Con Claude Code, un bug me tomó 27 iteraciones para ser arreglado finalmente.
Sin un log de cada intento, el modelo repetía las mismas soluciones fallidas una y otra vez.
Escribí sobre esto y otros patrones que uso a diario 👇
goanpeca.co/es/blog/2026...
🚀 #Jupyter #MCP Server 0.21.0 is out! A single-PR release that speeds up tool discovery by adding caching — super handy for agents that love to use (or abuse) the get_tools API 😄
🔗 github.com/datalayer/ju...
Just returned from JupyterCon, where I had the opportunity to present on #Jupyter #MCP Server. It was fantastic to connect with the community and witness all the incredible innovation happening in the space. I’ll be sharing more insights and follow-ups from my talk in a blog post soon—stay tuned!
agent built on top if pydantic AI (migrated from langchain)
@goanpeca.bsky.social hello. welcome to bluesky
2025 is the year of the agents
Crazy how many hours it takes to get a decent dark theme. But hey, users love it—so we grind on. 🌑
Data analysis from our desktop app?
That’s exactly what our new Electron app is all about.
🧵⤵️
Have you ever spent more time dealing with CSVs, chasing versions, configuring environments than doing your actual research? 😅
We’ve built a Datalayer for biologists to focus on making progress, rather than configs’ issues. Here’s how it helps:
🧵⤵️
5/5
📖 Read the full post to find out more: datalayer.blog/2025/08/11/j...
1/5
🔔 We just launched a new update for Python dev working with Jupyter
Introducing jupyter-mimetypes - a new Python package that makes variable exchange between kernels and clients much (much) simpler.
🧵⤵️
Doing data analysis can feel tedious and messy at times. 😩
You’ve got tabs open, files scattered, and Python tutorials running in the background.
It’s chaotic.
But it doesn’t have to be.
🧵⤵️
The AI for Earthquake Response challenge is heating up 🔥 (organized by @esaearth.esa.int : platform.ai4eo.eu/ai-for-earth...).
Our Datalayer team is holding strong in 2nd place as Phase 1 wraps up! 🌎
