A/B testing feature flags
A user reported a frustrating experience with Anthropic's Claude Code, where a new, unannounced feature caused confusion and disrupted the user interface. The feature, identified as 'tengu_prompt_suggestion,' appeared to be an A/B test that injected responses as if they were from the user. The user managed to disable the feature by manually editing a configuration file, but noted it would likely reset upon restarting the application. AI
IMPACT A poorly implemented feature flag in a coding assistant can degrade user experience and trust in AI tools.