🔍 Read the full analysis: Amazon’s Warning To Engineers Makes AI Outcomes The Bigger Story on ThorstenMeyerAI.com
Get hardware and tech essentials 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 updated guidance for software engineers on October 27, warning that token counts and other AI-use metrics can become harmful targets when disconnected from results. The memo still says AI adoption metrics correlate with outcomes today, but directs managers to examine whether heavy and light users are actually benefiting.
Amazon published new guidance for its software engineers on October 27, warning that token counts and other measures of AI use should not become performance targets. The memo says teams should judge work by its impact on customers, the business and colleagues, as the company prepares to report quarterly results and continues to develop and deploy AI products.
The memo, published on Amazon’s news site, was written by Terk Terkowitz, a vice president and distinguished engineer in the company’s stores business, and reflects a revised role guideline for software engineers. It says measures such as AI-tool adoption, token use, code changes and code reviews can serve as indicators, but are “deeply flawed” when used as targets. In the memo’s view, rewarding inputs rather than results can increase costs and complexity.
The guidance does not tell engineers to stop using AI. It says these input metrics currently “correlate well” with outcomes that matter, and that teams adopting AI are delivering more while teams not adopting it are falling behind. The proposed distinction is between monitoring usage to find patterns and scoring workers on usage alone.
Managers are told to examine both low and high adoption. The memo recommends looking at engineers around the 10th percentile of AI adoption to learn what is preventing effective use, and the 90th percentile to check whether high usage is producing better work. It also says engineers should document when current AI tools do not help, and that leaders should test the tools on their own production work.
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 AI Usage Is Not the Score
The guidance addresses a basic problem for employers adopting generative AI: usage is easy to count, but business value is harder to establish. Token totals can show that a tool is being used, yet do not by themselves show whether a customer received a better service, whether work became faster, or whether the result was worth its cost.
That distinction matters for employees as well as managers. If a company rewards AI activity as an end in itself, workers may have reason to generate unnecessary prompts or code. Amazon’s memo says such measures can encourage higher costs and more complicated systems. Its alternative is not a new score, but judgment about the task and its result, including a documented explanation when a tool does not help.
The memo also gives Amazon a public way to describe how it wants AI adoption assessed across its engineering organization. It is a statement of internal guidance, not evidence that every team has changed its incentives or that the approach has already improved results. The practical effect will depend on how managers apply the instructions.
Products Worth Considering
Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, and Liquid Retina Display
Logitech MX Keys S Wireless Keyboard, Graphite
Logitech MX Mechanical Full-Size Wireless Mechanical Keyboard, Tactile Quiet Switches, Backlit Keys, Bluetooth, USB-C, Multi-Device, Graphit
As an affiliate, we earn on qualifying purchases.
From Usage Leaderboards to Role Guidance
The memo follows earlier reports about internal tracking of AI use. In May, the Financial Times reported that some Amazon employees gave an internal agent called MeshClaw unnecessary tasks to raise token counts. The report also described a weekly AI-use target and internal leaderboards. Those accounts relied on anonymous employees and should be treated as reported claims, rather than independently confirmed details.
Amazon did confirm one related development. At the end of May, a company spokesperson told Business Insider that a leaderboard called KiroRank had been “deprecated”. Amazon said it was an informal tracker created by a group of employees and was not 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 memo takes that message further by putting outcome-focused guidance into the company’s engineering role framework.
The source article also compares the memo’s emphasis on outcomes with Amazon’s AI investments. It reports that Business Insider said in July that most flagship Nova models had been moved to “keep the lights on,” while Amazon’s lifecycle page lists several Nova products as retired. Separately, Amazon invested in Anthropic; the source says an Amazon filing dated June 30 carried the holding at about $190 billion, compared with roughly $18 billion invested. Those comparisons concern distinct investments and do not establish that the memo was prompted by either outcome.
““Tokenmaxxing is at odds with frugality.””
— Amazon engineering memo, published October 27
How Teams Will Apply the Guidance
The memo sets out a principle and suggested management practices, but it does not specify a single replacement metric or explain how Amazon will evaluate individual engineers under the updated guideline. It is also unclear how widely the reported token-count behavior occurred, how many teams used the earlier trackers, or whether those practices affected formal performance reviews.
The memo’s statement that adoption measures correlate with results is Amazon’s own assessment; the published guidance does not provide underlying figures or a defined measurement period. Nor does it establish that high AI use caused stronger results. The source material also does not show whether Amazon’s reported Nova product changes or Anthropic investment valuations played a role in the decision to revise the engineering guidance.
Amazon’s Results and Internal Follow-Through
Amazon is scheduled to report its third-quarter results on October 29, the day after the memo’s publication. Those results may provide information about the company’s broader business performance, but they will not by themselves show whether the new engineering guidance has changed how teams work or whether AI use is improving customer outcomes.
The next evidence to watch is how managers put the instructions into practice: whether they assess results alongside usage data, review why some engineers use AI less, and test whether heavy users are producing better work. The memo says leaders should use the tools themselves and share examples of tasks where AI helped or did not. Amazon has not, in the source material available, announced a timeline or public reporting plan for measuring the guidance’s effect.
Key Questions
What did Amazon tell its engineers?
Amazon’s October 27 memo says engineers should not be judged on token counts, AI-tool adoption, code changes or reviews alone. It says results for customers, the business and the team should remain the focus.Is Amazon telling engineers to stop using AI?
No. The memo says AI adoption metrics currently correlate with important outcomes and that teams not adopting AI are falling behind. Its warning is against using activity measures as goals in themselves.What does “tokenmaxxing” mean in this story?
It refers to increasing AI token use for the sake of a usage count or leaderboard, rather than because the task needs it. The memo says that approach conflicts with Amazon’s principle of frugality.What remains unknown about the new guidance?
Amazon has not specified one replacement metric, explained exactly how the guidance will affect performance reviews, or published evidence showing whether it has changed work or customer outcomes.When will Amazon report its next quarterly results?
The source says Amazon is scheduled to report its third-quarter results on October 29, 2026. The results may provide business context but will not alone measure the effect of the engineering memo.Source: ThorstenMeyerAI.com
