Posts
Thoughts on AI agent infrastructure, protocols, and the tools we're building.
-
Everyone Is Asking 'A2A or MCP' — The Question Itself Is Wrong
A2A and MCP are not competing standards. One extends what agents can do; the other defines how agents talk. Here's why you need both.
-
Agent Runtimes Are Everywhere, But How Does the Message Reach the Agent?
The AI agent ecosystem has runtimes, frameworks, and tools. But nobody is talking about the gap between a user's message and the agent that processes it.
-
Two Layers of Agent Governance: Why Guardrails Alone Are Not Enough
Most teams think agent governance means guardrails on the agent's behavior. They're only seeing half the picture. The other half — message-flow governance — is where the real gaps hide.
-
87% of Multi-Agent Systems Fail in Production — The Reason Isn't What You Think
Multi-agent system failure rates in production are staggering. Research shows the root cause isn't model intelligence — it's coordination infrastructure.