects broader skepticism about whether current AI tools meaningfully advance software engineering practice beyond incremental improvements.