Amazon Finds 'Tokenmaxxing' Lacks Productivity Gains
5 61Ancient Slashdot reader Gilmoure shares a report from Tom's Hardware, adding: "Huh. Looks like consuming one's own slop does not turn the wheels of industry." From the report: Amazon wants its engineers to quit chasing token counts, warning that heavy AI usage does not automatically mean better work. In updated engineering guidelines published on October 27, the company laid out how developers and managers should gauge AI's real impact -- and the takeaway is blunt: consuming more tokens is not the same as delivering more value. [...]
According to the memo, metrics such as AI adoption rates, token consumption, code changes, and code reviews become "deeply flawed" when they are disconnected from actual results. Turning them into targets, Amazon says, encourages wasteful behavior that clashes with its emphasis on "frugality and simplicity" and can raise both compute bills and software complexity.
The company's preferred yardstick is long-term results. "We believe that it is best to measure, and be measured by, long-term outcomes," Amazon wrote. That does not mean managers should ignore usage data altogether, as input metrics remain useful when examined "with care and judgment" alongside other indicators. Amazon suggests looking at both the teams using AI the least and those using it the most to determine whether adoption is translating into better delivery.
[...] The memo also pushes back on the idea that every task needs an AI tool. If current tooling does not benefit someone's work, Amazon says, that is "a finding, not a failing." Engineers who can explain where today's tools fall short are demonstrating judgment rather than reluctance to adopt AI.
5 comments
from the files of "No Shit Sherlock" (Score: 5, Insightful)
by haruchai ( 17472 ) on Wednesday October 28, 2026 @02:11PM (#66398206)
how did the idiots who made the policy ever get hired?
Re:from the files of "No Shit Sherlock" (Score: 5, Informative)
by Valkyre ( 101907 ) on Wednesday October 28, 2026 @02:17PM (#66398224)
Came to post exactly this. You gave people a metric and said that their job depended on meeting it. What did you expect?
Re: from the files of "No Shit Sherlock" (Score: 5, Interesting)
by ArmoredDragon ( 3450605 ) on Wednesday October 28, 2026 @02:59PM (#66398308)
Cheese...This whole time I literally thought token maxing meant getting the most bang for your tokens. No joke. I never bothered reading anything about it because nothing about my job ever said "use no more than x tokens" and nobody ever said I was using too many. The work objectives have always been the deliverable, not how we delivered it. The only restriction we have is the obvious: Sandbox it, and don't just give it everything.
Organizations suboptimize to meet new metrics (Score: 5, Informative)
by spazmonkey ( 920425 ) on Wednesday October 28, 2026 @02:17PM (#66398222)
Studies show it takes 30 days for employees suboptimize efforts to adjust output to get results that fit new performance metrics.
Its not just that THIS performance metric was pants-on-head stupid. Its the entire idea that a new performance metric will accomplish anything other than employees tailoring efforts to meet that new metric at the expense of job performance and meeting actual organizational goals.
Re:Organizations suboptimize to meet new metrics (Score: 5, Interesting)
by Martin Blank ( 154261 ) on Wednesday October 28, 2026 @03:11PM (#66398338)
My employer announced that AI costs will be billed back to departments. Token use (almost entirely among developers) was cut in half without affecting productivity, and perhaps even enhancing productivity because they had to think a bit about the best use of what they had instead of trying random things.