
Refining prompts for newer Claude models
Anthropic suggests dropping phrases like think hard from prompts because current reasoning models already deliberate automatically. Stating your exact objective, defining the output format, and setting check-in points will generate faster responses without sacrificing output quality.
Try it yourself
- 1Remove phrases like think hard or take your time from your saved prompts.
- 2Define your target objective and specify the exact structure you want for the final response.
- 3Add checkpoints that tell the model when to pause and ask for your input.
- 4Use the standard Medium reasoning setting for routine tasks and switch to High only when the model struggles.
The Blend
Anthropic has advised users to adjust how they speak to its newer AI editions, suggesting that common habits such as ordering an assistant to think carefully before replying are no longer helpful. Modern systems now evaluate complex problems automatically behind the scenes, making extra coaxing redundant and potentially slower. The company suggests focusing instead on concrete deliverables, exact structural preferences, and interim progress reviews throughout longer assignments.
For regular office workers, this evolution means less time spent crafting elaborate, superstitious instructions just to get a reliable result. Direct briefs help the software produce usable summaries, spreadsheets, and slide decks without meandering explanations. Furthermore, keeping prompts concise can lower overall running costs, because the software expends fewer computing units parsing unnecessary guidance.
It remains uncertain whether users will comfortably abandon the rigid prompting formulas they learned over the past few years. As different technology vendors build distinct internal reasoning mechanics, navigating subtle etiquette shifts across competing workplace platforms might become a lingering headache for regular employees.
Written independently by AI News Smoothie from the reporting listed below. Facts belong to the original publishers. Follow the links for their full coverage.
Ingredients
- Which Claude model is best to use for knowledge work | Claude by Anthropic
Anthropic recommends providing advanced assistants with clear project briefs rather than coaxing phrases, since recent systems deliberate internally without prompt coaching.