tech
April 17, 2026
"Tokenmaxxing" is making developers less productive than they think
There's a lot more code—but it's a lot more expensive and requires a lot more rewriting.
TL;DR
- The practice of 'tokenmaxxing,' where developers aim to consume large amounts of AI processing power, is not a reliable measure of productivity.
- AI coding tools increase the amount of accepted code, but engineers often need to revise this code frequently, negating productivity gains.
- Companies like Waydev, GitClear, Faros AI, and Jellyfish report significant increases in code churn associated with high AI tool usage.
- While AI tools provide higher throughput, the cost of tokens and subsequent rework make productivity improvements disproportionate to the expense.
- The gap between senior and junior engineers is highlighted, with junior engineers accepting more AI-generated code and facing more rewriting.
- Despite challenges, developers do not anticipate abandoning AI coding tools and see them as a permanent shift in software development.
Continue reading the original article