🔍 Read the full analysis: Amazon’s Engineers Get A Reminder: AI Metrics Need To Track Results on ThorstenMeyerAI.com
Get everyday essentials 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 revised software-engineering guidance on October 27, warning that token counts and AI adoption are flawed performance targets when disconnected from results. The memo says usage measures can still help teams spot patterns, but employees should be judged by impact on customers, the business and their teams.
Amazon’s engineering leadership told software teams on October 27 to stop treating token use and other AI activity measures as performance targets, saying engineers should be judged by their impact on customers, the business and their teams. The company’s revised role guidance follows reports that some employees used an internal AI tool for unnecessary tasks to raise their token counts.
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, written by the company’s most senior engineers. Rather than define the job around specific AI tools, the document says the technology is changing too quickly: “Any definition we write now will likely look antiquated by this time next year.”
The guidance calls measures such as token usage, AI-tool adoption, code changes and code reviews “deeply flawed” when they are used as targets instead of measures. The memo says those inputs can raise costs and complexity without showing whether work has improved. It links token targets to Amazon’s frugality principle and lines-of-code targets to its principle of inventing and simplifying.
But the memo does not say usage data is worthless or that engineers should avoid AI. It says those measures currently “correlate well” with important outputs, and that teams adopting AI most are delivering more while teams not adopting it are “falling behind.” Its distinction is between using metrics to investigate patterns and using them to score employees. The guidance asks leaders to examine both low and high adoption, and to understand whether heavier use is producing better work.
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
Why AI Activity Is Not Performance
The guidance addresses a measurement problem facing companies adopting generative AI: activity is easier to count than value. Token totals and tool-use rates can show that employees are using a system, but they do not establish that it saves time, improves software or benefits customers. If employees are rewarded for the count itself, they may have reason to generate more activity without improving the work.
That distinction matters to engineers whose employers are introducing AI tools and tracking adoption. Amazon’s memo says a finding that current AI tools do not help with a particular task is not a performance failure. It calls that information useful if the employee explains the reason and shares it with peers and a manager. The guidance also puts responsibility on leaders to use the tools on their own production work and describe where they help and where they do not.
For companies, the practical message is to treat usage figures as diagnostic signals, not success measures. A business still needs to decide what outcomes count—such as customer impact, quality or delivery—and how to assess them over time. The memo offers a change in management guidance, not evidence that Amazon has adopted a single replacement metric across its engineering teams.
Products Worth Considering
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
As an affiliate, we earn on qualifying purchases.
From Token Leaderboards to Guidance
The memo follows earlier reports about internal AI-use tracking at Amazon. In May, the Financial Times reported that some employees gave an internal agent tool called MeshClaw unnecessary tasks to increase token counts. That reporting also described a target for more than 80% of developers to use AI weekly and internal tracking of token use. Those details were reported accounts, not facts established by Amazon’s October memo.
At the end of May, Amazon confirmed to Business Insider 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 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 document is a further step: it sets out the company’s engineering role expectations, while avoiding a fixed definition tied to today’s tools.
The source article also compares the memo’s focus on outcomes with Amazon’s broader AI investments. It reports that Business Insider said most flagship Nova models had been moved to maintenance in July, while Amazon’s filing valued its Anthropic holdings at about $190 billion on June 30 after an investment of about $18 billion. Those reported figures and product developments provide a wider frame, but they do not establish that the memo was prompted by either investment or the performance of either company’s models.
“Tokenmaxxing is at odds with frugality.”
— Amazon engineering memo
How Teams Will Be Evaluated
The memo does not specify a replacement scorecard, a common set of outcome measures or how managers should weigh customer, business and team impact in individual reviews. It says usage measures correlate with outputs at present, but does not provide figures or explain how that relationship was assessed.
It is also unclear how widely the reported token-maximising behavior occurred, how many employees or teams were involved, or whether the new guidance changes existing evaluation practices immediately. The May accounts about unnecessary tasks and weekly adoption targets were reported by the Financial Times; Amazon’s later confirmation concerned the retirement of KiroRank, not every detail of those reports.
Amazon’s Next Performance Signals
Amazon’s memo directs leaders to examine teams at both ends of AI adoption: those using the tools least and those using them most. Managers are asked to investigate why adoption differs and whether heavy use is producing better work, while leaders are expected to test AI tools on their own production tasks and discuss the results with their teams.
Amazon is scheduled to report third-quarter results on October 29, one day after the memo’s publication. The source article says it will not predict those results, and the memo does not link its engineering guidance to the earnings report. Further details about how teams will apply the guidance, and whether Amazon will publish outcome measures or updated performance practices, remain unannounced.
Key Questions
What did Amazon tell its software engineers?
Amazon’s October 27 guidance says engineers should be judged by their impact on customers, the business and their teams, not by token counts or other AI activity measures used as targets.
Is Amazon telling employees to stop using AI?
No. The memo says AI adoption measures currently correlate with important outputs and warns that teams not adopting AI are falling behind. Its objection is to scoring people on usage rather than results.
What does “tokenmaxxing” mean here?
It refers to using an AI tool for extra or unnecessary tasks to increase the number of tokens processed and, in the reported cases, improve a place on a usage leaderboard. The May reports described such behavior at Amazon; the October memo says token use should not be a target.
What should Amazon managers measure instead?
The memo points to long-term outcomes for customers, the business and teams, but does not prescribe one replacement metric. It asks leaders to examine adoption patterns and determine whether AI use is producing better work.
When will Amazon report its next quarterly results?
Amazon is scheduled to report its third-quarter results on October 29, 2026, the day after the memo was published. The memo does not say that the guidance is connected to the earnings announcement.
Source: ThorstenMeyerAI.com
