🔍 Read the full analysis: AI Outcomes Vs. Token Counts: The Perspective In Amazon’s Bets on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Amazon published engineering guidance on Oct. 27 saying software engineers should be judged by customer, business and team impact, not by token counts or AI-tool usage. The memo says those inputs can correlate with results but become flawed targets; the broader performance of Amazon’s Nova models and its Anthropic investment remains a separate, mixed picture.
Amazon published revised guidance for its software engineers on Oct. 27, saying performance should be judged by impact on customers, the business and teams—not by how many AI tokens an employee uses or how often they use AI tools. The memo follows reports of engineers giving an internal tool unnecessary tasks to raise their token counts and sets out how managers should assess AI adoption without turning activity into a score.
The guidance was written by the company’s senior engineers and published on Amazon’s news site. It was introduced by Terk Terkowitz, a vice president and distinguished engineer in Amazon’s stores business. The memo says that measures such as token use, AI-tool adoption, code changes and reviews can be useful indicators, but are “deeply flawed” when treated as targets rather than measurements. Used as goals, they can increase costs and system complexity.
The memo does not say that teams should stop using AI. It says input metrics currently “correlate well” with outcomes that matter, and that teams adopting AI most are delivering more while teams that do not adopt it are “falling behind.” It calls for managers to examine both low and high adoption: understand why some teams are not benefiting, and test whether the heaviest users are producing better results.
Amazon’s guidance says that when current AI tools do not help a person’s work, that should be recorded as a finding rather than treated as a performance failure. It also asks leaders to use the tools on their own production work and describe tasks where they helped and where they did not. The memo avoids defining AI as a fixed job requirement, saying, “Any definition we write now will likely look antiquated by this time next year.”
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
Amazon Shifts AI Performance Measures
The memo addresses a growing management problem: activity is easy to count, but value is harder to measure. A token tally can show that an employee used an AI system, but does not show whether the work improved a customer experience, reduced effort or helped the business. Treating usage as a target can reward unnecessary work and make teams appear productive without establishing a useful outcome.
For Amazon employees, the guidance signals that AI adoption alone is not the intended performance test. For other employers, it offers a distinction between using adoption statistics to spot patterns and using them to rank individuals or teams. Amazon’s own memo still regards usage metrics as potentially informative, but says managers need to test them against results and use judgment.
The company’s AI investments provide a wider, though incomplete, business backdrop. The source account describes Amazon’s Nova models as having been scaled back and its Anthropic stake as having risen substantially in reported value. Those developments do not establish whether Amazon’s overall AI strategy is succeeding: one concerns its own model products, the other a financial investment in an outside company, and neither substitutes for evidence about customer or operating outcomes.
Products Worth Considering
Dell UltraSharp U2725QE 27-Inch 4K UHD IPS Black Monitor
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
FLEXISPOT 32-Inch Electric Standing Desk Converter with Keyboard Tray
As an affiliate, we earn on qualifying purchases.
From KiroRank to Formal Guidance
The memo followed earlier reports about internal AI-use tracking. In May, the Financial Times reported that some Amazon employees gave an internal agent called MeshClaw unnecessary tasks to increase token counts. The accounts were anonymous and are reported allegations, not independently confirmed details in the supplied material. The reporting also described a weekly AI-use target and internal leaderboards.
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 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 October memo takes that message further by putting outcome-based judgment into official engineering role guidance.
The source account also cites a mixed record for Amazon’s AI bets. 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, the source says Amazon invested about $18 billion in Anthropic and reported holdings valued at about $190 billion on June 30. These are attributed figures and reports, not a full measure of the commercial performance of either effort.
“Tokenmaxxing is at odds with frugality.”
— Amazon engineering memo, published October 27
How Results Will Be Measured
The memo does not specify a single replacement metric or explain how managers will compare different kinds of work. It says outcomes can take time to measure and asks leaders to use judgment, but the supplied material does not describe how that judgment will be applied in reviews or whether teams will adopt common standards.
It is also unclear how much the reported tokenmaxxing affected performance evaluations, budgets or actual AI use. The accounts about employees assigning pointless tasks came from anonymous sources cited in earlier reporting; Amazon’s confirmation that KiroRank was deprecated does not independently verify every account about how the tracker was used.
Finally, the Nova and Anthropic figures do not provide a like-for-like measure of product success. The reported value of Amazon’s Anthropic holdings is a financial valuation, while Nova’s reported status concerns model products and lifecycle decisions. Amazon’s upcoming earnings may add financial detail, but the source material does not establish what the company will say about the returns from either AI effort.
Earnings and Workplace Practice
Amazon is scheduled to report third-quarter results on Oct. 29, 2026, one day after the memo’s publication. The results may provide additional company-level information, but no figures or AI-specific disclosures are confirmed in the supplied material. The source account says it will not forecast the results.
Within engineering, the next test is how managers apply the guidance: whether they use adoption data to identify teams that need support, check whether heavy AI use improves outcomes and allow employees to document cases where tools do not help. The memo establishes the stated principle—judge impact, not volume—but does not set a timetable for reviewing how the policy works.
Key Questions
What did Amazon announce?
Amazon published revised software-engineering guidance on Oct. 27, 2026. It says engineers should be assessed by impact on customers, the business and teams, rather than token counts or AI-tool usage alone.
Is Amazon telling employees to stop using AI?
No. The memo says teams that adopt AI are delivering more, while warning that usage statistics should not become performance targets. It asks managers to assess whether AI use actually improves results.
What does “tokenmaxxing” mean in this report?
It refers to increasing token use—such as by assigning an AI tool unnecessary tasks—to improve a usage tally or leaderboard position. Earlier reports described such behavior at Amazon; those accounts relied on anonymous sources.
What did Amazon say about the KiroRank leaderboard?
Amazon confirmed to Business Insider that KiroRank had been deprecated. The company described it as an informal employee-created tracker and said it was not intended to promote AI use for its own sake.
Does the memo settle whether Amazon’s AI investments are successful?
No. It sets guidance for evaluating engineering work. Reports about Nova’s status and the value of Amazon’s Anthropic holdings provide separate pieces of context, but do not establish the overall commercial or customer impact of Amazon’s AI strategy.
Source: ThorstenMeyerAI.com
