- MIT Technology Review argues companies should measure AI coding agents by cost per merged pull request, not cost per token.
- A single agentic coding task can use 10 to 50 times more tokens than a basic chat prompt because agents plan, generate code, run tests, and retry.
- Uber reportedly exhausted its entire 2026 AI budget by April after expanding its use of Claude Code, and GitHub Copilot has shifted from flat-rate pricing to metered credits.
- Gartner predicts AI coding token costs will surpass average developer salaries by 2028.
What Happened
MIT Technology Review published an analysis arguing that per-token pricing is the wrong way to measure AI coding agent costs, published November 2, 2026, proposing cost per merged pull request as a better metric.
Why It Matters
The piece lands as engineering teams grapple with the same cost pressures seen in Anthropic’s own usage-limit guidance for Claude Code, arguing that blanket spending caps restrict valuable workflows along with wasteful ones.
Technical Details
A single agentic coding task can consume 10 to 50 times more tokens than a basic chat prompt because agents plan, generate code, run tests, and repeat the cycle to fix errors. Uber reportedly exhausted its entire 2026 AI budget by April after expanding Claude Code use, GitHub Copilot moved from flat-rate pricing to metered credits, and Gartner projects AI coding token costs will exceed average developer salaries by 2028.
Who’s Affected
Engineering leaders setting AI tooling budgets, developers hitting usage caps mid-task, and vendors like GitHub and Anthropic adjusting their pricing models all face the same measurement problem.
What’s Next
The article recommends tracking cost per merged pull request and rework rates instead of raw token counts, so teams can distinguish wasted retries from productive agent work before setting spending limits.
