Tech Leaders Critique Anthropic's AI Protocol Direction, Eyeing Future of Dev Platforms and AI Agents

The latest ‘Ask Us Anything’ session, hosted by Victor and Scott, ignited a fervent discussion on the evolving landscape of AI development, platform engineering, and cloud infrastructure. A significant point of contention arose around Anthropic’s perceived shift from the Message Control Protocol (MCP) to ‘skills’, with participants highlighting the fundamental differences: MCP as a cross-agent communication protocol versus skills as glorified, project-local prompts. Experts expressed disappointment with Anthropic’s apparent disengagement from the MCP standard they initiated, advocating for an MCP 2.0 that would expose TypeScript APIs and enable LLMs to write execution code rather than rely on inefficient sequential tool calls. This broader conversation extended to best practices for AI agents, emphasizing the importance of breaking down tasks into smaller, manageable chunks, judicious context management, and leveraging deterministic tools and ‘hooks’ to validate AI-generated work, thereby improving reliability and self-correction capabilities.

Further insights delved into the role of internal developer platforms and the future of Kubernetes. Backstage was reaffirmed as a ‘portal, not a platform’, requiring dedicated product investment and a customer-centric approach for success. The consensus pointed towards Kubernetes becoming ‘invisible’ to end-users through higher-level abstractions like Crossplane, rather than being supplanted by opinionated managed container services such as Azure Container Apps. The discussion also revisited HashiCorp Nomad, which, despite its technical merits, was deemed a ‘failure’ due to its inability to foster a robust ecosystem compared to Kubernetes. The session concluded by acknowledging AI’s profound and multifaceted impact on DevOps and cloud jobs, enhancing the efficiency of senior engineers but presenting risks for junior roles, underscoring the necessity for deep domain expertise and a fundamental shift in established workflows and processes.