AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: What Amazon’s AI Bets Reveal About The Limits Of Token Counting on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Amazon published engineering guidance on October 27 rejecting token counts and other AI-use measures as employee targets, while saying those indicators can still help teams assess adoption. The memo’s message lands alongside mixed signals from Amazon’s AI investments: reporting says most flagship Nova models were scaled back, while Amazon’s Anthropic stake was valued far above its investment in a June 30 filing.

Amazon published new guidance for its software engineers on October 27, warning that token counts, AI-tool adoption and code volume are poor targets for judging performance because they can reward activity rather than customer and business outcomes. The memo arrives as Amazon’s own Nova models and its investment in Anthropic offer contrasting tests of how the company’s AI bets are performing, though the available evidence does not establish a direct link between the guidance and those investment decisions.

The memo, published on Amazon’s own news site, was written by Terk Terkowitz, a vice president and distinguished engineer in the company’s stores business. It introduces revised role guidance for software engineers, drafted by the company’s senior engineers. The document says long-term outcomes are important but difficult to measure, which has led some teams to use proxies such as token usage, AI-tool adoption, code changes and reviews.

Those measures can provide information, the memo says, and currently “correlate well” with relevant outputs. But used as goals rather than indicators, they can increase costs and complexity. The memo says “tokenmaxxing is at odds with frugality” and that measuring lines of code conflicts with Amazon’s principle of inventing and simplifying. Its central distinction is between tracking whether tools are being used and rewarding people simply for using them.

The guidance asks leaders to examine both low and high AI-adoption levels: why some teams use the tools less, and whether heavy users are actually producing better results. It also says employees should document cases where current AI tools do not help. The memo’s wording is explicit: “that is a finding, not a failing.” Leaders are asked to use the tools in their own production work and explain where they helped and where they did not.

At a glance
analysisWhen: Published October 27, 2026; Amazon’s th…
The developmentAmazon published revised software-engineering guidance that warns against using token counts and other AI activity metrics as performance targets.
Thorsten Meyer AI
Timeline

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.

  1. May 2026Token leaderboards backfireFinancial Times (via The Decoder): staff give an internal agent tool pointless tasks to inflate token counts. Reported, anonymous sources.
  2. Late May 2026KiroRank “deprecated”Amazon spokesperson to Business Insider. SVP Dave Treadwell: “Please don’t use AI just for the sake of using AI.”
  3. 24 Jul 2026San Francisco AGI Lab site closesConfirmed by Amazon to GeekWire. Frontier model research continues under Pieter Abbeel.
  4. 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”.
  5. 14 / 30 Sep 2026Nova Sonic v1, Nova Reel, Nova Canvas reach end of lifeAWS Bedrock model lifecycle page. Requests to them now fail.
  6. 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.
  7. 27 Oct 2026The memo: outcomes, not tokenmaxxingUsage and throughput metrics are “deeply flawed” as targets. “Tokenmaxxing is at odds with frugality.”
  8. 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 Changing Its AI Scorecard

The guidance addresses a measurement problem facing companies adopting generative AI: usage is easy to count, while the quality, speed and customer value of resulting work can be harder to isolate. When a proxy becomes a performance target, employees may have reason to optimize the count instead of the work. Amazon’s memo acknowledges the risk without saying that adoption data should be discarded.

That distinction matters to engineers and managers alike. A low usage rate may point to a tool that is not useful for a particular task; a high rate does not, by itself, show that a team is delivering more value. The company’s proposed approach is to use metrics to find questions for review, then apply professional judgment and outcome evidence rather than treating a leaderboard as a verdict.

The company’s model and investment record provides a broader, but incomplete, lens on the same issue. Business Insider reported in July that most flagship Nova models had been moved to a maintenance posture, while Amazon’s June 30 filing valued its Anthropic holdings at about $190 billion. These figures suggest different outcomes across Amazon’s AI efforts, but they do not show that token counts caused those outcomes or reveal the full commercial return on either bet.

Products Worth Considering

Kindle Paperwhite Signature Edition 32GB (2024 Model) - Metallic Black

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

As an affiliate, we earn on qualifying purchases.

From Internal Rankings to Formal Guidance

The memo follows earlier reporting about internal AI-use tracking. In May, the Financial Times reported that Amazon employees assigned unnecessary tasks to an internal agent tool called MeshClaw to raise their token counts. The report also described a target for more than 80 percent of developers to use AI weekly and internal token leaderboards. Those accounts were based on anonymous sources and should be treated as reported allegations, not independently established company policy.

Amazon did confirm to Business Insider at the end of May that a leaderboard called KiroRank had been deprecated. The company described it as an informal tracker created by a group of employees and said it 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 guidance goes further by putting the outcome-focused principle into the company’s engineering role guidelines.

Amazon’s AI portfolio offers relevant investment context. Business Insider reported in July that most flagship Nova models had been moved to “keep the lights on”; Amazon’s lifecycle page lists Nova Canvas, Nova Reel and the first Nova Sonic as retired. Separately, a June 30 filing valued Amazon’s Anthropic holdings at about $190 billion, compared with roughly $18 billion invested, according to the source material. That filing valuation is not the same as cash proceeds or a realized return.

““Tokenmaxxing is at odds with frugality.””

— Amazon engineering memo

What the Metrics Cannot Yet Show

The memo does not provide a detailed measurement system for comparing AI-assisted work with work done without AI. It says adoption metrics correlate with outputs at present, but does not specify the data, time period or methods behind that assessment. It also does not set out how managers should weigh speed against quality, cost, reliability or customer impact across different types of engineering work.

The Nova and Anthropic figures also need careful interpretation. The reported Nova changes do not establish that every Nova product or Amazon model effort has been abandoned. The reported Anthropic valuation is a filing figure, not proof that Amazon has sold its stake or received a comparable amount in cash. The source material does not establish whether the valuation and model decisions are connected to the new engineering guidance.

Amazon’s third-quarter results had not yet been reported when the source article was written. The company was scheduled to publish them on October 29; until then, the available information does not provide that quarter’s financial results or a full account of the commercial returns from its AI work.

Amazon’s Results and Adoption Review

The next scheduled company milestone is Amazon’s third-quarter earnings report on October 29, 2026. Those results may provide additional information about the company’s finances, but the source material does not indicate that Amazon will report a standalone return on its AI engineering tools or its Nova models.

Inside engineering teams, the memo’s approach points toward continued review of adoption at both ends of the range, alongside explanations of where tools help or fail. Whether Amazon publishes further guidance or evidence about how it measures outcomes remains unclear. For now, the stated policy is not to maximize token use, but to judge whether the work produces useful results.

Key Questions

What did Amazon announce on October 27?

Amazon published revised software-engineering role guidance saying token use, AI adoption and code volume should not be treated as performance targets. It says such measures can still serve as indicators for investigation.

Is Amazon telling engineers to stop using AI?

No. The memo distinguishes between encouraging effective adoption and rewarding use for its own sake. It says teams that do not use AI may be falling behind, while also recognizing that a tool may not benefit particular work.

What does “tokenmaxxing” mean here?

It refers to maximizing AI token usage to improve a usage score or leaderboard position, rather than because the extra activity helps complete useful work. The source material says earlier reports described employees giving an internal tool unnecessary tasks, but those accounts were based on anonymous sources.

What is known about Amazon’s Nova and Anthropic bets?

Business Insider reported that most flagship Nova models were moved to a maintenance posture, and Amazon’s lifecycle page lists several Nova products as retired. A June 30 filing reportedly valued Amazon’s Anthropic holdings at about $190 billion against roughly $18 billion invested; that valuation does not mean Amazon realized that amount in cash.

When will Amazon report its next results?

Amazon was scheduled to report third-quarter results on October 29, 2026, the day after the memo’s October 27 publication. The figures were not available in the source material.

Source: ThorstenMeyerAI.com

You May Also Like

Today’s AI Workplace Picture: November 14, 2026

A November 14 news roundup connects AI-related financial risks, robot models, a vendor-linked Uber breach and an FAA maintenance disruption.

Smart Home Devices Should Work Without The Internet

A home automation builder says core devices should keep working offline, after cloud outages and vendor API changes disrupted connected products.

Jensen Huang’s AI Legacy And The Question Of His Exit

Nvidia says its board accepted Jensen Huang’s resignation Nov. 11. The stated record and separate governance and intelligence allegations remain distinct.

Claude Haiku 5.5 On AWS: Getting Acquainted

An AWS page is titled “Introducing Claude Haiku 5.5 on AWS,” but the available material does not confirm release timing, access, pricing or capabilities.