Are you steering towards AI burnout? Afraid of loosing your job to someone with little programming skills and a huge Claude account? Disappointed about the code quality in your projects? This post addresses how to keep enjoying programming while integrating LLMs. There are legitimate ethical concerns about frontier LLMs run by big tech companies, but this post focuses on the practical impact on the developer experience. A key theme is the degradation of the programmer's role from actor to cog in a machine. As a Haskell programmer, the author enjoys the process of expressing thoughts in code, a sentiment likely shared by many.

When generating code, much of that enjoyment is at risk. The solution is not to abstain entirely, but to use LLMs for "nearly everything else" — the boring, nuisance tasks — rather than the core coding. Keeping writing code is essential to avoid becoming an LLM wasteland and to maintain skills, as there is a low threshold to handing work over to agents.

After weeks of not coding, returning becomes difficult. LLMs are worse at producing good, human-readable code than advertised. The goal is to become moderately more productive without losing the joy of programming, ensuring the codebase remains understandable to humans, not just agents.