testkube

A test fails. The login button got renamed to "Sign in," so the old locator no longer matches. Classic, and it happens constantly. In this clip from the Cloud Native Testing podcast, Lakshmi walks through how an agent can handle this on its own: spot the broken locator, update the script, and rerun to confirm the fix. How it works, and where the human still comes in 👇 #SoftwareTesting #AITesting #TestAutomation

In 2019, Jensen Huang said English would become the new programming language. A lot of people rolled their eyes. In this clip from the Cloud Native Testing podcast, Lakshmi explains why he was right. The hard part was never the idea you had in your head. It was translating that idea into a language the computer would understand. AI has mostly removed that barrier. His take on what that shift actually means 👇 #AI #SoftwareDevelopment #VibeCoding

Should AI be allowed to fix your failing tests on its own? In this clip from the Cloud Native Testing podcast, Ole Lensmar lays out where he draws the line: the AI can trace the failure, fix it, and open a PR explaining what it changed. But a human still checks the work before it merges. His take on full self-healing, and why he still wants a person in the loop 👇 #SoftwareTesting #AITesting #QA

Last week we unveiled Testkube AI, the umbrella for everything we're building to help testing keep up with AI-driven engineering teams. MCP server, AI agents, autonomous capabilities, and more on the way. This article from @Sonali Srivastava is a hands-on look at one piece of that picture. She breaks down how the Failure Categorizer agent and AI Triggers automate the first level of test failure triage. The agent reads execution logs and Kubernetes events, then tags failures so engineers stop spending hours separating real regressions from flaky tests and infra noise. Worth a read: hubs.li/Q04gDmMS0 On May 21, we're going live with the full Testkube AI story. If you want to see how it changes the way modern teams test, that's where we'll demo it. Register: hubs.li/Q04gDvQb0 #TestkubeAI #AIOps #Kubernetes #TestAutomation

AI writes code. Pipelines run themselves. Releases ship at a pace that would have looked absurd three years ago. Testing is the last part of the workflow still waiting on humans. We've been building something that closes that gap. May 7, we tell you what it is. May 21, our CTO Ole Lensmar runs it live. Real cluster, real tests, real questions. Testing will never be the same. Save your seat. → hubs.li/Q04dSjTw0

On May 7, we're pulling the curtain on something we've been building. Want to see it in action and hear directly from the team behind it? Join us on May 21 for the launch webinar. Live walkthrough, real demo, your questions answered. Grab a seat: hubs.li/Q04d2Tmy0

A humanoid robot just set the world record for a half-marathon in Beijing. 50 minutes, 26 seconds. Then it crashed meters before the finish line and had to be carried across by a team of humans. Same thing is true for software. Your pipeline passing means your system works under the conditions your pipeline thought to check. The last mile, the real environment, the edge cases your simulator never modeled, that's where the wobble happens. That's the gap we've been working on. Different kinds of testing for a different kind of velocity, especially as AI-generated code starts outpacing the humans reviewing it. Shipping and surviving aren't the same thing. Here's what we're doing about it. 🗓️ Details drop May 7 🎥 In-depth webinar May 21 → hubs.li/Q04cLtWr0

No sales pitches. Just pure product. Tomorrow, our Sr. Solutions Architect, Jonathan Fortunati, is hosting our monthly Testkube live demo! Perfect for DevOps, QA Leads, & Platform Engineers wanting to learn how to: → Improve quality with cloud-native test orchestration → Decouple test automation from CI/CD → Deploy test catalogues to any environment on demand → Reduce reliance on testing service throttles Bring your questions for the live Q&A at the end! Choose the time that works best for you: 🕐 3:30 PM GMT: hubs.li/Q04c8K1r0 🕐 1:00 PM EST: hubs.li/Q04c8KDM0 #Testkube #DevOps #Testing #PlatformEngineering

Your test coverage dropped 23 points last quarter. Nobody deleted a single test. How? AI coding assistants. One engineer with Copilot generates 3-5x more code. Multiply that across a team. All those new code paths have zero corresponding tests. Your suite still passes. Your coverage number just doesn't mean what it used to. A service sitting at 78% branch coverage three months ago can be at 55% today. Just from the new code that shipped without tests. The denominator grew. The numerator didn't. This is the part that sneaks up on teams. Nobody made a bad decision. Nobody cut corners. The testing strategy just wasn't sized for this rate of change. And the defects that slip through don't show up in unit tests. They show up when services interact under load. On a Friday. So what actually helps? Decoupling test orchestration from your CI/CD pipelines. Get a single view of what ran, what passed, and what's still untested. Let it scale with how fast your team ships now. Full breakdown on our blog 👇 hubs.li/Q04bCGMP0

AI-generated tests sound great until you ship the edge case it missed. Sahana Nagabhushan, Founder of Dev Plaza, has been watching this space closely. The tooling is improving. But hallucinations, gaps in coverage, edge cases that never get a test written for them. That's not a theoretical risk. That's the quality problem teams are already hitting. Faster test generation only helps if the tests are actually right. → Full episode: hubs.li/Q048FZDp0

Your test suite is green. Main is passing. Nobody's sure which environments that actually covers. Three teams, four CI tools, results in Slack, S3, and a Grafana dashboard someone set up eight months ago. You know tests ran. You don't know if the right ones did. That's the gap between "we have tests" and "we have test orchestration." One control plane. Every test, every environment, every result. Triggered automatically, running in your own containerized infrastructure, not inside a pipeline runner. → hubs.li/Q048GQGK0

We're moving fast. But are we moving with confidence? Sahana Nagabhushan, Founder of DevPlaza, is asking the question a lot of teams are quietly sitting with: when AI writes the code, who's actually validating it's ready to ship? Functional requirements. Quality signals. Production trust. None of that comes free just because the code came fast. → Full episode: hubs.li/Q048FT9p0

Your coverage number didn't change. Your codebase did. 78% branch coverage last quarter. Still 78% today, according to the dashboard. But your team adopted AI-assisted dev in Q2 and has been shipping 3x the code volume since. The test suite hasn't kept pace. The real number is closer to 55%. Not because anyone deleted tests. Because coverage is relative to the codebase, and the codebase grew faster than the tests did. Defects don't disappear when velocity increases. They move into untested surface area, and the metric that's supposed to catch them looks fine. → hubs.li/Q049dgSw0

We’re heading to the Cloud Native London Meetup 🇬🇧 today! Join the lineup of Meghan McGowan of Cloudsmith, Mauricio Salatino of Diagrid, and Ole Lensmar at Thought Machine HQ or via livestream. Ole will be presenting: Keeping Up: Testing in the Age of AI and Cloud-Native As AI accelerates development and GitOps drives continuous delivery, testing needs to evolve just as fast. He’ll share how teams can scale continuous testing and use AI to better understand failures, automate remediation, and keep quality aligned with velocity. 📍London: hubs.li/Q0498bKV0 🎥 Livestream: hubs.li/Q0498jPd0 If you’re there in person, come say hi and grab a plushie 👋

Our team spent three days in Amsterdam talking to engineers, platform leads, and architects. The pattern across every conversation: teams aren't evaluating cloud-native anymore. They're operating in it, and the infrastructure has gotten complex enough that "it deploys" is no longer a sufficient signal. AI workloads on Kubernetes. MCP in production. Multi-cluster environments that need to behave correctly across every node, every edge deployment, every customized distribution. Here's what stood out across those three days → hubs.li/Q048Y76G0

Most teams building on Kubernetes have automated their tests. Few have actually orchestrated them. There's a difference. Automated means tests run. Orchestrated means consistent execution across environments, centralized visibility, one platform for all your tools. That's Level 3 on the testing maturity curve, and it's where the biggest ROI actually lives. Most teams are stuck at Level 2. Tests exist, pipelines run, results scatter across tools and dashboards. Nobody has a unified view. The jump from 2 to 3 is the one worth making first. Full model (and what to do at each stage) in the blog. → hubs.li/Q048GtQt0

AI writes the code. Humans still have to own the outcome. Sahana Nagabhushan, Founder of DevPlaza, said it well: AI can be extremely helpful, but it needs to be the friend, not the main person. Shipping faster without the confidence to back it up is just risk at scale. The teams getting this right aren't slowing down. They're building the layer between "code generated" and "code trusted." → Full episode: hubs.li/Q048FV_r0

We're at KubeCon. The floor is moving fast. AI for code. CI for builds. Testkube for tests. Every part of the stack now has a purpose-built layer. Testing shouldn't still be bolted onto your pipeline, slowing builds down and scattering results across environments. When testing runs as its own orchestration layer inside your infrastructure, you get faster feedback, less flakiness, and full visibility into what ran and why. That's how teams keep quality up without slowing down. Come talk to us at booth 1162 📍

"We don't write tests. We integrate the tools that do." That's how Sahana Nagabhushan, founder of Dev Plaza, describes her approach. She's building a production readiness layer that sits between "I finished coding" and "I can actually ship this." The problem she keeps running into: teams skip the steps that matter. Not because they don't know better. Because the tooling to enforce those steps was always something you had to build yourself. Her take on AI-generated code: the confidence to ship doesn't come from the model. It has to come from what happens after the model writes the code. The conversation covers why every company she's worked at rebuilt the same internal solution (at a cost of $3M to $16M per year), where AI-generated tests are still falling short, and why production readiness matters more now, not less, in an age of AI velocity. New episode of the Cloud Native Testing Podcast with Ole Lensmar is live. 🎥 hubs.li/Q047ZSkv0

Taking Kubie to the Alps ⛰️ Some things just work better when they're in the right environment. Same goes for testing. Tests that run inside your containerized environment, close to where your code actually lives, behave differently than tests bolted onto the outside of your CI pipeline. Less flakiness. More signal. Faster feedback. That's the idea behind test orchestration: stop treating testing as a CI afterthought and run it where it belongs. See Testkube in action → hubs.ly/Q046VQyN0