Kubecon North America 2025 - Crisci Solutions

The Kubecon in Atlanta is now done. I had a good time attending the conference and meeting a few people among the speakers and people from selected booths.

I attended the conference without a specific problem in mind to solve so I visited very few booths and only ones that I had a specific interest in. I was mostly there for the talks ( AI and supply chain in particular ) and the overall experience. I didn’t meet as enough people as I would have hoped to but it was still a good experience to attend the conference. It felt a bit emptier than the previous Kubecon London conference which I attended earlier this year.

AI was omnipresent

A few things come to mind after this experience. AI/ML has been dominating the show and new features such as DRA have been heavily mentioned. My understanding is that it allows you to share GPUs resources on a node through special device plugin such as the NVIDIA DRA driver. It is also used to enhance Kubernetes networking capabilities with projects from Google such as dranet.

It also became apparent that I have a lot to catch up in terms of AI/ML solutions. I have yet to find a customer who would need to implement a Kubernetes stack that requires the level of tuning and sophistications that were demoed in those talks but I always welcome challenging and technical content. I feel like 2026 will bring me such projects to work on and as a contractor, I have to be aware of how the sausage is made and what the options are for my customers to ultimately deliver with my help.

I definitely got my money worth of technologies and details on what to study and skill up on next. There is so much to learn with this new AI/ML ecosystem, it’s exciting. I can envision new challenges ahead and customers in need of my skills and that’s what makes everything that I do so rewarding.

Supply chain security is gaining momentum

I was also happy to see how supply chain security is picking up with various Keynotes and talks on the subject. Having spent the last 2 years working on a project with software attestations using the in-toto Attestation Framework, I was not surprised. It is also very important these days as the supply chain becomes ever more complex with many moving parts that need to be audited and attested. Your current stack is as strong as its weakest link and that weakest link could be a simple minimal transient dependency library that you are not even aware of.

Attestations are such a great way to provide evidence of security controls being met within your organisation and a self-service / decentralised way to capture the evidences and have them travel alongside your container images. They can be stored within an OCI artifacts inside a registry alongside the image. New features of the OCI Image and Distribution specs v1.1 make it easy to discover OCI artifacts attached to a container image via the referrer api.

I was hoping to meet the in-toto people as they celebrated their 10th birthday and recently graduated within the CNCF but unfortunately I didn’t manage to do so. Once again, a great piece of software / logic and they are now busy working https://github.com/in-toto/attestation-verifier which would provide a way to standardise a verification process of attestations. Have a look at this example layout.

Testing distributed systems in a simulation

I also loved meeting people from Antithesis. The way that they setup a simulated environment to test distributed systems like etcd and other databases sounds interesting. They also allow you to not only trigger an unexpected fault but to also rewind on the execution and capture the exact context / environment that triggered the fault. Once you have fixed the issue, you can replay your software tests within that specific context and ensure that the issue has been fixed. That sounds quite valuable where the alternative is to increase your logs to DEBUG / TRACE and pray that you trigger that issue again in your tests environment or , worse, in production ( with all the performances impact and potential issues that running your software in such a state give you ). My gut feel tells me that this is big. I first ran into them watching this video on Youtube.

I love these kind of talks because a lot of this stuff I can understand but some things like TLA+, formal methods, property-based testing, metastable failure condition, I had to research and understand which is always great. You are learning that there are ways to deal with the inherent chaos of running your app in production. I had a discussion with Marek who is the SIG etcd lead and he was very keen on the solution which he used himself with etcd.

They also organised a conference recently around testing and other topics. You can find the videos on their Youtube channel.

WASM and Kubernetes

I also had a discussion with the people from Spin on WASM and I definitely need to keep an eye on this whole ecosystem as it might prove useful for certain use cases ( mostly sandboxed execution of portable, lightweight and untrusted code ). They have a cliand also a project to deploy WASM workloads on Kubernetes called spinkube.

Apparently Helm v4 also supports WASM for its new plugin system.

LLMs and your internal documentation

Finally I bumped into people from Atolio. A lovely bunch with an interesting product. They basically allow you to aggregate data from various internal sources and have a way to then query the internal AI from a unique place instead of spending hours / days to search for the key documentation / person holding that piece of data that you need to onboard / progress on your deliveries.

This is fantastic and certainly what LLMs are for. As a contractor, I often move from business to business and the amount of time spent onboarding, finding the right person to talk to to get more information on x topic, making sense of what is current vs outdated documentation, browsing through different sources of informations to find anything related to x topic, is time that is not productive and is better spent on actually delivering on the Statement Of Work’s deliverables. As much Information as possible should be available from day 1 to allow me to be productive as soon as I start. If Atolio can even solve 50% of what I have described, I’m already sold on it.

Platform engineering

The main thing that caught my attention was the mention of a marketplace within the Thursday keynote. Instead of providing all the features / components of your platform as an engineer, you should let your tenants / third parties plug in their own features within a marketplace. A bit like the AWS or GitHub marketplaces. Every single successful platform usually has a strong marketplace for its tenants.

News and Talks

Things to mention

Talks that I would advise you to watch

I have attended a few talks and I would advise you to watch the following:

Technologies on my radar

What I consider essential to know to remain competitive in this market:

AI

Supply Chain Security

Misc

Visiting Atlanta

I would also recommend a visit to Atlanta. I thought that the people were friendly and courteous. Varuni Napoli and the Fox bros BBQ have to be visited! The Hyatt House where I stayed is also a really nice central place with a really helpful staff.

Contact us

If you need someone to help you make sense of some of the technologies above, feel free to contact us.