Amazon tells tokenmaxxing devs to slow their AI roll

Amazon has instructed its engineers to prioritize delivering value, rather than focusing on the use of AI tokens, according to a memo sent to engineering teams on Tuesday. The memo, authored by Amazon VP and distinguished engineer Terk Terkowitz, highlights the evolving role of software development engineers (SDEs) in the age of AI. Terkowitz stated that an engineer's role is not defined by the use of AI tools but by the delivery of useful software and results for customers and the business.

Terkowitz emphasized that defining what constitutes good AI-powered engineering practices is challenging, as the inputs, tools, and techniques continue to change rapidly. However, the company has decided to focus on what remains constant during this evolution. The memo argues that metrics related to AI input, such as token usage, are flawed and disconnected from what truly matters.

Using these metrics as targets rather than measures could lead to increased costs and complexity, ultimately at odds with Amazon's mission to "invent and simplify."

The memo discourages a practice known as "tokenmaxxing," which involves using AI to meet arbitrary token usage targets without delivering meaningful results. Instead, Amazon encourages engineers to measure outcomes, not token consumption, when utilizing AI tools for software development. The company also emphasizes the importance of developers understanding when AI does not enhance their work, demonstrating the judgment required to make informed decisions about its application.

Amazon's directive aligns with a similar approach taken by Microsoft in August, which also prioritized outcomes for customers and the business over the use of AI token metrics. Microsoft's EVP Jay Parikh stressed the need to focus on maximizing outcomes that move the needle for customers and the business. As major tech companies reconsider the usefulness of AI token consumption as a measure of productivity, it may signal a shift in the industry's approach to integrating AI technologies, with a growing emphasis on leveraging cost-effective models and ensuring AI improves customer delivery and business outcomes.