🔍 Read the full analysis: Amazon’s Engineers Are Told To Stop Tokenmaxxing—Here’s The Bigger Point on ThorstenMeyerAI.com
Get the latest gadgets 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 guidance on October 27 telling software engineers not to treat AI token use or other activity metrics as performance goals. The memo says those measures can help track adoption but should not replace evidence of customer, business or team impact.
Amazon published new guidance for its software engineers on October 27, warning that token counts and other measures of AI activity should not be used as performance targets. The memo says engineers should be evaluated on impact for customers, the business and their teams, as the company faces questions about how it measures returns from its AI investments.
The guidance was written by Amazon’s most senior engineers and published on the company’s news site. It introduces a revised role guideline for all software engineers. The memo says AI tools are changing too quickly for a fixed definition of AI use to remain current, so it anchors the role in outcomes that are less likely to change: customer, business and team impact.
Amazon’s memo calls token usage, AI-tool adoption, code changes and reviews “deeply flawed metrics” when used as targets. These measures can serve as indicators, it says, but turning them into goals may increase costs and system complexity without improving the work. The memo gives two examples: “tokenmaxxing is at odds with frugality” and measuring lines of code is at odds with “invent and simplify.”
The guidance does not tell staff to stop using AI. It says activity metrics currently “correlate well” with results that matter, and that teams adopting AI are delivering more while teams that do not adopt it are falling behind. It asks leaders to examine both low and high adoption: understand why some teams are not using the tools, and check whether heavy users are producing better outcomes.
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
Amazon Shifts AI Performance Measures
The memo makes a distinction that matters beyond Amazon: AI use is not the same as AI value. Token counts can show that a tool is being used, but alone they do not show whether it saves time, improves software or benefits customers. Treating activity as a target can reward unnecessary work and raise computing costs.
For engineers, the guidance also gives room to report when current AI tools do not help a task. The memo says that finding should be documented and shared, rather than treated as a performance failure. For managers, it calls for firsthand use of the tools on production work, including examples of tasks where they help and where they do not.
The approach may matter to other companies setting AI targets for staff. Amazon’s position is not that usage data has no value; it is that such data should help identify patterns and questions, not act as a scorecard on its own. The company has not yet published evidence showing how the revised guidance will change evaluation or spending.
Products Worth Considering
Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
Apple iPad Air 11-inch (M2), 512GB, Wi-Fi, Space Gray
As an affiliate, we earn on qualifying purchases.
From Usage Leaderboards to Guidance
The memo follows reports of internal efforts to track AI use. In May, the Financial Times reported that some Amazon employees gave an internal agent tool, MeshClaw, unnecessary tasks to increase token counts. The reporting also described an AI-use target and internal token leaderboards. Those accounts came from anonymous sources and have not been confirmed in the memo.
Amazon did confirm a separate development at the end of May: a spokesperson told Business Insider that an informal employee-created leaderboard called KiroRank had been deprecated. The company said it was never intended to promote 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 October guidance formalizes the broader principle in an engineering role guideline.
The memo lands as Amazon’s AI strategy draws attention. Business Insider reported in July that most flagship Nova models had been moved to “keep the lights on.” Amazon’s lifecycle information lists Nova Canvas, Nova Reel and the first Nova Sonic as retired. Separately, the source material reports an approximately $18 billion investment in Anthropic and a June 30 filing value of about $190 billion for the holdings. Those figures describe different measures of the investment and its reported valuation, not a realized return.
““Tokenmaxxing is at odds with frugality.””
— Amazon’s engineering memo
How Amazon Will Judge Outcomes
The memo does not specify how the new guidance will affect individual reviews, compensation, promotions or team-level targets. It also does not provide a replacement formula for measuring AI’s effect on productivity, quality or customer experience. The advice is to use judgment and examine outcomes, but the company has not described a uniform method for doing so.
The earlier reports about engineers inflating token counts relied on anonymous accounts; they should not be read as a company-wide finding. Amazon’s memo acknowledges that adoption metrics correlate with results at present, while arguing they are inadequate as targets. How that relationship may change as tools and work practices evolve remains unknown.
Amazon’s third-quarter financial results were scheduled for October 29, one day after the memo’s publication. The source material provides no results, and the memo itself does not quantify the financial return from Nova or its Anthropic investment. The reported investment valuation is not the same as realized profit.
Amazon’s Results and Follow-Through
Amazon is due to report its third-quarter results on October 29. Those results may offer a broader view of the company’s finances, but they will not by themselves establish whether AI tools are improving engineering outcomes; the memo offers no specific AI-return metric or evaluation timeline.
For now, the stated next step is for engineering leaders to apply the guidance in their teams: examine low and high AI adoption, use the tools on real production work, and record where they help or fail to help. Whether Amazon later reports how those practices affect customer outcomes, costs or engineering performance is not yet clear.
Key Questions
What does Amazon mean by “tokenmaxxing”?
Tokenmaxxing means using an AI tool in ways that increase token consumption or improve a usage ranking without a useful work outcome. Amazon’s memo says token counts should not be treated as a performance goal.
Is Amazon telling engineers to stop using AI?
No. The memo says AI adoption currently correlates with important results and asks leaders to examine how teams use the tools. Its warning is against scoring people on usage alone.
What should engineers be measured on instead?
The memo points to long-term impact on customers, the business and the team. It does not set out a single replacement score or formula for measuring those outcomes.
Did Amazon confirm that employees inflated token counts?
The Financial Times reported accounts of employees giving an internal tool unnecessary tasks to increase token counts, based on anonymous sources. Amazon’s later memo criticizes token use as a target, but the supplied material does not show the company confirming those specific accounts.
When are Amazon’s next quarterly results due?
Amazon’s third-quarter results are scheduled for October 29, 2026, according to the source material. No results are included in the October 27 memo or the supplied reporting.
Source: ThorstenMeyerAI.com
