coding.berlin | KERNSTACK

A European Cloud Needs a Foundation Innovation in Europe is happening - but it’s happening on the backs of hyperscalers. European startups, enterprises, and public institutions build ambitious products, only to run them on infrastructure they do not own, control, or meaningfully influence. We innovate locally, but the economic gravity pulls everything outward. The result is simple: innovation is not paying off for us. We are missing something fundamental. Europe knows how to host websites. We know how to run virtual machines. We know how to build and operate world-class data centers. But modern cloud architectures are not built from isolated parts. They are built on integrated service stacks - deeply interconnected systems for compute, storage, data, networking, and execution, designed together from the ground up. That is the real advantage of hyperscalers. Not scale alone, but tight integration. Without it, everything becomes harder: Developers glue services together instead of building products. Capital flows away from the place where value is created. Today, there are effectively zero real alternatives. KERNSTACK exists because that is not acceptable. We believe Europe needs its own foundational cloud layer - not a copy of hyperscalers, and not a collection of loosely coupled services, but a cohesive, modern platform designed from first principles around how software is built an run today A platform where infrastructure is not an obstacle. Where primitives compose naturally. Where innovation compounds locally instead of leaking away. This is not about sovereignty as a slogan. It’s about economic gravity, technical independence, and long-term leverage. KERNSTACK is our answer to the missing foundation.

@sundeep @steipete Consuming tokens has become such an integral part of work that consuming too little is just hard to justify x.com/codingber/stat…

“If your $500K engineer isn’t burning at least $250K in tokens, something is wrong.”

@mntruell Love it! 🫶 Flexibility and fast adaptability is a great edge think .. no need to be „pure“ in anything at this stage with this rate of change in the industry 💪

Composer 2 is out! Cursor is an example of a new type of company, not a pure app maker and not a model provider. Our aim is to build the most useful coding agents by combining the best API models and our domain-specific models.

@melvynx I‘m surprised about how it actually generates nicer UIs, restyling some elements rn 😅

Just try Composer 2: → for new apps, it makes WAY cleaner and smarter code than Opus and Codex → makes reducers, respects clean code → respects the UI and features I ask for perfectly It's a good model. Really.

@ryolu_ Look forward to seeing it sneak into my token distribution going forward. Going hard on testing it rn and loving it so far💪 x.com/codingber/stat…

Frontier intelligence. Fraction of cost. Fast. Composer 2.

@mark_k Never left … on a streak here x.com/codingber/stat…

Are you going to give Cursor another chance, now that Composer 2 is here?

@benln On it … 💪 having loads of fun and great results with it so far 🙌

You can now build with Composer 2

@cursor_ai Great model! Seems to be fitting my workflow just right so far 👍 It seems to have a much better sense for aesthetics in UI as well.. far less issues with weird margins or font style creep and similar 🙂💪

Composer 2 is now available in Cursor.

@kristaletz Getting some impressive results so far .. super fast, accurate and concise in summarizing the changes 🫶 Will definitely stick with it for a while 👍

Composer 2 is available now in Cursor. Frontier-level intelligence, faster, and more cost-effective than comparable models. cursor.com/blog/composer-2

On a streak with @cursor_ai lately 💪

@broodsugar I heard that little guy kicks humans, so be careful around him 😁👍 looks really cool though 🙌

Our lab is cooler than yours.

@RoundtableSpace kernstack.com

What are you currently building?

@kristaletz cursor is throwing exciting things at me every day.. anyways I’m ready for it 🙌

exciting things at cursor coming soon 🎻

@NVIDIAAIDev @karpathy @DellTech I need one of these 🙌

🙌 Andrej Karpathy’s lab has received the first DGX Station GB300 -- a Dell Pro Max with GB300. 💚 We can't wait to see what you’ll create @karpathy! 🔗 #dgx-station" target="_blank" rel="nofollow noopener">blogs.nvidia.com/blog/gtc-2026-… @DellTech

@AstasiaMyers less talking, more building. Love it 👌 We can plan a meeting to talk about how to get started some day .. or we can just show off a first version of the product on day 1 and take it from there

One of the biggest shifts in product development I’ve seen over the past year: Don’t write a memo. Just build it ‼️ • Don’t write docs → ship prototypes • Don’t debate one idea → build 5 and test them • Don’t ask users → watch what they actually 👉 Emphasize discovery

@morganlinton Big fan here as well .. and of AutoML to create them with little to no expertise in actual training

Insanely bullish on small, special purpose models.

@andrewchen Yes! And the opportunity cost of not sitting at the computer to hit run on the next command is soo hurtful 🤯😁

AI is supposed to save me time, but now I find myself building stuff all evening and weekend and it's actually increasing my time in front of the computer WTF

No pipeline. Just data -> searchable instantly

@GoogleCloudTech this type of clear structure really gets me excited - so many pieces clicking into place after billions of tokens of „raw“ experimentation 💪

1B tokens deep already and it's only mid March 🤯

@AstasiaMyers supply shock incoming 😩

CPUs are cool again! For years: GPUs = AI Now: AI = systems problem Agentic AI + inference-heavy workloads are driving a real CPU resurgence handling orchestration, memory, and coordination that GPUs don’t. GPUs think. CPUs run the show.