HaiAI123

Curated Global AI Tools Directory

Industry update
Flat

Why multi-agent workflows blow up token costs

11/20/2026 — 11/20, 11:16·1 sources (all)·1 reports (all)

Key takeaways

  • An analysis runs the numbers on a five-step multi-agent workflow: without changing any business logic, simply letting the model decide who a…

Background & analysis

On November 20, 2026, the AI section of DEV Community published an article titled "Why Multi-Agent Systems Get More Expensive — Routing, Handoffs, and Quadratic Cost," which runs a concrete cost calculation for a multi-agent workflow. The piece sets up a five-step multi-agent workflow and shows that with not a single line of business logic changed, handing only the decision of which agent goes next over to a model pushed the token overhead from 750 to 18,000 tokens. The starting figure of 750 comes from five hops multiplied by 150-token summaries, and the extra 17,250 tokens all trace back to one mechanism: everything each agent produced was re-read by every agent that came after it. The author explicitly notes that nobody wrote a bug, so the growth is not being treated as a coding mistake, and the article then lays out four ways to trim that part of the spend. At this point, the available material verifies only this single article and the numbers it presents. There is no reported follow-up on whether any vendor has acted on the findings, no independent confirmation of the measurement, and no evidence yet on how well the proposed cuts would work in practice. The event therefore currently rests at the stage where the analysis and the reduction options have been published, with no subsequent development on record.

AI-generated from 1 reports · updated 2 days ago

updream
24-hour heatHeat index 25 · peak 50 · 23h ago
24 hours agonow

Reports on this story headlines open the original

Nov 20
  1. A five-step multi-agent workflow kept its business logic unchanged and only handed the routing decision to a model, yet token overhead jumped from 750 to 18,000. The extra 17,250 tokens come from one source: every agent's output is re-read by all the agents after it. The post walks through the math and gives four cuts that bring the cost down.

    DEV Community · AIAI score 57

Other stories people are talking about

How is heat calculated?About the method

Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.

This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.

Surge
Discussion rising fast
New
First report within 6 hours
Rising
Still gathering discussion

Back to AI industry updates →