Amazon’s Anti-Tokenmaxxing Memo: The Renaissance Developer vs. the AI Slop Machine

(SeaPRwire) –

Amazon’s new memo is a rare moment of sanity in a tech sector drunk on AI metrics. The e-commerce giant just told its software engineers that using AI for every single task is not a virtue. Terk Terkowitz, vice president and distinguished engineer at Worldwide Amazon Stores, frames the shift with the term “Renaissance Developer”—someone who blends AI with their own expertise and judgment. This is not a tech utopia. This is a corporate audit of a deeply flawed assumption: that more AI tokens equals better work.

The memo’s core directive is brutally simple. Engineers must tell management when AI is not useful to their work. “If your work does not benefit from current AI tooling, that is a finding, not a failing,” Terkowitz wrote. “Document why, share it with your community of peers, bring it to your manager, and work to find solutions.” This is a direct refutation of the hype cycle. It treats an engineer’s judgment as more valuable than a raw token count. The memo goes further, calling token counting and lines of code “deeply flawed” measures of productivity. They do not reflect quality. They inflate cost. Terkowitz’s phrase “tokenmaxxing is at odds with frugality” is a shot across the bow at the entire AI adoption playbook.

Amazon’s internal data tells a cautionary tale. Employees were running AI agents on meaningless tasks just to climb an unofficial leaderboard called Kirorank. That practice got nixed in May after it drove up computing costs. The memo also references a company that accidentally burned through $500 million on Claude in one month because there were no usage limits. These are not isolated incidents. They are the inevitable outcome of treating AI usage as a badge of honor. Amazon has not abandoned its AI adoption metrics entirely. Terkowitz writes that they still “correlate well with the outputs that matter.” But he now urges managers to compare the 10th and 90th percentile of AI users on their teams. That is a data-driven way to see who is actually gaining productivity and who is just racking up tokens.

Amazon is not alone in this reckoning. Shopify’s CEO Toby Lütke recently warned about “slop grenades”—AI-generated emails and code that create more work for colleagues to review. This is the industry’s dirty secret. AI can generate volume, but it cannot generate judgment. The tech sector’s obsession with token counts and leaderboards is a symptom of a deeper problem: a lazy belief that productivity is a number you can measure in a dashboard. Amazon is now saying that belief is costly and misguided. The company is not discouraging AI use. It is demanding that engineers think about what AI is actually for. That is a far more mature position than the “AI everywhere” mantra that has consumed corporate America. The real supply chain here is not silicon. It is the judgment of the people using the tools.