Last week, Joao and I headed down to the port of Hamburg for ContainerDays—a few packed days filled with sharp talks and industry-leading platform engineers. Here is part one of our main takeaways from the event.
The increasing demand for Cloud sovereignty
Cloud sovereignty is moving higher on the priority list for more and more companies. Local, regional, and European cloud providers, as well as cloud-exit strategies and costs, were “hot” topics at this year’s conference.
We talked with the STACKIT and T Cloud Public teams , both of which provide cloud infrastructure in the DACH region, to understand what the main requirements are from companies wanting to move their data & services away from hyperscalers.
While some clients see cost savings as the main advantage, others are simply unable to fulfill their respective regulations using current hyperscaler guarantees. Regardless of the main motivation, we think that choosing the right architecture and vendor should not be done too hastily, and are currently working on making sure we can help our clients make the right choice moving forward.
Open source and its current hurdles
On the topic of open source, there were a bunch of new tools and projects unlocking new functionality or optimizing existing use cases. However, two talks stood out for us, giving good insights into the current challenges and development of the open-source landscape.
The first featured Christian Hüning & Niklas Voss from BWI, describing how members of the NeoNephos foundation are taking advantage of open-source components while still following the strategic objectives of IPCEI-CIS. In their examples, they use Kubernetes running on bare metal, managed with Gardener and relying on data diodes to serve risk-averse clients. It’s great to see how with proper design and implementation, air-tight companies are able to take advantage of open-source innovation
Tackling a different angle, Adrian Mouat from Chainguard talked about how open-source projects are very vulnerable to exploits, as their remediation time is longer and defense budgets are limited. Public PRs and a lack of dedicated resources make these projects easy targets for attackers who exploit vulnerabilities even before they get publicly acknowledged.
Besides reducing unneeded dependencies, getting rid of LLTs once and for all, and trying to use memory-safe languages, Adrian suggested actively defending and consistently evaluating our systems in sandboxes, even if we don’t have access to Mythos-level capabilities. Ending on a positive note, he brought up the Athena Coalition, which more and more companies are joining. The main goal is to use AI resources to find and fix vulnerabilities in critical open-source dependencies.
Human-centric development in the age of AI
Kelsey Hightower, the reason many of us love Kubernetes and platform engineering, delighted us with a long Q&A session where he touched on how much Kubernetes has grown, AI dependence, and his predictions for the upcoming years.
Kubernetes is declarative, which sets clear boundaries for its capabilities. These boundaries, according to Kelsey, are bound to not be sufficient for future platforms, but right now, “Kubernetes is the new infrastructure hammer, and everything looks like nails.” He pointed out that current developers are hesitant to work outside current industry standards, and that he’s surprised not many are trying to craft new tools instead of tweaking the hammer to no end.
When it comes to the state of AI today, Kelsey pointed out how developers spend more time documenting skills, which might lead to a universal query language used to communicate and even generate data across platforms. However, the rate of tokens being burned to fix repetitive tasks is out of hand, and Kelsey urges developers to use code generation and then set up cheap & deterministic flows with the given logic. One question addressed the concept of a “zero-token architecture”—his way of naming the practice of trying to understand and design a solution before using any AI tool, forcing critical thinking rather than taking the path of least resistance.
Last but not least, Kelsey shared his thoughts on the importance of learning new skills and putting in the work to avoid falling behind the ever-increasing competence level of agentic and AI development. “We now have senior engineers and junior people”—personal development (such as mentorship) is more important now, and hiring should focus more on skills that can’t be replicated by computers. We should all invest in and be excited about becoming better individuals, not just using better tools.
I agree with a lot of the points Kelsey made, and I do think that now, more than ever, we should make sure we’re focusing on the right skills and placing ourselves in the right professional environment to adjust to this new era of software and platform engineering.
