November 2026 Summaries

Tailscale has made OAuth apps generally available, enabling AI agents such as Grok Bot to securely connect to private resources on a user’s tailnet through OAuth 2.0 without exposing internal infrastructure or sharing credentials. Unlike existing OAuth clients that create devices under a shared service identity, OAuth apps provision devices tied to the authorizing user, preserving normal tailnet policies, audit logs, and device quotas. Grok Bot can use this integration to reach private Model Context Protocol servers, while administrators can restrict its network access independently from the user’s other devices through custom device posture attributes on Premium and Enterprise plans or tag-based policies on other plans. Tailscale’s example policy permits engineers’ regular devices to access production databases while limiting their Grok Bot instances to designated MCP servers. Users can connect Grok Bot through its desktop application’s Tailscale plugin, and developers can use Tailscale’s OAuth Apps documentation and device provisioning guidance to build similar integrations.

AI coding tools could expand programming to roughly a billion people by allowing non-specialists to describe and create small applications without learning traditional languages or inspecting source code. Using a coin-flipping app example, the piece argues that generating code is becoming easier while deployment, networking, identity, storage, and sharing remain unnecessarily complex, pushing even simple software toward cloud-centric professional infrastructure. It presents Tailscale and its Aperture initiative as efforts to make devices, models, services, and security controls more easily connected and interchangeable, enabling people to run software on local machines, private hardware, or cloud resources as needed. It warns that AI providers may recreate the vendor lock-in associated with modern cloud platforms by bundling models, agents, connectors, hosting, and identity systems into closed stacks. Instead, it advocates open, modular systems with strong interfaces and security boundaries, allowing users, organizations, and countries to change models and providers without abandoning their applications or data. The discussion also frames AI sovereignty as the practical ability to move workloads and retain local control, rather than requiring every country to build an entire domestic AI stack, and concludes that greater user ownership of powerful personal computing resources could support a more competitive and decentralized software ecosystem.