I Think I Have LLM Burnout
Alec Scollon
A developer says heavy daily LLM use is starting to wear on him—not from bugs, but from the writing itself. Same hallucinations, same clipped tone, same emoji spam, every single time.
Based on reporting by Alec Scollon — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
There's a specific kind of exhaustion that comes from repetition, and one dev writing for TLDR Dev says he's hit it. He's not anti-AI. He spends hours a day with Claude Code at work and Codex at home, reviewing and revising whatever the assistant hands back rather than blindly accepting it. His job has quietly shifted shape: less writing code from scratch, more designing it, describing that design to a model, and then combing through what comes out before he ever touches the keyboard himself.
The bigger project on his plate right now is building a framework for large-scale, unsupervised code generation, which means his days are split between working with Claude to build tooling and sifting through output from another model, Qwen, running without supervision. Either way, he's reading machine-generated text almost constantly. Even his search habits have changed—he'll default to ChatGPT or a Gemini overview instead of clicking through links, partly because plain web search has gotten cluttered with low-quality AI-written articles. He still falls back to old-fashioned browsing when the answer is wrong, but for casual questions the shortcut usually holds up.
What's changed isn't his workflow, though. It's how he feels walking into it. He describes a creeping dread before he even opens the output, because he already knows what he's going to get: confident wrong answers, choppy overconfident fragments, and a scattering of unnecessary emoji. None of these things is a dealbreaker by itself. Stacked together, day after day, they've made him tired of LLM writing in a way that surprised him.
He's careful to draw a line between frustration at a flaky tool, which he understands, and irritation at a writing style, which he didn't expect to bother him this much. Humans make mistakes too, he points out, and are plenty unreliable on their own. But people don't all make the exact same mistakes in the exact same voice. LLMs do, and personalization settings only smooth over so much of it—especially once other people's AI-generated content enters the mix, where he has zero control over the style at all.
He doesn't have a fix. He's not quitting the tools, since he still feels more productive with them and thinks learning to use them well is worth the effort. He's just bracing for more of the same, teeth gritted, hoping the feeling passes before it curdles into something worse.
My take — AI-written commentary, not fact-checked reporting
This is the tell nobody wants to admit: the tools are useful enough to keep using and annoying enough to dread. That combination is exactly how burnout works, in software or anywhere else—not one big failure, but the same small friction repeating until it grinds a person down. Vendors chasing benchmark wins should worry more about voice fatigue; nobody churns over a wrong answer as fast as they churn over reading the same tic for the thousandth time.
Read more about this at: Alec Scollon