# Amazon Just Told Engineers to Stop Gaming Its AI Usage Leaderboard

> Some Amazon engineers were running AI agents on pointless tasks just to climb an internal ranking. A new internal memo says that if a job doesn't actually benefit from AI, that's useful information, not a failure.

- Source: Morning Week
- Canonical URL: https://morningweek.com/article/amazon-just-told-engineers-to-stop-gaming-its-ai-usage-leaderboard
- Author: Morning Week Newsroom
- Section: Tech
- Published: 2026-10-29T11:25:51.279Z
- Updated: 2026-10-29T11:25:51.279Z
- Tags: Amazon, AI, workplace, engineering culture, Claude, productivity metrics

---

Amazon has a new message for its engineers about artificial intelligence, and it runs counter to the instinct many workplaces have been pushing: more AI usage is not automatically better. In a memo revising the company's internal AI guidelines, Terk Terkowitz, a vice president and distinguished engineer at Amazon Stores, pushed back against the idea that heavier AI usage should be treated as a goal in itself.

## How the gaming started

The behavior the memo is responding to had a name inside the company: tokenmaxxing. Some engineers discovered that running AI agents on low-value or essentially meaningless tasks inflated their token usage numbers, which in turn boosted their position on an internal leaderboard called Kirorank. Amazon shut the leaderboard down in May, citing concerns that the unnecessary computing activity was driving up costs without producing anything of real value.

## A finding, not a failing

Terkowitz's memo reframes what it means when a given piece of work doesn't lend itself to AI assistance. Rather than treating low AI usage as a red flag, he argues it should be read as useful information about the nature of the task itself, writing that if the work does not benefit from current AI tooling, that is a finding, not a failing. The distinction matters because it removes the implicit pressure for engineers to force AI into workflows where it does not actually help.

## The 'Renaissance Developer'

In place of raw usage volume, the memo promotes what it calls the Renaissance Developer, an engineer who applies AI tools with judgment rather than defaulting to maximum usage on every task. Terkowitz warned that treating metrics like token counts and lines of code generated as targets in their own right risks driving up costs and runs against Amazon's longstanding internal principle of invent and simplify, a core tenet the company has used for years to guide engineering decisions.

## Amazon isn't alone in confronting this

The memo lands amid a broader reckoning across the tech industry about what unconstrained AI usage actually costs. One unnamed company reportedly spent 500 million dollars on Claude usage in a single month after failing to set any usage limits. Shopify CEO Toby Lutke has described a related problem from the other direction, with employees producing what he called slop grenades, meaning AI-generated content sloppy enough that it creates more review work than it saves.

Amazon's new guidance treats AI usage less as a performance metric to maximize and more as a tool whose value depends entirely on the judgment of the person deciding when to reach for it.

Amazon is not abandoning measurement altogether. Managers are still expected to compare usage patterns across their teams, but the memo reframes the purpose of that comparison, shifting it from rewarding raw volume toward asking whether higher AI usage is actually correlating with better outcomes. For a company that has spent years optimizing internal metrics down to the smallest detail, the shift amounts to a quiet admission that some numbers are easier to game than they are to trust.

---

Originally published by Morning Week. Free to cite with attribution and a link to https://morningweek.com/article/amazon-just-told-engineers-to-stop-gaming-its-ai-usage-leaderboard.