AI coding agents consume 10 to 50 times more tokens than single-response tasks

When developers give AI agents coding tasks, the agents plan, generate code, run tests, and repeat to fix errors, with a single task consuming 10 to 50 times more tokens than a one-off answer.

Read the full story at MIT Technology Review AI ↗

Timeline · 1 report

2026-11-02 13:10 · MIT Technology Review AI
Stop counting tokens, start counting what ships

More stories

Introducing the Grid Protectors Program — OpenAI News
Anthropic reports $17B quarterly revenue, shrinking losses in fresh IPO filing — Reuters AI
Marvell to acquire optical components maker Lumentum in $115 billion all-stock deal — DigiTimes Asia
OpenAI to merge ChatGPT's Chat and Work tabs into a single desktop experience — TechCrunch AI
OpenAI weighs rebranding data centers as 'compute campuses' amid growing public backlash — The Verge AI
October 2026 AI roundup highlights local LLM releases and Google announcements — Google AI Blog
Anthropic limits Claude Code to prevent recreation of existing video games — Engadget
Claude Code ecosystem expands with hooks, multi-agent workflows, and a security scare — Microsoft Research Blog

Get the daily brief of stories like this at 6:30 every morning →