🔍 Read the full analysis: Amazon’s AI Investments Point To A Broader Measure Of Success 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 engineering guidance on October 27 saying software engineers should be judged by customer, business and team outcomes, not AI usage totals such as tokens or code volume. The memo says adoption data can still help teams understand performance, but Amazon’s Nova model changes and its rising Anthropic stake underscore that use and investment are not the same as durable results.
Amazon said on October 27 that its software engineers should be evaluated on the impact of their work, not how many AI tokens they use or how frequently they use AI tools. The guidance matters beyond employee reviews: it arrives as Amazon scales back some of its own Nova models and holds a substantially more valuable stake in Anthropic, illustrating why AI activity is not itself a measure of business success.
The memo, published on Amazon’s news site, introduces revised role guidance for all software engineers. It was written by the company’s most senior engineers and is associated with Terk Terkowitz, a vice president and distinguished engineer in Amazon’s stores business. It says the role should center on impact for customers, the business and the team, rather than prescribe a fixed set of AI tools or behaviors.
The memo calls token use, AI-tool adoption, code changes and code reviews “deeply flawed metrics” when used as targets. It warns that treating them as goals can raise costs and make systems more complex. At the same time, it says these measures currently “correlate well” with outcomes that matter. The guidance therefore does not tell teams to stop using AI or to discard usage data; it says the data should inform judgment rather than become a scorecard.
Amazon’s suggested approach is to examine both low and high adoption. Leaders should investigate why teams with low AI use are not adopting the tools, and check whether heavy users are delivering better results. The memo also asks leaders to use AI themselves on production work and describe tasks where it helped and where it did not. If a task does not benefit from current tools, the memo says, that is a finding to document, not 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
From AI Usage to Customer Impact
The guidance addresses a growing management problem: usage is easy to count, while productivity and customer benefit are harder to measure. A token total can show that a tool was used, but it cannot establish whether the work became faster, less costly or more useful. Turning such a count into a target can reward activity without showing that the activity helped.
For engineers, Amazon’s wording gives formal backing to the idea that not every task benefits from AI. For managers, it places responsibility on them to test whether adoption produces results instead of assuming that higher usage is inherently better. The company’s emphasis on frugality also points to a direct cost concern: unnecessary model use can consume resources while adding little value.
The distinction is relevant to other companies setting AI targets. Amazon’s memo offers no new productivity figures or evidence that a specific usage threshold improves outcomes. Instead, it sets out a measurement principle: treat adoption as a signal to investigate, and judge success by what changes for customers, the business and colleagues.
Products Worth Considering
Bose S1 Pro+ Portable PA Speaker
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade
As an affiliate, we earn on qualifying purchases.
Amazon’s Contrasting AI Bets
The memo follows earlier scrutiny of internal AI-use tracking. The source account cites a May Financial Times report that some Amazon employees assigned unnecessary tasks to an internal agent tool, MeshClaw, to raise token counts. Those accounts were anonymous and are reported claims, not independently established facts in the material provided. The same account says an Amazon spokesperson confirmed to Business Insider that the KiroRank leaderboard had been discontinued, describing it as an informal employee tracker that was not intended to promote AI use for its own sake.
Amazon’s AI portfolio provides a separate backdrop. Business Insider reported in July that most flagship Nova models had been moved to a “keep the lights on” status. Amazon’s lifecycle information, as cited in the source material, lists Nova Canvas, Nova Reel and the original Nova Sonic as retired. These developments do not establish that Amazon has abandoned its broader model program, but they show that product availability and continued investment can change.
Amazon’s investment in Anthropic offers a contrasting financial measure. The source material says Amazon invested about $18 billion and that a June 30 filing valued its holdings at about $190 billion. That is a reported paper valuation, not cash realized from a sale or proof of operating returns. Amazon is due to report third-quarter results on October 29; those results may provide further company-wide financial context, but the figures were not available when the memo was published.
““Tokenmaxxing is at odds with frugality.””
— Amazon engineering memo
What the Metrics Cannot Yet Show
The memo does not provide quantified evidence that AI adoption has increased productivity, improved customer outcomes or reduced costs across Amazon. Its statement that usage measures currently correlate with relevant outputs does not establish that greater use causes better results, and the company does not specify how managers should weigh the different measures in individual performance reviews.
The reported valuation of Amazon’s Anthropic holdings is also a paper value; the source material does not establish what Amazon could realize from a sale or how the investment has affected its operating results. The precise status and future plans for Amazon’s Nova portfolio are not fully described here. Amazon’s October 29 earnings release may add financial detail, but it will not by itself settle how much of any reported performance is attributable to AI.
Amazon’s Earnings and Engineering Measures
Amazon is scheduled to report its third-quarter results on October 29, one day after publishing the engineering guidance. Those results will provide an updated view of the company’s finances, including any reported investment effects, though the source material does not establish which AI-related measures Amazon will disclose.
Within engineering, the memo points managers toward reviewing adoption at both ends of the distribution and documenting where tools help or fall short. The next test is whether teams use that guidance to connect AI activity to measurable customer and business outcomes, rather than replacing one leaderboard with another. Amazon has not announced a specific new company-wide productivity measure in the material available.
Key Questions
What did Amazon announce?
Amazon published revised software-engineering guidance on October 27 saying engineers should be evaluated on customer, business and team impact, not raw AI usage totals.
Is Amazon telling employees to stop using AI?
No. The memo says AI adoption measures can correlate with relevant outcomes. It cautions against using those measures as targets or performance scores, rather than discouraging AI use.
What does “tokenmaxxing” mean here?
It refers to using AI tools in ways that increase token consumption or leaderboard rank, rather than because the work needs the tool. The memo says that behavior can conflict with frugality.
What is known about Amazon’s Nova models and Anthropic investment?
The source material cites reporting that most flagship Nova models were moved to maintenance and says several named Nova products are retired. It also reports an approximately $18 billion investment in Anthropic and a June 30 filing valuation of about $190 billion for Amazon’s holdings; that valuation is not a realized cash return.
When will Amazon provide its next financial update?
Amazon is scheduled to report its third-quarter results on October 29, 2026. The memo does not specify which AI-related financial or productivity measures the company will include.
Source: ThorstenMeyerAI.com
