🔍 Read the full analysis: Amazon’s Engineering Guidance Puts AI Results Ahead Of Activity 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 revised guidance for software engineers on October 27, saying outcomes—not token counts or AI-tool usage—should define good work. The memo still treats adoption metrics as useful signals, while warning that turning them into targets can encourage waste and complexity.
Amazon published revised guidance for its software engineers on October 27, telling them to judge AI-assisted work by its impact on customers, the business and teams, rather than by token counts or how often they use AI tools. The company’s engineering leaders say usage data can help identify patterns, but using it as a performance target risks raising costs and adding complexity without proving better results.
The memo was written by Amazon’s most senior engineers and published by the company. It introduces a revised role guideline for software engineers, without prescribing a fixed set of AI tools or usage requirements. Terk Terkowitz, a vice president and distinguished engineer in Amazon’s stores business, is identified as the memo’s author. The guidance says a detailed definition of AI use could quickly become outdated, so it anchors the role in enduring outcomes instead.
Amazon’s memo calls token usage, AI-tool adoption, code changes and reviews “deeply flawed metrics” when used as targets. They are inputs, not the customer or business results the company ultimately wants. The memo warns that pursuing those counts as goals can drive up costs and make systems more complex. It uses Amazon’s leadership principles to frame the issue: tokenmaxxing conflicts with frugality, while measuring lines of code conflicts with simplifying and inventing.
The guidance does not say AI adoption is pointless. It says these input measures currently “correlate well” with important outputs, and that teams adopting AI most heavily are delivering more, while teams that do not adopt it are falling behind. It recommends examining both lower and higher adoption levels: leaders should investigate why some teams use AI less and whether heavy users are producing better work. Engineers who find that AI does not help their task are asked to explain and share that finding, rather than treat it as a failure.
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 Amazon Rejects Usage Targets
The guidance draws a practical distinction for companies adopting generative AI: track usage to learn, but do not confuse usage with value. A token total or number of AI-assisted code changes can show that a tool is being used; on its own, it does not show that customers are better served, work is more reliable or a team has become more productive.
That distinction matters to workers whose performance may be evaluated during rapid AI rollouts. Amazon’s memo explicitly makes room for professional judgment, including a reasoned decision not to use a tool for a particular task. For managers, it directs attention toward the quality and effect of the work rather than activity that is easy to count. The memo’s recommendation that leaders use the tools on their own production work also asks them to examine where AI helps and where it does not.
The company’s caution is not a claim that its AI rollout has produced no gains. The memo says adoption measures currently correlate with outcomes, while warning that using them as targets can create perverse incentives. That balance is the core point: adoption may be a useful signal, but it is not a substitute for measuring results.
Products Worth Considering
Jackery HomePower 3000 Portable Power Station with Two 200W Solar Panels
DJI Osmo Pocket 3 Creator Combo Vlogging Camera with Mic 2, 1 CMOS, 4K/120fps, 3-Axis Stabilization, 128GB Card
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
As an affiliate, we earn on qualifying purchases.
From Internal Leaderboards to Guidance
The memo follows earlier reporting about internal AI-use tracking at Amazon. In May, the Financial Times reported that some employees assigned unnecessary jobs to an internal agent tool, MeshClaw, to increase token counts. That account relied on employee reports and should be treated as reported behavior, not as a company-confirmed description of every team. The same coverage described an internal target for weekly AI use by more than 80% of developers and token-use leaderboards.
Amazon did confirm that a leaderboard called KiroRank was “deprecated” at the end of May, according to Business Insider. The company said it had been an informal tracker created by a group of employees and 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 memo turns that concern into formal engineering guidance, while preserving a role for adoption data as an indicator to investigate.
The source article also compares the guidance with Amazon’s broader AI investments, including its Nova models and its investment in Anthropic. Those comparisons are separate from the memo and do not establish why Amazon adopted this policy. The clearest confirmed development here is the company’s stated approach to evaluating engineering work: assess impact, and use activity metrics cautiously.
““Tokenmaxxing is at odds with frugality.””
— Amazon engineering memo
How Teams Will Apply the Guidance
The memo sets out principles and suggested checks, but the source material does not provide a detailed company-wide evaluation system, targets or enforcement process. It is not clear how individual teams will translate customer and business impact into reviews, or whether Amazon will continue tracking token use and adoption in internal dashboards beyond using them as signals.
The memo’s statement that high adoption currently correlates with better outputs is Amazon’s characterization; the material provides no underlying data, measurement period or comparison method. The earlier claims about employees inflating token counts came from reporting based on employee accounts, while Amazon’s confirmation concerned KiroRank’s deprecation and stated purpose. These are distinct claims and should not be treated as a single company-wide finding.
What Engineers and Managers Do Next
The memo asks leaders to use AI tools on their own production work and identify at least one task where the tools helped and one where they did not, explaining the difference. It also calls for leaders to examine adoption at the lower and upper ends of the distribution, including whether heavy use is translating into better delivery. Those steps are intended to replace leaderboard-style incentives with investigation and judgment.
Amazon’s next test is how the guidance appears in day-to-day engineering management and performance reviews. The source material does not give a timeline for further policy changes or report how teams have responded so far. Amazon was scheduled to report its third-quarter results on October 29, one day after the memo’s publication; the source article says it would not speculate about those results. Company financial disclosures may provide broader business context, but they will not by themselves show how the new engineering guidance is being applied.
Key Questions
What did Amazon tell its software engineers?
Amazon’s October 27 guidance says engineers should be judged by customer, business and team impact, not simply by token counts or how often they use AI tools.
Is Amazon telling engineers to stop using AI?
No. The memo says adoption measures currently correlate with important outputs and warns that teams avoiding AI may fall behind. Its objection is to making usage itself a performance target.
What does “tokenmaxxing” mean in the memo?
It refers to increasing AI token use for its own sake, rather than because the work benefits. Amazon says that approach conflicts with its principle of frugality.
Does the guidance say engineers will be penalized for not using AI?
The memo says that if current AI tools do not help a task, that is a finding, not a failing. It asks engineers to document their reasoning; the source material does not describe a specific penalty or review process.
What remains unknown about the policy?
The memo does not specify how each team will measure outcomes, what data will be used in performance reviews, or how Amazon will implement the guidance across the company.
Source: ThorstenMeyerAI.com
