Gergely Orosz (Pragmatic Engineer) discusses 'token maxing' — the practice of artificially inflating AI token usage at companies like Meta, Microsoft, and Salesforce where AI usage is tracked via leaderboards or spending targets. He explains how leadership pressure to adopt AI tools led to gaming these metrics, drawing parallels to lines-of-code metrics from earlier productivity tools. The conversation covers how AI is reshaping the software engineer role (absorbing tester, DevOps, and product responsibilities), why large companies are building custom internal AI infrastructure (MCP gateways, background coding agents), and why teams that succeed with AI tend to be low-ego and open to abandoning prior assumptions. Orosz also shares the origin story of the Pragmatic Engineer newsletter, which became the top paid tech newsletter within months of launch.
Sort: