Worth of coding, in my mind

Short summary

A 20-year veteran developer weighs the productivity wins of LLMs like Claude against the risk of skill atrophy. They've noticed that AI assistance has eroded their ability to read unfamiliar code critically, search for solutions independently, and maintain mental context—capabilities built over two decades. They're now deliberately limiting AI in personal projects to preserve this judgment, accepting slower work as the price of maintaining expertise.

• •Developer with 20 years experience feels less capable after 1.5 years of heavy LLM use
• •Claude can complete a half-day of work in 20 minutes, but productivity gain comes at a skill-atrophy cost
• •Choosing to code manually on personal projects to preserve ability to evaluate code quality and maintain independent problem-solving

Generated with AI, which can make mistakes.