The tweak via the markdown file could be effective to a degree in helping enterprises taming costs around AI as they move to production, analysts say.

InfoWorld is a source of news, analysis, and commentary on technology trends, IT strategies, and business innovation. With a focus on enterprise technology and digital transformation, InfoWorld offers insights and guidance for IT decision-makers, software developers, and technology professionals. From  articles on cloud computing and cybersecurity to product reviews and industry trends, InfoWorld helps readers navigate the complexities of modern IT environments and make informed decisions to drive business success.

InfoWorld

A data analyst at Amazon published a markdown file (CLAUDE.md) on GitHub that claims to reduce Claude's output token usage by ~63% by imposing strict behavioral constraints on the model. The instructions eliminate sycophantic responses, boilerplate phrases, unsolicited suggestions, and stylistic quirks. At 100 prompts/day, estimated savings are ~$0.86/month; at 1,000 prompts/day, ~$8.64/month. Analysts note the approach offers real but modest operational benefits for enterprises running high-volume AI workloads, though actual enterprise savings are likely in single digits since output tokens are only a fraction of total inference costs.

A GitHub tinkerer teaches Claude to talk less, and that may matter more than it seems