Amazon warns engineers against excessive AI token use in new memo
Amazon issued a memo to software engineers in September, later published on Tuesday, advising against ‘tokenmaxxing’—the practice of overusing AI tools without regard for cost or efficiency. The memo, part of broader AI integration under CEO Andy Jassy, acknowledges AI’s growing role in software development but warns that metrics like token usage or adoption rates are unreliable for measuring…
Key points
- Amazon’s memo discourages ‘tokenmaxxing’—excessive AI usage without cost or efficiency checks
- Company calls token usage and AI adoption metrics ‘deeply flawed’ for measuring productivity
- Amazon promotes ‘feature velocity’ and long-term outcomes over short-term AI tool adoption
The memo follows Amazon’s shutdown of an internal leaderboard in May after employees manipulated token usage to inflate rankings. While the company has not imposed token caps, it urges engineers to use AI judiciously, even when it doesn’t improve workflows. Marc Brooker, a vice president at Amazon Web Services, told CNBC that the company prioritizes ‘return on investment’—focusing AI efforts on high-value problems rather than speculative tasks. The memo also frames skepticism about AI’s utility as a sign of good judgment.
Amazon's anti-tokenmaxxing memo says engineers should be pickier about using AI
CNBC Technology · 27 October 2026
Loading the full article…
This text was published by CNBC Technology and written by Annie Palmer. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Enterprise & Industry
All →- Avante raises $30M Series A to expand AI benefits platform · 2 src
- Infisical raises $40M Series B at $600M valuation to secure AI agent access · 1 src
- Check Point uses OpenAI Daybreak to cut vulnerability research time · 1 src
- Microsoft Teams adds AI phishing detection for November 2026 rollout · 3 src
- Oracle launches Media Fabric AI platform for media production and streaming · 1 src
Comments
via GitHub Discussions