Amazon Revises AI Guidelines to Foster Critical Thinking Among Engineers

Amazon, the world's largest cloud and e-commerce company, has completely revised its guidelines for engineers on the use of artificial intelligence (AI). The company aims to curb the practice of 'token maxing,' which focuses solely on increasing AI usage or code output, and instead promote the development of 'Renaissance Developers' who combine AI technology with human judgment.

On October 28, Terk Terkowitz, Amazon's Vice President, sent a memo outlining the new role guidelines to the entire Software Development Engineer (SDE) and Software Development Manager (SDM) organizations.

The guidelines, established by Amazon's Distinguished and Senior Principal Engineers, criticize the practice of evaluating developers based on short-term input metrics such as AI tool adoption rates, token consumption, and code modification counts.

Terkowitz stated, "Metrics like token usage or code output are 'very flawed evaluation methods' that are disconnected from actual customer outcomes. Setting these metrics as goals encourages token abuse, which contradicts our principle of frugality and only increases system complexity."

However, he recommended using these metrics in a limited way to identify the top 10% of teams that effectively leverage AI and the bottom 10% that struggle with AI adoption, to share best practices and address obstacles.

Amazon has defined the ability to honestly identify situations where AI implementation is ineffective as a core competency for engineers.

The memo emphasized, "If your work does not currently benefit from AI tools, this is not a 'failure' but an important 'finding.' An engineer who can accurately describe areas where AI cannot assist and discuss them with peers and managers embodies the critical judgment we seek."

It further urged engineers to evolve into 'Renaissance Developers' who create new value by integrating AI tools with human judgment and business expertise, moving beyond simple coding tasks.

Meanwhile, Amazon clarified that the revision of these guidelines is not intended to impose artificial limits on employees' use of AI. Mark Bruecker, Vice President of Amazon Web Services (AWS), stated, "If there is an important issue, we will invest the necessary tools, time, tokens, and human resources. What we do not want is to waste time and tokens solving unimportant problems."

While the global big tech industry had embraced unlimited AI token usage until early this year, the rising costs of AI infrastructure have led to a shift towards more efficient practices, such as 'model maxing' and performance-based AI operation strategies.