mvizard retweeted
A single 64-core server running 1.42 million isolated WebAssembly sandboxes on Kubernetes is what Cosmonic showed off in its KubeCon + CloudNativeCon North America 2026 demo, enough environments to host a great many AI agents without giving each one its own pod.
Its CTO argues Wasm suits the bulk of enterprise agents because most are focused on a narrow range of tasks, while a coding agent that needs more room still belongs in a micro VM.
See how the sandbox density argument fits the wider race to make Kubernetes fit for agents: zpr.io/XUSs4Je9zVVR
#Kubernetes #CloudNative #WebAssembly #AIAgents #WasmCloud #CNCF #KubeCon
mvizard retweeted
Measuring AI spend isn't enough. Tying it to value is. Mike Vizard talks with @tempohq Chief Product Officer Kevin Nanney about tying AI spend to business value.
Watch: techstrong.tv/videos/intervi…
#AIGovernance
mvizard retweeted
Airbyte has added semantic search to the MCP server it already provides so AI agents can understand context, meaning and intent across data stored in multiple repositories, running the embedding model on its own infrastructure instead of sending text off to a separate platform.
Its product marketing lead argues this makes it possible for anyone to be their own data engineer, since you can run analytics from a chat interface, though none of that helps if the data the agent reaches was unreliable to begin with.
Here's how semantic search changes what agents can do with your data: buff.ly/4UyMkFf
#AI #AIAgents #DataEngineering #MCP #Airbyte #SemanticSearch
mvizard retweeted
Broadcom used KubeCon North America 2026 to contribute a VM API to the CNCF as a sandbox project, aiming to make virtual machines first-class citizens in cloud-native environments, and threw its weight behind a new Workload Conformance Working Group that wants an objective standard for whether workloads are truly portable.
The timing is telling, since a Linux Foundation report out today finds VMs still account for 53% of production applications, with just over a third of respondents expecting to expand their VM footprint as they deploy AI and agentic workloads.
See what the contribution means for teams still weighing their platform options: zpr.io/FDFME7c2kEZM
#Kubernetes #CloudNative #VMware #Broadcom #CNCF #KubeCon
mvizard retweeted
Microsoft outage, AWS AI-built cloud tools & sea sensors: @mvizard @poller @TracyRagan @jswartz @JeffReichCSO Amanda Razani. #Cloud #AI #AWS #Sensors #Resilience techstrong.tv/techstronggang www.teapawt.org/i/broadcasts/2qVXZYlln…
mvizard retweeted
Red Hat at KubeCon + CloudNativeCon North America 2026 extended OpenShift 5 with AI capabilities, including a built-in MCP server that makes the platform reachable by any AI agent through a simple update and a Lightspeed agent trained to automate tasks across Argo and Red Hat Developer Hub.
With most organizations lacking the software engineering expertise to manage clusters at scale, the pitch is that AI agents finally let an average IT administrator handle cloud-native environments at a scale that previously demanded a highly trained engineer.
Learn what else version 5 adds, from multi-tenant isolation to a post-quantum control plane: zpr.io/WqxP4M8kCdAT
#Kubernetes #RedHat #AIAgents #CloudNative #MCP #KubeCon #CloudNativeCon
mvizard retweeted
IT teams are struggling to inject AI agents that generate probabilistic outputs into workflows that need to be executed the same way every time, and Mirantis just made available a consulting service through which teams can train those agents in OpenStack or Kubernetes using a set of best practices.
The platform ships with built-in Prism agents and third-party agents that detect production issues, diagnose root causes and propose fixes, with defined access, isolated sandboxes, human approval before fixes, spending limits and audit trails, because as its product chief warns, deploying an agent in a sandbox without understanding the larger system around it is a recipe for disaster.
The details, including the warnings about overdependence on a few engineers, are here: buff.ly/WilMDez
#Kubernetes #CloudNative #AIAgents #Mirantis #KubeCon #DevOps
mvizard retweeted
AI agents require different security controls depending on their capabilities and workloads.
At KubeCon, the Techstrong Gang explores AI agent security, AI SRE in production, and agent governance, including the RAMP test.
Watch the discussion: 👉 bit.ly/4eBaNgp
#KubeCon #AISecurity
mvizard retweeted
Diagrid shipped Dapr 1.19 with a reworked workflow engine that lets workers keep their own execution history instead of receiving the full record on every run, which matters when AI agents run parallel tasks for hours and a disruption means recovering a workflow rather than restarting it from scratch.
The update, rolled out around KubeCon + CloudNativeCon North America 2026, also adds an opt-in mode where the Scheduler handles actor placement, plus an alpha vector embeddings API and a Binary Store for streaming large files.
See what the release changes for teams running agents at scale: buff.ly/uK0yo66
#Kubernetes #CloudNative #Dapr #AIAgents #Diagrid #CNCF #DevOps
mvizard retweeted
Kong is expanding its AI platform with new cost controls, observability and governance tools. Its Token Vault also helps secure AI agent access to enterprise resources.
Read more: buff.ly/CSFCVap
#AgenticAI #AIGovernance #Observability
mvizard retweeted
Attackers want your AI agents. The surface is wide open. Mike Vizard talks with SentinelOne Distinguished AI Research Scientist Gabriel Bernadett-Shapiro about AI agent hijacking.
Watch: techstrong.tv/videos/intervi…
#AISecurity
mvizard retweeted
ControlTheory's Dstl8 2.0 adds metrics and traces to its AI observability platform. Analyzing telemetry at the source could help DevOps teams cut costs and detect issues faster.
Read more: buff.ly/vaHZ0il
#Observability #DevOps #AIOps
mvizard retweeted
Apify has extended its marketplace with a natural language interface that invokes an AI agent to build workflows from tools hosted there, and made Server Actors generally available so those workflows can span multiple services like a set of microservices.
Its Actor runs already pass 267 million per week across 100,000 Actors built by 4,500 developers, and as CEO Jan Curn puts it, there is going to be a Cambrian explosion of software.
See how the marketplace model is meant to keep pace with it: buff.ly/l0tKVW5
#AI #AIAgents #Apify #MCP #LowCode #Automation #TechNews
mvizard retweeted
IT is changing, but just throwing AI at it isn't the answer. Start by solving 3 fundamental problems slowing down your everyday IT. Mike Vizard (Techstrong Group) with Rakesh Jayaprakash (@manageengine) techstrong.tv/videos/intervi… #AIStrategy #ITOps
mvizard retweeted
Broadcom has contributed VM API to the CNCF Sandbox, an open-source project built on VM Operator that lets virtual machines be provisioned and managed through the same Kubernetes APIs and declarative workflows teams already use for containers.
A Linux Foundation survey of 492 practitioners found VMs remain a majority workload at 53% of production applications and only 10% of organizations run none, while AI and agentic workloads are cited by 34% as a reason to expand their VM footprint.
Here is why provider-neutral governance could matter for mixed estates: zpr.io/bZe23AKJU6sy
#Kubernetes #CloudNative #VMware #Broadcom #CNCF
mvizard retweeted
CrowdStrike has added a kill switch for AI agents to Falcon Guardian, using its endpoint software to terminate an agent, block it from launching again and cut off network access, alongside agreements to acquire Fiddler AI and Dreadnode and extend its AIDR platform to Claude Cowork and AWS Bedrock AgentCore Guardrails.
The trouble is that cybersecurity teams are often unaware an agent was even deployed until there is an actual incident, and without visibility it becomes all but impossible to manage and secure them even though the security team will be held accountable for the mayhem.
Read the full article on what CrowdStrike is building and what it costs: buff.ly/GV6bDUh
#Cybersecurity #AIAgents #CrowdStrike #AIGovernance #InfoSec
mvizard retweeted
Teams can't scale manually. Confidence scores unlock automated remediation. Mike Vizard talks with @qualys CTO Dilip Bachwani about the shift to autonomous fixes.
Watch: techstrong.tv/videos/qualys-…
#CyberSecurity
mvizard retweeted
Cloudsmith has launched a registry for AI agent skills, giving DevSecOps teams greater control over access, security and usage. The move aims to reduce software supply chain risks.
Read more: buff.ly/45niz3B
#DevSecOps #AgenticAI #Cybersecurity
mvizard retweeted
Teleport's answer to AI agents wandering production with standing privileges is Teleport Beams, now generally available, a runtime that hands each agent a unique identity and privilege set tied to a single task and revokes both the moment the work is done.
An agent proposes an execution plan and the access it needs, a human or an LLM acting as judge approves or revises it, a separate runtime executes with only the approved roles, and Agentic Classifiers score each session against MITRE ATT&CK to flag goal hijacking or emergent coordination.
Walk through how the runtime provisions, judges and revokes access: buff.ly/OQzMJph
#Cybersecurity #AIAgents #Teleport #IdentitySecurity #ZeroTrust #InfoSec
mvizard retweeted
SUSE used KubeCon + CloudNativeCon North America 2026 to make a bevy of updates to its Kubernetes management tools, including specialized AI agents invoked through its Liz natural language assistant, which can now call MCP servers over OAuth2 to integrate with external tools, agents and applications.
The additions include a beta Hosted Control Plane mode for Rancher Prime Virtual Clusters, a CLI for programmatically testing infrastructure updates and updated Observability StackPacks, all designed to reduce time to value and address the chronic skills gap that Puri says is the biggest challenge most IT teams face deploying workloads on Kubernetes clusters.
See what SUSE added and why distribution choice now comes down to extra tooling: zpr.io/ttEhHm8PpRb7
#Kubernetes #CloudNative #SUSE #AIAgents #KubeCon #PlatformEngineering






