Getting the most out of Opus 5.5 in Claude and Claude Code
Anthropic published a playbook detailing how to prompt and steer Opus 5.5 for long-running, multi-step coding tasks in Claude Code.
summary is enough It outlines useful prompt adjustments and configuration patterns for the new model, though the core takeaways can be scanned quickly.
summary
Anthropic's guide explains how to adapt prompting styles for Opus 5.5's built-in adaptive thinking and longer execution windows. It recommends removing manual reasoning triggers, defining clear completion states, and setting explicit execution boundaries in project configuration files. The playbook also details how to delegate complex work to parallel subagents and manage safety-filter fallback behaviors. Commenters strongly praise the model's multi-step autonomy while cautioning about high token costs.
key points
- Remove explicit 'think carefully' prompts since Opus 5.5 features native adaptive thinking.
- Use configuration files like CLAUDE.md to set clear boundaries for when the model should pause or continue autonomously.
- Delegate large-scale code migrations and audits to coordinated parallel subagents to maximize efficiency.
who should read it
Software engineers using Claude Code for complex, long-running development tasks.
discussion warm
consensus
Commenters overwhelmingly praise Opus 5.5 for its unprecedented capability in handling long-horizon, multi-step coding and engineering tasks.
pushback
Some users report issues with high token consumption, overeager execution past boundaries, and spurious security flags.
It’s a really good model. Over the past few days, I give Opus some general directives to basically speed up our CI, and telling it I care both about billing minutes and wall clock time. I told it to create a plan after analyzing everything in our CI, run the plan by a Fable subagent, and then focus on low-risk, high-reward changes.
9 hours later, I had 12 PRs ready to be merged, and the net result is CI time has dropped from ~10 minutes to ~4 minutes, and billing minutes have dropped around 60%. Less than an hour of my attention.
It details a concrete real-world success story of the model autonomously completing a major multi-PR CI optimization over nine hours.
Generated by an LLM from the article and thread. Check the sources before relying on it.