🔍 Read the full analysis: Amazon’s AI Bets Offer A Different Lens On Engineering Metrics 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 engineering guidance on Oct. 27 saying AI usage measures such as token counts should not be targets for evaluating engineers; customer, business and team impact should count instead. The memo comes as reports describe a reduced role for several Nova models and Amazon’s investment in Anthropic has risen substantially in reported value, though the company’s third-quarter results had not yet been released.
Amazon’s engineering leadership told software engineers on Oct. 27 not to treat AI token counts, tool adoption or code volume as performance goals, arguing that evaluation should center on impact for customers, the business and teams. The guidance arrives as Amazon’s own Nova model effort and its investment in Anthropic present contrasting results, although key financial and product details remain incomplete.
The revised role guideline was published by Amazon and written by senior engineers. In a memo, Terk Terkowitz, a vice president and distinguished engineer in Amazon’s stores business, said input measures—including AI tool adoption, tokens used, code changes and reviews—can be misleading when set as targets. The memo argues that such goals can raise costs and add complexity without proving that customers or the business are better served.
The guidance does not say that AI tools should go unused. It says these measures currently correlate well with important outputs, while cautioning that correlation is not a reason to score employees by usage. Leaders are asked to examine both low and high adoption: understand why some teams use AI less, and determine whether heavy users are producing better results. Engineers whose work does not benefit from current tools are asked to document that finding rather than treat it as a failure.
The company’s own AI portfolio provides a broader backdrop. Business Insider reported in July that most flagship Nova models had been moved to a “keep the lights on” status. Amazon’s lifecycle page lists Nova Canvas, Nova Reel and the first Nova Sonic as retired. Separately, Amazon’s June 30 filing valued its Anthropic holdings at about $190 billion, compared with roughly $18 billion invested, according to the source material. Those figures are reported valuations, not cash proceeds.
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 Rejecting Usage Scores
The memo addresses a common management problem as companies adopt generative AI: activity is easy to count, while value is harder to measure. Token volume and tool adoption can show that a system is being used, but they do not establish that the work is faster, better or more useful to customers. Turning those counts into targets can reward unnecessary usage and make engineering systems more costly or complicated.
Amazon’s instructions also make the policy more specific than a general call to use AI responsibly. Managers are asked to judge outcomes, investigate why adoption differs across teams and use the tools on their own production work. That approach leaves room for a team to conclude that a tool is not helpful for a particular task. It also puts pressure on leaders to show that adoption delivers results, rather than simply asking employees to raise usage.
The comparison with Amazon’s AI bets should be treated as context, not proof that the memo was prompted by those results. Still, the reported status of Nova and the rise in the recorded value of Anthropic show why investment scale and activity metrics are not the same as demonstrated product impact. Amazon had not yet reported its third-quarter results at the time of the article, so the memo and available portfolio information do not provide a complete financial scorecard.
Products Worth Considering
Apple MacBook Pro 14.2-inch with M5 Max, 18-Core CPU, 20-Core GPU, 48GB Memory, 2TB SSD, Space Black
Trezor Safe 5 Crypto Hardware Wallet with Color Touchscreen, Black Graphite
Ledger Stax Crypto Wallet with Touchscreen
As an affiliate, we earn on qualifying purchases.
From Internal Rankings to Formal Guidance
The Oct. 27 memo follows earlier reports about internal AI-use tracking. In May, the Financial Times reported that some Amazon employees assigned unnecessary tasks to an internal agent tool called MeshClaw to increase token counts. The accounts were anonymous and are reported allegations, not independently established facts. The same reporting described an adoption target and internal token-use rankings.
Amazon did confirm one related development. A company spokesperson told Business Insider at the end of May that a leaderboard called KiroRank had been “deprecated,” describing it as an informal tracker made by a group of employees and saying 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 carries that principle into official engineering expectations.
The memo also avoids defining a fixed AI requirement in the job description. Its authors say any such definition could become outdated quickly. Instead, they anchor the role in outcomes that are expected to endure: customer, business and team impact. That framing reflects the pace of change in tools while leaving teams responsible for judging whether a particular tool is useful.
““Tokenmaxxing is at odds with frugality.””
— Amazon engineering memo
What the Guidance Does Not Establish
The memo does not set out a replacement numerical score for AI’s effect on engineering work, nor does the source material identify a single method Amazon will use to measure customer or business outcomes across teams. It says input metrics correlate with results today, but does not quantify that relationship or show how it varies by task.
The earlier claims about employees inflating token counts came from anonymous accounts reported by the Financial Times; Amazon’s confirmation that KiroRank was deprecated does not independently verify those accounts. The source material also does not establish why Amazon reduced the role of flagship Nova models or whether those changes amount to a broader retreat from its model work. The reported Anthropic valuation is based on a company filing and is not a realized return. Amazon’s third-quarter figures were still pending, so their financial implications were unknown.
Amazon’s Results and Team-Level Measures
Amazon was scheduled to report its third-quarter results on Oct. 29, the day after the memo’s publication. Those results may provide more information about the company’s financial position and investment accounting, but they will not by themselves establish whether AI tools improve engineering outcomes. The memo sets no announced date for a further policy update.
For now, the practical next step is for managers to examine how AI performs across actual work: where it helps, where it does not, and whether heavier use leads to better delivery. Amazon’s memo asks leaders to use the tools in production work themselves and share examples of both successful and unsuccessful use. Whether that approach changes internal evaluation practices, or how those outcomes will be reported, remains to be seen.
Key Questions
What did Amazon tell its engineers?
Amazon said engineers should not be judged by token counts, AI tool adoption, code changes or reviews as targets. The memo says evaluations should focus on customer, business and team impact.
Is Amazon telling employees to stop using AI?
No. The guidance says adoption measures correlate with important outputs, but warns against rewarding usage for its own sake. It asks teams to determine whether AI helps with their actual work.
What happened to Amazon’s KiroRank leaderboard?
Amazon told Business Insider that KiroRank had been deprecated and described it as an informal tracker created by employees, not an official tool intended to promote AI usage.
What is known about Amazon’s Nova and Anthropic bets?
Business Insider reported that most flagship Nova models had been moved to a “keep the lights on” status, and Amazon’s lifecycle page lists several Nova products as retired. The source material says Amazon’s June 30 filing valued its Anthropic holdings at about $190 billion; that is a reported paper valuation, not cash received.
When will Amazon’s next results provide more information?
Amazon’s third-quarter results were scheduled for Oct. 29, 2026. The source material did not include those results, so their effect on the company’s AI investments was not yet known.
Source: ThorstenMeyerAI.com
