The podcast explores the emerging enterprise battle over AI model ownership, sparked by Thinking Machines Lab's release of Inkling, a 975-billion-parameter open-weight model. While Inkling shows promise with its unique, non-distilled training approach, early tests reveal it still lags behind top frontier models. The core debate revolves around control: open-weight models like Inkling offer enterprises data sovereignty and cost advantages, contrasting with Microsoft's Frontier Tuning approach which requires trust in a provider. This signals a shift from a limited model selection to a diverse ecosystem, driven by near-state-of-the-art performance and geopolitical factors. The episode also covers Cursor's ambition to become a top-tier model developer, Apple's efforts to re-enter the AI race, and Microsoft's strategy to compete with OpenAI and Anthropic by emphasizing cost-effectiveness and security of its in-house MAI models. The experimental period presents opportunities for nimble enterprises to test various models and fine-tuning options.