Stop counting tokens, start counting what ships
MIT Technology Review · 2 Nov 2026, 23:10 AEST
AI agents in coding tasks can significantly increase token usage, leading to financial pressures for companies. A shift in focus from token counting to measuring the value of software shipped is essential for optimizing AI development costs.
Key facts
- A single task can use 10 to 50 times more tokens than a basic chat prompt.
- Uber reportedly exhausted its entire 2026 AI budget by April after expanding its use of Claude Code.
- Gartner predicts that AI coding token costs will surpass average developer salaries by 2028.
- A better measure is cost per merged pull request: total AI-assisted development cost divided by the number of changes that pass review and are merged.
- With owned capacity, most of the cost is committed upfront, so each additional call gets cheaper as usage grows.
Read aloud
A I agents in coding tasks can significantly increase token usage, leading to financial pressures for companies. A shift in focus from token counting to measuring the value of software shipped is essential for optimizing A I development costs. A single task can use ten to fifty times more tokens than a basic chat prompt. Uber reportedly exhausted its entire twenty twenty six A I budget by April after expanding its use of Claude Code. Gartner predicts that A I coding token costs will surpass average developer salaries by twenty twenty eight. A better measure is cost per merged pull request: total A I-assisted development cost divided by the number of changes that pass review and are merged. With owned capacity, most of the cost is committed upfront, so each additional call gets cheaper as usage grows.
Sources in this cluster
- Stop counting tokens, start counting what ships
MIT Technology Review · 2 Nov 2026, 23:10 AEST