Just spiking a Terraform -> Pulumi migration - with Claude doing all the heavy lifting, it's actually almost a no-brainer. Also, with state and plans, there's a pretty effective feedback loop for Claude

Luma hides the receipt / invoice really well, so we had a bunch of requests for this already.

Claude Cowork is surprisingly useful - Was just finding relevant people from a company: "<company url> Find people which are based in Europe and might be a good fit to speak at agentic.hamburg/conf-2026/ " Which actually surfaced the right people - and no, there was no location info on that page.

New alias for nanobanana cli

TIL that OTRS exists and I'm now retroactively grateful for every single piece of software I've ever complained about.

A PM threw some design mockups over the fence. Instead of manually iterating, I set up Claude Teams (mockup-artist, ux-designer) with access to the Nanobanana CLI to riff on them.

Had overall good results with Claude Code teams - but haven't seen them actually "communicate" between each other. More like reporting back to the orchestrator and then continuing with the next one. Am I holding it wrong?

Been reviewing a few companies' tech stacks lately. A pattern keeps showing up: massive microservice complexity around a core product that's honestly a weekend build now. Most of the code is boilerplate. This is exactly why agentic coding is about to reshape entire engineering orgs.

> "Mystery solved. That was all you, Sebastian! Here's the timeline (all UTC, +1h for CET)" #claude

🤯 www.anthropic.com/engineering/...

“This is our single point of failure Nginx fronting our highly available Traefik ingress to our Nomad cluster” Poetry.

`claude --from-pr` is pretty helpful 🎉

These ice blocks near Hamburg in Geesthacht made it to major news in Germany – so it was quite crowded even at -5°C and strong wind 🧊❄️

delivering emails from a new account is the worst