This Week in AIOps https://aiopsfieldnotes.com/podcast/ en-us © 2026 AIOps Field Notes The AI news that matters if you actually have to run it in production. Every week, a fast two-host rundown of the models, agents, and protocols reshaping operations — and what each one means for the systems you build and keep alive. AIOps Field Notes AI news for people who run the infrastructure. A fast weekly two-host rundown of models, agents, MCP, and the reliability work that gets AI from demo to production. episodic false AIOps Field Notes james.son20@icloud.com https://aiopsfieldnotes.com/podcast/media/cover.jpg This Week in AIOps https://aiopsfieldnotes.com/podcast/ Ep. 1 — The Week AI Became an Ops Problem: Agent Reliability, Ox Alpha & Stateless MCP https://aiopsfieldnotes.com/podcast/ aiopsfieldnotes-aiops-ep1-20260828 Fri, 28 Aug 2026 12:00:00 -0400 The debut of This Week in AIOps: AI news for people who have to run it in production. This week's theme — the industry got honest that shipping agents is an operations problem, not a capability one. Cisco's data shows 85% of enterprises piloting AI agents but only 5% in production, and the consensus is that reliability, not capability, is the blocker — teams are entering a "rebuild era," adding the durable execution, state, observability and recovery they skipped. Meanwhile the cheap models are coming for the frontier: Z.ai confirmed it's behind "Ox Alpha," an open-weight reasoning model topping leaderboards, with weights out this week; and Inherent's Faraday — running on a 27B model — beat frontier systems at reproducing scientific papers. Finally, MCP went stateless with an OAuth 2.1 auth overhaul, so you can run MCP servers behind a standard load balancer on Kubernetes. Homework: build the boring reliability layer, keep an eval harness warm for open models, judge the workflow not the parameter count, and read the stateless MCP spec before you scale. This week's stories, for people who have to run AI in production:

  • The agent reliability reckoning. Cisco's figures: 85% of enterprises piloting AI agents, only 5% in production. The consensus this week — the blocker is reliability, not capability. It's a "rebuild era": durable execution, state management, observability and recovery are the work that gets an agent past risk, legal and compliance.
  • Z.ai's "Ox Alpha." The lab behind the GLM series confirmed it's behind the open-weight reasoning model that quietly topped leaderboards on OpenRouter. Weights released this week — an open, agentic model you can self-host changes the build-vs-buy math.
  • Inherent's Faraday. A 27-billion-parameter agent that reproduces published scientific findings — and beat much larger frontier systems at it. The lesson: the engineering around the model matters as much as the model.
  • MCP goes stateless. The Model Context Protocol dropped protocol-level sessions and added an OAuth 2.1 / OpenID Connect auth overhaul, so MCP servers finally scale behind a standard load balancer on Kubernetes.

The homework: build the boring reliability layer, keep an eval harness warm for open models like Ox Alpha, judge the whole workflow (not the parameter count), and read the stateless MCP spec before you scale your servers.

]]> 0:04:40 1 false