Qovery Skills give your AI agent real deployment superpowers. Deploy, optimize, and troubleshoot on Kubernetes (and beyond) directly from Claude Code, Codex, Opencode, and 25+ other clients. See it in action: www.qovery.com/blog/qovery...
I’m in NYC this week - happy to catch up on Friday. I have a few open slots in the afternoon to meet CTOs and CEOs to talk about AI and infrastructure in 2026
Claude Design just made Webflow and Framer feel obsolete. Wild time to be building.
This is our last month on Webflow.
With Claude Design and Claude Code, every marketer gets a full team of seasoned designers and engineers - on demand - to build landing pages fast. No more back-and-forth. No more waiting for your idea to catch up to the build.
We are living crazy times. Qovery users: happy to share our SKILL and get your feedback. Preview is available right now.
3. Deploy (~15 min): I used Qovery to deploy it straight from my opencode (we’ve got a SKILL for that ;)). It took ~15 minutes to fix the Dockerfile, but after that it was fully automated. I just had to wait :)
2. Code (~45 min): I picked my design from Claude Design and asked Opus 4.6 to code it. With a few adjustments, it took ~45 minutes.
1. Design (~1h): Claude Design
From Design to Live in Prod: 2h of work! Here is my stack:
Claude Design is exactly what I needed in the AI era. No more wasted time picking the right pencils. I just explain what I have in mind, and everything is handled by Opus 4.7.
What a crazy time to live for every builder. It’s super exciting.
I’m very visual, and I’ve always liked drafting my ideas - first with Microsoft Visio early in my career, then draw(dot)io, then Excalidraw, then tldraw…
Claude Design is, for me, the killer product launched in 2026. As a CEO of a product company, I need to make my ideas crystal clear for my team.
We’re looking for 2 Enterprise AEs and one Alliance Manager. Very competitive compensation package. Jobs.qovery.com
There’s a lot of FOMO around AI - the challenge for entrepreneurs is removing the noise from the signal. Focus on your customers and you’ll gain clarity. Use AI (or not) to solve real problems that matter to them, and you’ll rise above the noise.
Happy Friday.
Excited to have officially kicked off our partnership with Devoteam.
Devoteam and Qovery partner to accelerate Enterprise Kubernetes adoption with an AI Platform mindset.
That’s just the beginning. Let’s go teams!
You can stack up to 5 Mac Studios, with RDMA (Remote Direct Memory Access), and scale up to 2.5 TB of RAM - a monster for running best-in-class open-source models. Of course, there’s a huge price tag (~$50k).
Clustering Mac Studios with Thunderbolt 5 - look at the perf. It’s insane.
Let’s meet the Qovery team at AWS summit Paris Today! Qovery is a proud partner of AWS. Let’s go!
Tomorrow is AWS Summit Paris - are you going? Let’s meet my team and happy to meet in person 🤝🏼
Take a look at how our MCP server gives superpowers to Kubernetes and can save you tons of time on incident management. proof in the article. enjoy the ride, it’s crazy 🔥
www.qovery.com/blog/mcp-se...
Impressive: a 397B model (Qwen3.5-397B-A17B) that runs with 48GB RAM.
github.com/danveloper/...
Running powerful LLM models locally is one step closer to reality.
The future of AI is local models - stop spreading your data across the world to providers that will monetize it on your behalf once they need to increase their income and cut costs.
Today, 90% of what I was doing with the Anthropic and OpenAI APIs has been replaced by local models. For the remaining 10%, I use Anthropic Opus 4.6.
That’s the best combination - and a huge $ saving.
I use llama4:16x17b for translation, OCR, and text improvement.
I use qwen3-coder-next to automate my work and as a coding assistant with opencode.
I use qwen3.5:122b for solving complex problems.
Current experimental results suggest that while LLMs show basic compiler-like capabilities, they still achieve low compilation success rates without intervention.
Exciting times!!
Compiler technology has been extensively studied since the mid-1950s, creating frameworks that transform source code into target code. The LLM-driven approach is a novel direction that aims to address challenges in traditional design.
We are reaching a point where LLMs may let us bypass traditional intermediate source languages and generate assembly directly from human language - though practical reliability still depends on prompt optimization and better reasoning methods.
Maybe the future of programming is Assembly?
AWS Summit Paris is next week. My team will be there - what about you? Let’s meet in person, and why not grab a drink? DM me.