According to Anthropic's current documentation on chaining complex prompts for Claude, what is described as the most common prompt-chaining pattern, and how is it structured across separate API calls?
- Load balancing: send the exact same prompt to several different Claude API keys simultaneously and return whichever response arrives first, discarding the rest
- Weight merging: average the model weights used to generate two separate draft responses into a single hybrid model before producing the final answer
- Cache warming: repeatedly resend an unchanged prompt purely to keep it available in the prompt cache, with no reviewing or refining step involved
- Self-correction: generate a draft response in one API call, have Claude review that draft against stated criteria in a second call, then have Claude refine the draft based on that review in a third call, so each step is a separate call that can be logged, evaluated, or branched on
Why D? And why not the others?
Correct answer: D. Self-correction: generate a draft response in one API call, have Claude review that draft against stated criteria in a second call, then have Claude refine the draft based on that review in a third call, so each step is a separate call that can be logged, evaluated, or branched on
Anthropic's documentation states that explicit prompt chaining, breaking a task into sequential API calls, remains useful when you need to inspect intermediate outputs or enforce a specific pipeline structure, and it identifies self-correction as the most common chaining pattern: generate a draft, have Claude review that draft against stated criteria, then have Claude refine the draft based on the review, with each step run as its own separate API call so results can be logged, evaluated, or branched on. The second option is wrong because sending identical requests to multiple API keys and keeping the fastest response is a redundancy or load-balancing pattern unrelated to improving reasoning quality, and it is not the documented chaining pattern. The third option is wrong because the Messages API does not expose model weights for merging, and no such technique is described. The fourth option is wrong because it describes maintaining a prompt cache with no reviewing or refining step, which is a caching concern rather than the review-and-refine chaining pattern the documentation describes.
Source: Anthropic, "Prompting best practices" -- "Chain complex prompts," platform.claude.com prompt engineering documentation