🔍 Read the full analysis: From Tokens To Outcomes: The Measure Behind Amazon’s AI Bets on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Amazon published an engineering guideline on October 27 saying AI token use and tool adoption should not be treated as performance goals. The memo says those inputs can help identify patterns, but engineers should be judged by customer, business and team outcomes. Amazon’s Nova models and investment in Anthropic offer a wider test of whether its AI spending is producing value.
Amazon published new guidance for its software engineers on October 27, rejecting AI token counts and tool adoption as performance targets and directing teams to focus on outcomes for customers, the business and colleagues. The memo comes as Amazon’s in-house Nova models and its investment in Anthropic offer contrasting examples of the company’s AI strategy, though the financial and operational results are not yet fully clear.
The memo, posted on Amazon’s news site, was written by Terk Terkowitz, a vice president and distinguished engineer in Amazon’s stores business. It introduces a revised role guideline for software engineers, developed by the company’s most senior engineers. The guidance does not define the job around particular AI tools, saying those definitions could quickly become outdated. Instead, it anchors engineering work in its impact on customers, the business and the team.
Amazon’s memo calls measures such as token use, AI adoption, code changes and reviews “deeply flawed metrics” when used as targets. The company says these inputs can raise costs and complexity if workers are rewarded for activity rather than results. It also says the measures currently “correlate well” with outputs that matter, so they can still be useful as indicators. The distinction is whether they are used to spot patterns or to score employees.
The guidance asks leaders to examine both low and high AI adoption, including why some teams make little use of the tools and whether heavy users are delivering better work. It also says engineers whose tasks do not benefit from current AI tools should document that finding rather than treat it as a failure. Leaders are asked to use the tools on their own production work and discuss 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
A Shift From AI Activity to Results
The guidance matters because internal usage targets can shape how employees work, even when those targets are only intended as rough indicators. If engineers gain recognition for more tokens, code changes or AI sessions, they may have an incentive to increase activity without improving a product or service. Amazon’s stated standard instead puts the focus on whether AI makes work better for customers or the business.
That distinction also affects how companies evaluate AI investment. Usage data can show whether staff have access to tools and are trying them, but it does not, by itself, establish productivity gains, lower costs or improved customer outcomes. The memo offers an internal policy signal, not a published calculation of AI’s financial return at Amazon.
For employees, the wording makes room for professional judgment: a task that does not benefit from current AI is a result worth reporting, not evidence that the worker failed to adopt the technology. For other employers, Amazon’s position highlights the risk of turning easily counted activity into a substitute for harder-to-measure outcomes.
Popular right now
Ninja SP151 Flip Toaster Oven and Air Fryer, 8-in-1, Stainless Steel
Dyson V8 Cyclone Cordless Vacuum
Dreame L40 Ultra Gen 2 Robot Vacuum and Mop
As an affiliate, we earn on qualifying purchases.
Amazon’s Nova and Anthropic Bets
The memo follows earlier reports of employee usage tracking. The Financial Times reported in May that some Amazon staff gave the internal MeshClaw agent unnecessary tasks to increase token counts. The accounts were anonymous and remain reported allegations, not independently confirmed details. The FT also reported a weekly AI-use target and token leaderboards.
Amazon did confirm to Business Insider in late May that its KiroRank leaderboard 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 not to use AI simply to use it. The October memo formalizes a broader principle in the engineering role guidance.
The company’s AI portfolio provides another part of the picture. Business Insider reported in July that most flagship Nova models had been moved to a “keep the lights on” status. AWS lifecycle information lists Nova Canvas, Nova Reel and the first Nova Sonic as retired. By contrast, Amazon’s June 30 filing valued its Anthropic holdings at about $190 billion, against roughly $18 billion invested, according to the source material. That reported paper value is not the same as cash returned or proof of operating gains. Amazon’s third-quarter results are due October 29.
““Tokenmaxxing is at odds with frugality.””
— Amazon engineering memo
AI Returns Still Need Evidence
The memo does not provide a company-wide estimate of how much AI has improved productivity, quality or customer outcomes. It says adoption measures correlate with relevant outputs today, but does not publish the data, define those outputs in detail or quantify the relationship. It is also unclear how individual teams will apply the guidance in performance reviews.
The source material’s account of Amazon’s Anthropic stake gives a paper valuation, not a realized return. The reported status of Nova models also does not, by itself, establish the full commercial performance of Amazon’s broader AI offerings. The May reports about employees inflating usage relied in part on anonymous accounts; Amazon’s confirmation that KiroRank was deprecated does not confirm every detail in those reports.
Third-Quarter Results Due October 29
Amazon is scheduled to report third-quarter results on October 29, one day after the memo’s publication. Those results may provide information about the company’s business and investments, but the source material does not establish that Amazon will report a direct measure of returns from AI tools or explain how the new engineering guidance will be assessed.
The next practical test will be whether teams use the memo’s approach to connect AI activity with specific work outcomes. For now, Amazon has stated a change in what engineers should treat as success; evidence of the policy’s effects remains to come.
Key Questions
What did Amazon tell engineers about AI tokens?
Amazon said token counts should not be performance targets. They may help teams understand usage, but engineers should be evaluated on outcomes for customers, the business and their teams.
Is Amazon telling employees to stop using AI?
No. The memo does not reject AI tools. It says adoption measures currently correlate with important outputs, while warning against using activity as a substitute for results.
What is tokenmaxxing?
In the memo’s usage, tokenmaxxing means increasing AI token consumption as an end in itself, rather than using AI to improve work. Amazon says that approach conflicts with its principle of frugality.
What remains unknown about Amazon’s AI results?
The memo does not quantify AI’s effect on productivity, costs or customer outcomes, and it does not explain how teams will incorporate the guidance into reviews. The reported Anthropic valuation is a paper value, not a realized return.
Source: ThorstenMeyerAI.com
