I Spent $23 Testing the “YAML Saves Tokens” Hack. It Doesn’t Work

This article debunks the idea that switching from Markdown to YAML significantly reduces LLM token costs. Through controlled experiments, it shows that YAML often increases prompt size due to structural overhead, and any marginal savings are negligible compared to the dominant cost of large system prompts. The key takeaway is that meaningful optimization comes from trimming system context, compressing specs, and improving caching—not changing file formats.