Claude Code ships with a verb spinner — a rotating list of whimsical gerunds that flash in orange while the model processes your requests. Shenaniganing. Zesting. Smooshing. Nucleating. Metamorphosing. If you've spent any time with the terminal-based coding assistant, you've seen them.
Developer jdlms figured out how to turn it off, and the post has been making the rounds this week. The fix is straightforward: open ~/.claude/settings.json, add a spinnerVerbs block with mode set to 'replace', and put a single blank space string in the verbs array. Anthropic's curated list of terminal whimsy gets swapped out for nothing. The spinner mechanism stays intact; the words disappear. For the opposite effect, 'append' mode lets you bolt your own vocabulary onto the existing list.
The post spread fast, partly because the workaround is so clean, and partly because it validated something a lot of developers had been quietly annoyed about. The verb spinner sits squarely in the 'AI should have personality' school of design — make the tool feel playful, reduce the intimidation factor. It's a reasonable instinct. It also doesn't always survive contact with someone running their fortieth refactor of the afternoon.
The more interesting detail is that the escape hatch exists at all. Anthropic built in a settings override for a feature it presumably wanted users to see, then left it largely undocumented. That's a small but telling gap between what a product team ships and what it actually expects power users to tolerate.