"Token maxing"—the practice of artificially inflating AI token usage to meet internal performance metrics—has emerged as a controversial byproduct of corporate AI adoption. At large organizations like Meta and Salesforce, engineers are incentivized to maximize token counts to avoid negative performance evaluations, leading to the creation of low-value autonomous agents. While this trend reflects a broader, often misguided, push for AI integration, it highlights a shift in the software engineering role toward orchestration and managing AI agents rather than traditional coding. Companies are increasingly investing in custom internal infrastructure, such as MCP gateways and proprietary coding agents, to gain competitive advantages and streamline development. Despite the inefficiencies and arbitrary metrics, these investments are driven by the necessity to recruit top talent and maintain a technological edge in a rapidly evolving landscape.
Sign in to continue reading, translating and more.
Continue
