Agentic AIs are on everyone’s lips. But are they your tools, or your partners? And what if they turn into a Gremlin after midnight? Either way, management and observability are the keys
The Norwegian Labor and Welfare Administration (NAV) pays out 33% of Norway’s state budget. A thousand developers rely on its central logging infrastructure every day. At OpenSearchCon Europe 2026, platform engineer Hans Kristian Flaatten and Aiven Product Director Dmitry Kan walked through how they swapped out the underlying search engine while the lights stayed on – and what happened when a last-minute integration problem turned up three days before Christmas.
_1.png02d9.webp "NAV's Hans Kristian Flaatten and Aiven's Dmitry Kan at OpenSearchCon 26 © AlyxMacQueen")
Even SUSE’s personable, diplomatic CEO can’t evade questions about US politics, and to his credit, he tackles them head on. But what has all this got to do with enterprise choice and resilience?
Until this week, few would have branded SUSE a hotbed of insurrection, and not just innovation. But we live in strange times, and the conference heard that the demand for change is coming from a surprising source: enterprise customers.
Resilience and sovereignty are all the rage at SUSECON 26 in Prague this week. But against both markers, how does SUSE shape up?
OpenSearch Software Foundation Executive Director Bianca Lewis and Governing Board Chair Carl Meadows on why the new Long-Term Support accreditation program matters for enterprise AI infrastructure - and why the industry's obsession with vector search benchmarks is chasing the wrong metric.
_0.png89cb.webp "Carl Meadows speaking at OpenSearchCon Europe © AlyxMacQueen")
Three keynote speakers, one consistent argument. The hard part of enterprise AI in 2026 is not the model – it is getting the right data, with the right context, into the right place at the right time. OpenSearch thinks it has an answer.
.png9fc3.webp "Bianca Lewis at OpenSearchCon Europe")
Mark Collier briefed me on two updates under embargo at KubeCon Europe 2026 last month: Helion, which opens up GPU kernel programming to a far wider pool of developers, and Safetensors, which fixes a security problem in open source AI that was stubbornly overlooked for too long. The embargo's lifted - here we go.
.pnge0a3.webp "Mark Collier of PyTorch at KubeCon Europe © CNCF Events")
How many vendors at KubeCon Europe 2026 said they were not an AI company? I only found one. Peter Smails, General Manager of Cloud Native at SUSE, was fairly insistent about it – and what he said instead is worth looking into.
Every time a new chip ships and a CEO takes the stage to announce it, there is a question that does not get asked from the podium: can it actually run AI workloads? The answer, almost without exception, runs through PyTorch.
Four years after Log4Shell, 14% of affected artifacts are still being downloaded in vulnerable versions. OpenSSF's Chief Security Architect explains why AI agents are compounding the maintainer burden, and what enterprises should be doing about it right now.
.png5914.webp "Christopher Robinson AKA CRob at KubeCon Europe")
A roundtable discussion at KubeCon Europe revealed that the biggest change in observability is not technical, it's linguistic. The community is moving from describing what went wrong to predicting what the business should do next - and not everyone has caught up with the vocabulary.
_1.pngdc38.webp "CNCF Observability Roundtable © Alyx MacQueen")
Kubernetes formally archived one of its most widely deployed components on day one of KubeCon Europe 2026. Steering committee member Kat Cosgrove explains why the project's own flexibility became its fatal flaw - and why anyone still running it should be treating migration as an emergency.
A conference about prepping craft turned out to be about something more fundamental - the communities that make everything else possible.
.png28ed.webp "Sue Smith's slide from Monkigras © Monkchips - Bluesky")
AI is burying open source maintainers under a flood of automated security reports they don't have the time or tools to process. The Linux Foundation's $12.5 million coalition funding aims to fix that - but not in the way you might expect.
.png25fc.webp "Puppy riding on a unicorn © Michael Windser - LinkedIn")
The Linux Foundation's CAMARA project is exposing telecommunications network capabilities to AI agents through the Model Context Protocol (MCP), making network intelligence programmatically discoverable while defining the security and operational foundations for production deployment.
.pngcce6.webp "Cloud Computing Network Illustration © luthfil aziman effendi - Canva")
