Anthropic's Cherny Says Structured Prompting Is No Longer Necessary
Boris Cherny, creator of Claude Code, said on X that the latest Claude models no longer need elaborate structured prompts for most tasks.
Boris Cherny, who created Anthropic's AI coding tool Claude Code, said on X on October 6, 2026 that elaborate, highly structured prompting is no longer necessary for most tasks with the latest Claude models. He advised users to talk to Claude the way they would talk to a coworker.
Cherny said a prompt should communicate three things: what the AI should do, how much effort to spend, and how to verify the result. According to Business Insider, he shared a real prompt in which he asked Claude to build an artifact based on a podcast he called amazing, using sufficient tokens and creating visualizations better than those of The New Yorker or The New York Times.
Cherny said Claude Opus 5.5 has good taste and allowed the model to decide on its own whether to include watercolor graphics, which appeared in beige and orange tones in the output.
Back in the Sonnet 3.5 days, prompts mattered a lot, according to Cherny. Users previously shared various phrasings to get more out of chatbots.
Anthropic recommends structuring input and specifying verification methods for complex tasks involving long documents or multiple sources.
Quick answers
What did Boris Cherny say about prompting Claude?
He said elaborate, highly structured prompting is no longer necessary for most tasks with the latest Claude models, and that users should talk to Claude as they would to a coworker.
What three things should a prompt communicate according to Cherny?
What the AI should do, how much effort to spend, and how to verify the result.
Does Anthropic still recommend structured prompts?
Anthropic recommends structuring input and specifying verification methods for complex tasks involving long documents or multiple sources.