Prompt Optimization
Systematically testing and improving prompts based on results — treating prompts like code that needs iteration, not one-time writing. This means running the same prompt against test cases, measuring how often it gives the right answer, changing something, running it again, and comparing. The difference between a prompt that works 60% of the time and one that works 90% of the time is usually a few rounds of this kind of deliberate testing.
In practice
Your Claude-powered classifier is 78% accurate. You systematically rewrite the prompt — testing different instruction phrasings, adding examples, adjusting tone — and test each version on 100 labeled examples. After six iterations, you're at 91%. That iterative improvement process is prompt optimization.
Related concepts
Where Prompt Optimization shows up
2 articlesMost prompt failures come from one of five fixable problems. Here's a diagnostic framework for figuring out what went wrong — and how to fix it without starting from scratch.
On September 8, 2026 Anthropic published its own cost-reduction guidance for the Claude Platform, and the surprising part is how much of it is about deleting text. Prompts written for 2024-era models contain instructions that now make frontier models slower, more expensive, and sometimes worse. Anthropic's case studies show 52–73% cost reductions with accuracy held flat or improved.