Agentics Day: MCP + Agents at #KubeCon dug into the production realities of running AI agents.
My 8 takeaways, spanning openness, security, early lessons learned and the infrastructure needed:
https://t.co/KVYx3huB64
#MCP#AIAgents
@KristopherLeads@Zuplo Seems like a sensible approach. I wonder if that practice could also discover what approaches/workflows are too expensive to AI-ify.
Sales doesn't use AI like engineering does. Finance doesn't use AI like marketing does. So, why give every department the same AI budget?
My latest for @Zuplo looks at setting limits, monitoring usage, and optimizing AI spend across departments.
https://t.co/7BU41rv1Z1
In case you missed yesterday's AI Pen-Testing Conference, here's the full recap. It's got the session recordings, highlight reel, panel discussion, demos, and keynotes.
https://t.co/HyAxObP6Mq
So, we're all using LLMs now. Cool. But are you actually monitoring them in production?
My latest for @Zuplo looks at 7 LLM observability metrics to track, from latency and accuracy to token usage, costs, and more.
https://t.co/LlyA6msT5E
And that's a wrap! My final Road to KubeCon column for @thenewstack is out.
This week: Kubernetes workload conformance, OpenTelemetry OTTL v1, shared GPUs, and agent security horror stories.
Thanks to TNS, HPE, and @CloudNativeFdn for a fun series!
https://t.co/wQjTGg0d5U
"Agentic workloads have no patience for a cold node."
Bursty AI agents are a recurring pain point on K8s. My latest Road to KubeCon column for @thenewstack digs in:
https://t.co/1IGsyASnkv
Here are my slides for my opening keynote at @nordicapis Summit 2026! My 25-minute talk looked at how AI agents are forcing major shifts in how they interact with APIs.
https://t.co/A1xh95nuxe
This week’s Road to KubeCon on @thenewstack
looks at new projects for hosting agentic workloads and running agents on Kubernetes.
Also: Platform Engineering Day, AI coding, compliance ownership, and one cautionary tale about "infrastructure theatrics."
https://t.co/T0sxEsJ7oS
As AI agents make code easier to generate, the pendulum is swinging back toward up-front planning, understanding, and judgment.
Happy to contribute to Arthur D. Little's PRISM publication:
https://t.co/tyNoHGySDl
More agentic development is coming, predicts Angie Jones (@techgirl1908), VP at @AgenticAIFdn.
That said, it'll take more skills development around nondeterminism, MCP governance, and agentic observability.
My latest for @TheLeadDev: https://t.co/mVuejAiHN0
Excited to share that I'll be speaking at SecureMaine this week! I'm giving a talk on Oct. 8 on securing agentic AI, then moderating a roundtable later in the day.
https://t.co/tCp3vBulzV
Meta is using AI agents to turn senior engineers’ judgment into reusable skills, helping recover hundreds of megawatts of power and speed up regression detection/remediation.
Interesting case study of AI agents tied to real efficiency gains:
https://t.co/SYSMma0KoA
Why go to a conference when there's endless content to consume? Expert guidance, curated experiences, hallway discussions, serendipitous connections, and firsthand knowledge can still set in-person events apart. https://t.co/ZHvqwAOd2h