Stop counting tokens, start counting what ships
MIT Technology Review
AI coding agents consume significantly more tokens than basic chat prompts because they iteratively plan, code, and test.
When developers assign coding tasks to AI agents, the process involves planning, generating code, and running tests repeatedly to fix errors. Because of this iterative workflow, a single task can utilize 10 to 50 times more tokens than a standard chat prompt. As these agents become increasingly useful, the volume of API and token calls grows substantially.
Why it matters: The high token consumption of AI coding agents highlights a shift in resource demands as autonomous tools become more capable and complex.
