I'm sick of Claudisms, & what will happen next in AI-boosted software dev

A frustrated software developer protests the current mandate-heavy push to use large language models, calling out "Claudisms" - the bloated, over-commented, em-dash-heavy prose and convoluted code that some models (notably Claude) produce. Management drives widespread adoption by celebrating raw output and velocity without inspecting readability or maintainability, so teams ship frequent but low-quality changes: long, messy PR descriptions, overly complex generated code, and inlined commentary that obscures design and process knowledge. That creates developer fatigue and a growing backlash from engineers who value clear, maintainable code over blindly increased throughput.

The writer predicts this phase is a peak on an AI-assisted development bell curve and expects a pullback: rising inference costs (potentially 5-10×), visible drops in long-term maintainability, and enough developer rebellion to force executives to limit AI use. The likely equilibrium preserves AI as a useful tool but discourages indiscriminate reliance, with some "true AI believers" continuing heavy usage. The piece warns that continued mandates could drive experienced engineers away, urges practitioners to keep their code quality high despite pressure, and expresses cautious optimism that sensible balance will return.