# Amazon Engineers Stop Chasing AI Token Counts

> Amazon tells engineers to stop measuring productivity in AI tokens after internal token leaderboard was abused.
- Title: Amazon Engineers Stop Chasing AI Token Counts · BigGo Finance
- Summary: Amazon tells engineers to stop measuring productivity in AI tokens after internal token leaderboard was abused. Amazon updated engineering guidelines to stop…
- Keywords: amazon, token counts, engineering, software development, token leaderboard, ai tools, ai adoption, finance, markets, Engineers, Stop, Chasing
- Source: BigGo Finance — https://finance.biggo.com/news/b8c13de1-4e8b-47ce-bc5b-426589cdf818
- Author: BigGo Finance
- Published: 2026-10-28T16:55:14+00:00
- Read time: 1 min
- Topics: amazon, ai, token counts, engineering, software development, token leaderboard, ai tools, ai adoption, finance, markets
## Amazon updates engineering guidelines
Amazon published updated engineering guidelines declaring tokenmaxxing a deeply flawed practice.
## Tokenmaxxing defined
Tokenmaxxing is maximizing AI usage to appear productive but invites unnecessary complexity and higher compute bills.
## Internal token leaderboard shut
Amazon shut an internal token leaderboard in May after employees inflated token counts with an ai agent.
## Engineers judged on outcomes
Engineers should be judged on long-term business outcomes rather than token consumption or tool adoption rates.
## Terk Terkowitz's view
Terk Terkowitz says ai is changing software development faster than cloud computing and engineers should focus on customer impact.
## AI tool usage guidance
If work does not benefit from ai tools, that is a finding not a failing showing good judgment.
## Manager responsibilities
Managers must examine both top and bottom 10% of ai users to understand adoption effectiveness.
## Industry context
Uber burned through 2026 budget for claude code by april and ai token costs are rising across companies.
## Key takeaway

Amazon updated engineering guidelines to stop engineers from chasing AI token counts as token usage metrics can backfire.