← All news·2026-10-04·3 min read

Anthropic Shares Tips for Working With Opus 5.5 in Claude Code

Anthropic published a guide on working with Opus 5.5 in Claude Code, covering how to phrase tasks, when to stop, and how to handle code review.

aiclaude-codeanthropic

Anthropic released a tutorial on how to get the most productivity out of their latest model, Opus 5.5, in Claude Code. This is the mode where you get a coding assistant that makes changes to your code and runs commands. According to one user, Opus 5.5 can find bugs in code better than earlier versions of Opus. That holds true even when Opus was running at its highest effort level. On top of that, this new version has started pausing to think through its answers. So you no longer need to remind it to "think." Their main piece of advice? Lay out the whole task clearly in a single prompt, along with a description of what success looks like. Do not feed it partial information in pieces across multiple prompts.

This practice matters. Give the AI agent all the information it needs for the task in one prompt, and describe the finish line clearly. That is the key to getting the most out of an AI agent when you let it run autonomously for a long stretch. Otherwise you will likely waste time on unnecessary iterations or silent mistakes. I would recommend adding a rule to your project's instruction file that spells out when the AI agent should keep working without stopping, and when it should pause and bring in a human. That way you only get interrupted when it is truly necessary. Finally, there is a "quick response" feature. It costs more per response, though, so I would not switch everything over to it by default.

Source: claude.dev

Free course

Stop reading about AI — start building with it

The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.

Start free →
EAEvgenii Arsentev

Author

Evgenii Arsentev

PhD · Founder & CEO, Dad4Kids