On Cloud Dialogues, I break down the current operating reality for AI and cloud: regional model limits, data governance blockers, the widening adoption curve, and why reliability and sustainability matter as much as cost.

I joined Karl and Jon on the Logicata Podcast to unpack a stack of AWS news: GPU price increases, MCP-driven modernization with Quiro, new Dublin data centers, my S3 image-serving post, and AWS layoffs, plus a short detour into the "champing vs chomping" debate.

I walk through how I build scalable, self-orchestrating agent workflows with A2A and MCP. We cover agent discovery, shared context, and the patterns that keep multi-agent systems from turning into chaos, plus a practical AWS serverless stack to make it real.

Part 5 of the Andres Moreno series dives into meta-agents: where they sit in the agent loop, when they make sense in production, and how they act as output guardrails for your primary model.

Part 4 of our enterprise agent series with Andres Moreno covers input/output validation, when to skip the model, and how guardrails keep agents reliable, cheaper, and safer in production.

Allen Helton and Andres Moreno talk about why tool design is real engineering: treating tools as APIs for reliability, auditability, multi-tenant safety, and scalable enterprise agent systems.

Part two of our enterprise agent series focuses on context: how much is enough, what happens when you overload the model, and how to keep prompts sharp without slowing everything down.

Allen Helton and Andres Moreno show why system prompts are the difference between reliable enterprise agents and chaos, using a fraud-detection demo to test prompt depth and context overload.

At POST/CON 25 I share practical ways to make your API impossible to ignore on the Postman API Network, from table-stakes setup and Easy Auth to storytelling, branding, and verification.

On Between 2 Devs at Postcon, Hannah Seligson and I unpack MCP’s “full access” problem for LLMs, why that’s a trust and safety issue, and how Postman’s endpoint-level controls make agent tooling safer and more practical.