🔍 Read the full analysis: The Difference Between Using AI And Getting Results, Through Amazon’s Lens 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 impact, not token counts or AI-tool usage. The memo says usage measures can still help teams spot patterns, but turning them into targets can raise costs and complexity. Amazon’s Nova and Anthropic investments provide separate examples of how its AI bets have performed, though the full financial picture was not yet available.
Amazon published new guidance for its software engineers on October 27, saying performance should be judged by outcomes for customers, the business and teams—not by token counts or how often workers use AI tools. The memo responds to concerns about usage targets encouraging unnecessary activity, and comes as Amazon’s own AI models and its investment in Anthropic offer contrasting examples of the company’s broader AI strategy.
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 developed by senior engineers. It says long-term outcomes are the measures that matter, while acknowledging that they can be difficult and slow to assess. Some teams have instead tracked AI-tool adoption, token usage, code changes and reviews as proxies.
Amazon calls those measures “deeply flawed” when used as goals rather than indicators. The memo warns that targets can increase costs and make systems more complex. It also says the measures currently “correlate well” with important outputs, and that teams adopting AI more heavily are delivering more. The guidance does not call for abandoning AI use; it says managers should not treat usage itself as proof of performance.
The memo asks leaders to examine both lower and higher levels of adoption: why some teams use AI less, and whether the heaviest users are producing better results. It says work that does not benefit from current AI tools is “a finding, not a failing,” and asks engineers to record their reasons. Leaders are also told to use the tools on their own production work and describe tasks where they helped and where they did not.
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 Activity to Measurable Impact
The guidance matters beyond Amazon because companies are trying to assess whether AI changes the work employees produce, rather than simply whether they have access to or use new tools. Usage counts are easy to collect, but the memo argues they can reward activity that does not improve a product or service. If those counts become targets, teams may spend more on model usage or add complexity without a corresponding customer benefit.
Amazon’s approach preserves usage data as a diagnostic signal, not a scorecard. That distinction could shape how managers set expectations: adoption patterns may identify teams that need better tools or training, while output and customer impact remain the basis for judging results. The memo’s position also recognizes that a worker can make a sound decision not to use AI for a task when the tools do not help.
The company’s wider AI record adds a business dimension, but the cases are not directly comparable. Amazon’s Nova models have faced reported cutbacks and retirements, while its Anthropic investment has risen in paper value. Those developments underline that AI activity and AI value are different measures; they do not, by themselves, establish the financial return from Amazon’s overall AI strategy.
Products Worth Considering
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
DJI Osmo Pocket 4 Gimbal Camera Pro Kit, 18-Piece Bundle
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
As an affiliate, we earn on qualifying purchases.
The Metrics Behind Amazon’s Memo
The memo follows reports of internal usage tracking. In May, the Financial Times reported that some Amazon employees gave an internal agent tool, MeshClaw, unnecessary jobs to increase token counts. The report also described a target for more than 80 percent of developers to use AI weekly. These details were based on employee accounts and should be treated as reported claims, not as facts independently confirmed in the memo.
At the end of May, Amazon confirmed to Business Insider that an employee-built leaderboard called KiroRank had been deprecated. The company said it was an informal tracker and “was never intended to promote the use of AI for usage’s sake.” Senior vice president Dave Treadwell also told staff, “Please don’t use AI just for the sake of using AI.” The October guidance puts that principle into the company’s revised engineering role guideline.
Amazon’s AI investments provide further context. Business Insider reported in July that most flagship Nova models had been moved to maintenance status. Amazon’s lifecycle page lists Nova Canvas, Nova Reel and the first Nova Sonic as retired. Separately, the source reports that Amazon invested about $18 billion in Anthropic and valued its holdings at about $190 billion in a June 30 filing. These figures describe an investment’s reported paper value, not cash realized from a sale.
“Tokenmaxxing is at odds with frugality.”
— Amazon engineering memo
What the Guidance Leaves Open
The memo does not set out a single replacement metric or define exactly how every team should measure customer, business and team impact. It acknowledges that meaningful outcomes can take time to appear, leaving managers to apply judgment while deciding which indicators are useful for a particular kind of work.
It is also unclear how Amazon will assess whether teams with high AI adoption are actually producing better results over time. The memo says adoption measures currently “correlate well” with outputs that matter, but the source material does not provide the underlying data, define those outputs or quantify the relationship. Earlier reporting about token targets and employees gaming usage measures remains attributed reporting, rather than a confirmed account of company-wide practice.
Amazon’s financial results were due October 29, one day after the memo’s publication. Until those results are reported, the figures cited for Nova and Anthropic do not establish the overall return on Amazon’s AI spending or explain how the company measures its internal AI tools’ impact.
Amazon’s Results and Measurement
The next scheduled milestone was Amazon’s third-quarter earnings report on October 29. That report could add information about the company’s finances, but the source material does not establish that Amazon planned to break out returns from its AI investments or internal engineering tools.
For engineers and managers, the immediate test will be how the new role guideline is applied across teams. The memo calls for leaders to examine both high and low adoption and document where AI helps or does not help. Whether Amazon later publishes clearer outcome measures, or details how it evaluates those findings, remains unknown.
Key Questions
What did Amazon’s new engineering guidance change?
It says engineers should be assessed on impact on customers, the business and their teams, rather than on token usage or AI adoption alone. The guidance does not ban tracking those measures as indicators.
Is Amazon telling employees to stop using AI?
No. The memo says teams adopting AI more heavily are delivering more, while warning against treating usage as a target. It asks managers to examine where tools help and where they do not.
What does “tokenmaxxing” mean here?
It refers to increasing AI token usage for its own sake, including by giving a tool unnecessary tasks. The memo says that approach conflicts with frugality and does not establish that more usage produces better work.
What remains unclear about Amazon’s AI results?
The memo does not publish the data behind its reported correlation between AI adoption and outcomes or provide a universal replacement metric. The cited investment figures also do not establish the overall return on Amazon’s AI spending.
Source: ThorstenMeyerAI.com
