MIT Review · Note ·
Stop counting tokens, start counting what ships
When a developer gives an AI agent a coding task, the agent does more than answer once. It plans, generates code, runs tests, and repeats the process to…
What moved
When a developer gives an AI agent a coding task, the agent does more than answer once. It plans, generates code, runs tests, and repeats the process to correct errors. A single task can use 10 to 50 times more tokens than a basic chat prompt. The more useful the agent becomes, the more calls…
Why it matters
Filed as a public note from MIT Review, 2026-11-02. The desk tags this brief LLM / Agents.
On the record
- Filed from the MIT Review official RSS on 2026-11-02.
- Primary source host: technologyreview.com.
- When a developer gives an AI agent a coding task, the agent does more than answer once.
Primary source
technologyreview.com
technologyreview.com
Desk
Logged as brief 6946
Logged as brief 6946
Cite this brief
Stop counting tokens, start counting what ships (MIT Review, 2026-11-02) — https://labledgerdesk.pages.dev/b/2026/11/02/stop-counting-tokens-start-counting-what-ships/