🔍 Read the full analysis: Amazon’s Tokenmaxxing Warning Raises A Bigger Question About AI Metrics on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Amazon published engineering guidance on October 27 warning that token use, AI adoption and other activity measures can distort engineers’ priorities when used as targets. The memo says those measures can still help identify patterns, but the focus should be on customer, business and team outcomes. It does not set out a replacement numerical scorecard.
Amazon published new guidance for its software engineers on October 27, warning that token counts and other measures of AI activity should not become performance targets. The memo says engineers should be judged by their impact on customers, the business and their teams, as the company tries to distinguish useful AI adoption from activity that simply raises usage.
The memo, published on Amazon’s own news site, was written by Terk Terkowitz, a vice president and distinguished engineer in the company’s stores business. It introduces a revised role guideline for software engineers, developed by Amazon’s most senior engineers. The document does not prescribe a fixed AI-use requirement, saying definitions of AI work could quickly become outdated.
It calls measures such as token usage, AI-tool adoption, code changes and code reviews “deeply flawed” when they are used as targets rather than as indicators. The memo warns that target-driven measurement can increase costs and system complexity. It also says these measures currently “correlate well” with outcomes that matter, preserving a distinction between using data to understand work and using it to score people.
The guidance asks leaders to examine both low and high AI adoption. Looking at the 10th percentile can help explain why some teams are not using tools effectively; examining the 90th percentile can test whether heavy use is producing better results. It also says engineers who find that current AI tools do not help their work should document that finding, rather than treat it as a performance failure.
Six months at Amazon
From token leaderboards to an outcomes memo
Amazon spent 2026 learning the same lesson twice: once with how its engineers use AI, once with its own AI models.
- May 2026Token leaderboards backfireFinancial Times (via The Decoder): staff give an internal agent tool pointless tasks to inflate token counts. Reported, anonymous sources.
- Late May 2026KiroRank “deprecated”Amazon spokesperson to Business Insider. SVP Dave Treadwell: “Please don’t use AI just for the sake of using AI.”
- 24 Jul 2026San Francisco AGI Lab site closesConfirmed by Amazon to GeekWire. Frontier model research continues under Pieter Abbeel.
- 28 Jul 2026Most flagship Nova models wound downBusiness Insider, people familiar: Premier, Omni, Reel, Canvas to “keep the lights on”. Amazon: models “remain one of the most important things we’re working on”.
- 14 / 30 Sep 2026Nova Sonic v1, Nova Reel, Nova Canvas reach end of lifeAWS Bedrock model lifecycle page. Requests to them now fail.
- 26 Oct 2026Anthropic files its S-1Amazon notes: $6.2 billion principal, 12% ownership cap that Amazon can waive. Anthropic commits at least $110 billion to Amazon through April 2036.
- 27 Oct 2026The memo: outcomes, not tokenmaxxingUsage and throughput metrics are “deeply flawed” as targets. “Tokenmaxxing is at odds with frugality.”
- 29 Oct 2026Amazon’s third-quarter resultsAnnounced by Amazon. Not yet published when this article went live.
Sources: Amazon engineering memo (27 Oct 2026); Business Insider (29 May and 28 Jul 2026); GeekWire (24 Jul 2026); The Decoder (12 May 2026, summarising the Financial Times); AWS Bedrock model lifecycle (read 28 Oct 2026); Anthropic S-1 (26 Oct 2026).
Powered by Thorsten Meyer AI
Why Amazon Is Reframing AI Productivity
The guidance addresses a measurement problem facing employers adopting generative AI: activity is easier to count than value. Token volumes, tool logins and code changes produce visible numbers, but do not by themselves show whether a product improved, a customer problem was solved or engineering work became more effective.
Amazon’s warning matters to employees because usage quotas or leaderboards can reward behavior that looks productive without improving results. It also matters to companies setting AI policies: the memo argues for using adoption data to spot questions and bottlenecks, while leaving the evaluation of impact to informed judgment. That approach offers no simple replacement score, and may be harder to compare across teams, but it avoids treating more AI activity as proof of better work.
The memo’s practical test is whether tools help with specific tasks and outcomes. It asks leaders to use AI on their own production work, identify where it helped and where it did not, and explain the difference. That makes the policy less a rejection of AI than a warning against confusing tool use with performance.
Products Worth Considering
Dell UltraSharp U2725QE 27-Inch 4K UHD IPS Black Monitor
Apple iPad Air 11-inch (M2), 512GB, Wi-Fi, Space Gray
Dell UltraSharp 27 Thunderbolt Hub Monitor (U2726DER)
As an affiliate, we earn on qualifying purchases.
From Internal Leaderboards to Official Guidance
The memo follows reports of internal efforts to track AI use. In May, the Financial Times reported that some Amazon employees assigned unnecessary tasks to an internal agent tool, MeshClaw, to raise token counts. The report also described a target for more than 80% of developers to use AI weekly and internal token-use leaderboards. Those details came from reporting based on employee accounts, not from the October memo itself.
At the end of May, an Amazon spokesperson confirmed to Business Insider that a leaderboard called KiroRank had been deprecated. The company described it as an informal tracker created by a group of employees and said it was never intended to encourage AI use for its own sake. Senior vice president Dave Treadwell also told staff: “Please don’t use AI just for the sake of using AI.” The new role guideline makes the principle part of official engineering guidance.
The source material also compares Amazon’s AI products and investments, but those developments are separate from the memo. It cites Business Insider reporting that most flagship Nova models were moved to maintenance, and Amazon lifecycle information showing several Nova services retired. It also cites Amazon’s reported Anthropic investment and a June 30 filing valuation. Those figures and product changes do not establish why Amazon wrote the guideline; the memo itself focuses on how engineers’ work should be assessed.
““Tokenmaxxing is at odds with frugality.””
— Amazon’s engineering memo
How Teams Will Measure Outcomes
The memo does not name a single replacement metric or explain how Amazon will compare outcomes across teams doing different work. It also does not say whether the company will change existing performance reviews, set new reporting requirements or discontinue all internal AI-use tracking. Its stated position is narrower: activity indicators may be useful, but should not become targets in place of outcomes.
The earlier reports of employees inflating token counts rely in part on anonymous accounts. Amazon’s confirmation that KiroRank was deprecated establishes that the tracker existed and was discontinued, but does not independently confirm every reported practice or how widespread it was. The memo does not quantify how many employees or teams were involved.
Amazon Results and Policy Follow-Through
Amazon is scheduled to report third-quarter results on October 29, two days after publishing the memo. Those earnings may provide information about the company’s overall financial performance, but the available material does not establish that Amazon will report engineering productivity or the effects of its AI tools in a way that directly tests this guidance.
The next practical test will be how engineering leaders apply the role guideline: whether teams use adoption data to investigate where AI helps, and whether reviews emphasize demonstrated impact rather than counts of tokens, code or tool use. The memo calls for leaders to test tools on production work and share what they learn. It does not provide a timetable or a formal audit process, so how consistently the approach will be applied remains to be seen.
Key Questions
What does Amazon mean by tokenmaxxing?
Tokenmaxxing means using an AI tool in ways that raise token counts without a corresponding work benefit. The memo presents it as an example of why usage should not be a performance target.
Is Amazon telling engineers to stop using AI?
No. The memo says AI adoption can correlate with useful outcomes and describes teams that have adopted AI as delivering more. Its warning is against scoring people on usage alone, not against using the tools.
What does Amazon say should count instead?
The guidance centers on impact on customers, the business and the team. It asks engineers and leaders to assess whether tools improve particular tasks and to document cases where current AI tools do not help.
Did Amazon confirm employees inflated token counts?
The May reporting described employees giving an internal agent unnecessary tasks, based partly on anonymous accounts. Amazon separately confirmed that the KiroRank leaderboard was deprecated and said it was an informal tracker not intended to promote AI use for its own sake. The October memo does not confirm how widespread token inflation was.
Source: ThorstenMeyerAI.com
