Newshot

At KubeCon 2026, the illusion of AI control meets the reality of implementation

Summary

I revisited Salt Lake City for KubeCon/CloudNativeCon NA 2026 hoping to find practical approaches to delivering customer value amidst a rapidly shifting kaleidoscope of artificial intelligence-driven solutions that are making it increasingly difficult to trust our own eyes and ears. This year, the host Cloud Native Computing Foundation set its theme as “Cloud Native Without […] The post At KubeCon 2026, the illusion of AI control meets the reality of implementation appeared first on SiliconANGLE .

Original text(via SiliconANGLE)

UPDATED 11:41 EST / NOVEMBER 16 2026

At KubeCon 2026, the illusion of AI control meets the reality of implementation

GUEST COLUMN by Jason English

I revisited Salt Lake City for KubeCon/CloudNativeCon NA 2026 hoping to find practical approaches to delivering customer value amidst a rapidly shifting kaleidoscope of artificial intelligence-driven solutions that are making it increasingly difficult to trust our own eyes and ears.

This year, the host Cloud Native Computing Foundation set its theme as “Cloud Native Without Limits,” and judging by the record-shattering pace of new community developments with more than 230 active projects, and the increasing adoption rate of AI for building and consuming applications, the momentum will only grow.

“If you look back at last year’s KubeCon in Atlanta, OpenClaw was not released yet, MCP was less than a year old, and inference in cloud native was just getting started,” Chris Anzyzcyk, chief technical officer of CNCF, said in the opening remarks.

“We knew we were moving out of a phase where most of the AI compute was used for training and we were going to massively ramp up inference,” said Jonathan Bryce, executive director of CNCF. “Just in the last eight months, we’ve seen a 4X increase in inference traffic across the industry. That’s really an incredible increase in the AI footprint. Inference is no longer just an opportunity, it’s now a cloud-native production workload.”

With more than 350K CNCF contributors and 21.2 million cloud-native developers worldwide, fully three-quarters of them are already using AI coding agents, according to the latest State of Cloud Native Development Q3 2026 report, and a show of hands in the room hinted at a much higher presence of AI tooling on the engineer’s desktop.

Likewise, two-thirds of self-described AI-driven organizations are standardizing on Kubernetes and related open-source technologies.

There’s a clear connection here, but AI is not the whole story. Enterprises still need to modernize their application fleets, and deploy, scale, secure and observe their most critical systems, in order to get beyond the AI hype and realize business value.

With an open mind and a skeptical eye, here were some of the interesting new perspectives and vendor solutions I encountered at the show.

Building open paths to Kubernetes and AI productivity

With the rapid enterprise rollout of agentic AI, every developer is being challenged to apply human decision insight while working at agent speed, one of the major themes of the show.

Getting there will take more than new technologies, it will also require a lot of collaboration and understanding. The Kubernetes AI Conformance Program launched at KubeCon NA 2025 now has 42 major participating vendors agreeing to consistent standards, and this year there are new Kubernetes AI Certifications for upskilling engineers, whether they are just starting out, or if they are “Golden Kubestronauts” who already passed every course CNCF offers.

“Kubernetes has become the standard API for deploying a workload to a computer, including AI,” said Dan Ciruli, vice president and general manager for cloud native at Nutanix Inc. “The reason, in my opinion, is because the CNCF gave a safe haven for anyone from the smallest startup of two people to the largest companies on Earth to feel safe in donating labor, and code, and IP, and agreeing on things. And that, in turn, gave everybody, from governments to financial services to hot startups, the confidence that they could consume it.”

The firm recently introduced their Nutanix Agent Gateway and an NKP Metal solution that allows Kubernetes clusters and workloads to run on bare metal as well as virtual machines in EKS, AKS or GKE for what they call a “dual-native” experience.

“How are you going to safely and sustainably connect thousands of agents to the systems of record that actually run the business such as customer loyalty or inventory?” asked Matt DeBergalis, CEO, Apollo Graph Inc., doing business as Apollo GraphQL, which supports the popular open source GraphQL query language, presenting a new semantic domain model for deterministically interpreting AI prompts and controls for each record.

Eliminating resource constraints, both digital and human

Last year CNCF announced a Dynamic Resource Allocation or DRA capability, which unlocked new possibilities for the community to optimize AI workloads running in Kubernetes. This year, sandboxing, or running agents in secure instances to prevent escapes, hijacking and overprovisioning, was the watchword.

Development has moved fast in this area, with two new projects: a Agent Substrate layer contributed by Google that optimizes resource consumption of thousands of agent sandboxes, and OpenShell, an Nvidia Corp.-contributed project for secure sandboxed runtimes.

Startup Apelogic Inc. doing business as HyperShell built an operational agent envelope based on OpenShell that defines guardrails what systems an autonomous agent can access, and what actions it can take, based on the business or system intent the agent was intended to serve.

Cast AI started out as a cloud cost optimization layer for rightsizing clusters on major hyperscalers, but as cloud native infrastructure started taking over enterprise environments, it now finds itself automating and optimizing more AI workloads across CPUs and GPUs, memory and storage. New at KubeCon, their Kimchi code agent governs and deploys application infrastructures from a prompt, which allows developers shift left agentic reviews before production tuning.

Floxdev Inc., which I covered as a startup with their Nix-based lean containerized environment deployments in 2025, put its Flox platform directly in the AI delivery path, giving agents skills to stamp out identical complete environments that include all dependencies, without reproducing any element of the software supply chain inventory that wasn’t intended to change. This moves agentic loops faster with lower compute costs, but the real value is when human engineers can reduce future maintenance and security remediation costs.

Observability isn’t just for SREs anymore

This year’s CNCF graduation of the OpenTelemetry project couldn’t come at a better moment, making OTel the lingua franca of application performance, scalability and reliability.

While observability was traditionally considered the domain of engineers and operators, now everyone from data analysts to business executives may be using an LLM or an agent to interact with enterprise applications, potentially changing software, or delivering more agents, in the process.

Splunk, a division of Cisco Systems Inc. has added to its agent observability suite with a specialized language model called Eval-G1 that can run prompt-and-response evals in sub-second times at sub-cent costs, giving engineers the ability to interrupt a faulty response in real time, before it reaches customers. A new Evaluation Studio helps teams generate guardrails in a low-code graphical environment. Also cool, a tokenomics feature (from the recent acquisition of Galileo, which we just covered) that analyzes token spend versus system and business performance metrics.

“It’s been exciting to see how OpenTelemetry, which started out as tens of thousands of adaptors for every application and client, has transitioned with the AI revolution so that now, we have frameworks that just emit OTel natively,” said Morgan McLean, senior director of product management at Splunk and OpenTelemetry co-creator. “So Splunk, and our competitors, can just launch new observability features without putting 30 people on writing new integrations.”

Since a lot of AI projects start out as experiments, observability stalwart Dynatrace LLC has made several moves to meet smaller teams where they are in validating each feature for performance and reliability before enterprise-scale rollouts. It recently acquired Bindplane for OTel-native telemetry pipelines, BlueBox, which gives developers a custom agent for debugging and checks within a coding IDE, and Arize, which maintains the popular Phoenix open-source tool for AI evaluations.

New Relic Inc. released a new Autopilot Kubernetes Agent, a virtual SRE that users can interact with via natural language to look across telemetry and APM data to assess the health of the cluster and application, triage issues with a complete stack trace, and recommend resolutions.

“Especially when you bring agents into your environment, you can’t just take your telemetry and throw it over to a general-purpose LLM, because there are known deterministic patterns and historical insights about Kubernetes we’ve been working to resolve for years,” said Jemiah Sius, vice president of market strategy and developer relations at New Relic.

Making agent deployments safer, by making them easier to sustain

A lot of the governance difficulty with agents and sandboxes and policies isn’t a technical problem; it is about user experience. Human engineers don’t like being “in the loop” for AI, aka “Meat Proxies,” that continuously pass along approvals. Humans need to be in control, with less toil and more focus on what customers want out of AI applications.

“You can try to create a secure agent sandbox with a policy where everything is disabled by default until a human approves it, but eventually, someone gets tired of being in that approval loop,” said Simone Morellato, senior director of product marketing, Rafay Systems Inc., an AI infrastructure delivery platform used by leading neoclouds. “So you give an agent access to one thing, and eventually you end up giving them access to everything. And at that point, you are back at square one.”

Since I covered Edera Inc. last year, this hardened runtime container vendor is finding new traction in this agentic age by continuing to strengthen the stack from the bottom up, with an agent runtime and OCAP bounded authority that acts as a switchboard for discretely defined agent capabilities, rather than assigning agents credentials or access to MCP servers or APIs or other agents that can be expanded or exploited for another HuggingFace-type incident.

“You can consume capabilities in our system and publish capabilities,” said Edera founder and Chief Technology Officer Alex Zenla. So, for example, for an MCP server, we publish each tool call as a capability. And then when you spin up your agent, you ask for the tool calls that you want. And then it consumes just those capabilities. And we’re able to enforce policies because we have one central pathway for everything to go through, so we can know exactly which agents are using which capabilities.”

“I love the focus on reliability that I’ve seen in the talks and on the floor here at KubeCon,” said Kolton Andrus, CEO of Gremlin Inc., which was there demonstrating their new Foresight AI solution that proactively injects failure conditions in order to identify and remediate failure risks. “You can no longer wait for things to fall apart before taking action. “Everyone’s moving 100 miles an hour trying to keep up with AI, and waking up to the fact that reliability can’t be an afterthought anymore. You can no longer wait for things to fall apart before taking action.”

The Intellyx take

This being my eighth KubeCon, I keep coming back because the community atmosphere here is unparalleled. There’s a real sense of camaraderie among developers and engineers from wildly different backgrounds, as they build new fundamentals and enjoy the journey toward a cloud native future.

We’ve come a long way over the last eleven years since Kubernetes was introduced. Interoperability, portability, control planes, observability, networking, identity, security and many more mature capabilities are there for the taking. Now AI is being layered on top. But that’s not the story here.

What really impresses me is how much this open source community gives back for the benefit of everyone. All of these people have real jobs to do, and yet they invest their own time, as much as the software industry invests in them.

Indeed, as CNCF’s own Jonathan Bryce said: “Some problems can only be solved in the open.”

Jason English is a principal analyst and chief marketing officer at Intellyx. He wrote this article for SiliconANGLE. At the time of writing, Gremlin and Rafay are current Intellyx customers, Dynatrace, New Relic, Nutanix and Splunk are former customers, and none of the other companies mentioned here is an Intellyx customer. No AI was used to generate this content. CNCF covered the analyst’s attendance cost for the event, a standard industry practice. ©2026 Intellyx B.V.

A message from John Furrier, co-founder of SiliconANGLE:

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

  • 15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more
  • 11.4k+ theCUBE alumni — Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network

Are you an AWS customer? Support SiliconANGLE financially by buying your AWS services from our Marketplace portal page and links: https://siliconangle.com/aws-marketplace/

Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.

  • Alecto raises $20M seed funding to make UK property law more accessible to consumers with AI
  • Sumo Logic expands AWS Observability Solution with one-click deployment and new regions
  • Google says an AI framework turned security patches into exploits with no human in the loop
  • Goodbye, manual work: AI-native CRM platform Dreamteam raises $53M to automate enterprise sales
  • Palo Alto Networks report finds organizations struggle to track and secure AI agents in the cloud
  • Exaforce's ExaU lets security staff spin up a personal team of AI agents
  • Alecto raises $20M seed funding to make UK property law more accessible to consumers with AIAPPS - BY KYT DOTSON . 47 MINUTES AGO

Alecto raises $20M seed funding to make UK property law more accessible to consumers with AI

APPS - BY KYT DOTSON . 47 MINUTES AGO

  • Sumo Logic expands AWS Observability Solution with one-click deployment and new regionsCLOUD - BY DUNCAN RILEY . 3 HOURS AGO

Sumo Logic expands AWS Observability Solution with one-click deployment and new regions

CLOUD - BY DUNCAN RILEY . 3 HOURS AGO

  • Google says an AI framework turned security patches into exploits with no human in the loopSECURITY - BY DUNCAN RILEY . 3 HOURS AGO

Google says an AI framework turned security patches into exploits with no human in the loop

SECURITY - BY DUNCAN RILEY . 3 HOURS AGO

  • Goodbye, manual work: AI-native CRM platform Dreamteam raises $53M to automate enterprise salesAPPS - BY MIKE WHEATLEY . 3 HOURS AGO

Goodbye, manual work: AI-native CRM platform Dreamteam raises $53M to automate enterprise sales

APPS - BY MIKE WHEATLEY . 3 HOURS AGO

  • Palo Alto Networks report finds organizations struggle to track and secure AI agents in the cloudSECURITY - BY DUNCAN RILEY . 4 HOURS AGO

Palo Alto Networks report finds organizations struggle to track and secure AI agents in the cloud

SECURITY - BY DUNCAN RILEY . 4 HOURS AGO

  • Exaforce's ExaU lets security staff spin up a personal team of AI agentsSECURITY - BY DUNCAN RILEY . 5 HOURS AGO

Exaforce's ExaU lets security staff spin up a personal team of AI agents

SECURITY - BY DUNCAN RILEY . 5 HOURS AGO

The text above was extracted automatically from the source page and may miss images, tables or code blocks. The original is authoritative.

Published by SiliconANGLE
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗