Microsoft memo urges engineers to cut AI token use and default to OpenAI's GPT-5.6 Sol

A leaked internal Microsoft memo has sparked debate over how big tech companies control AI spending, after CoreAI leader Jay Parikh told employees to rein in heavy token use and default to OpenAI’s GPT-5.6 Sol for internal coding work. According to the memo, Microsoft wants teams to shift more workloads to OpenAI models to get more value from their token investment, while still allowing other model options when needed. The directive applies to Microsoft’s internal engineering teams, including GitHub and Visual Studio, and is meant to improve efficiency rather than change the customer version of GitHub Copilot. The episode highlights how quickly AI usage costs have become a management issue inside companies that are both building and buying these systems.