Back to Latest SignalMIT TECHNOLOGY REVIEW / RESEARCH

Stop counting tokens, start counting what ships

Source: MIT Technology Review · NOV 02, 2026

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…

IN THIS STORY
  • AI agents

Sources

  • Read the original at MIT Technology Review

On Bluesky

  • Agentic coding burns 10 to 50 times more tokens than chat; HPE urges tracking cost per merged pull request instead of rationing

    SkynetAndChill.com (@druce.ai) Nov 2, 2026

Related on Frontier

  • DeepSeek open sources Super-Agent, a coding agent research prototype it says proactively monitors developer group chats for coding issues and suggests solutions (Debby Wu/Bloomberg) Nov 1, 2026
  • OpenClaw release-publish/a3df6c6c8118-1793383326: fix(chat): older history stops loading from legacy transcripts (#188488) is available Oct 30, 2026
  • Qodo builds context into AI code review with Codex Nov 1, 2026
  • Google’s satellite AI server transmits its first response to Earth Oct 30, 2026
MORE LATEST SIGNALSIGNALS