certainly helps to provide the right spec to an coding agent 🤦

Missing Claude Code (there’s a pr), pretty interesting

First time I came close to hitting the limits of Claude Code. In a nutshell: Wasn't really sure which direction I wanted to go, fell into vibe coding mode by mostly winging it, which led to huge changes without having feedback loops in place (other than myself).

That's pretty interesting - via @tuhoanggggg.bsky.social #mcp

Was slightly confused about Claude Code coming up with weird takes for a problem I'm working on - Turns out I'm to blame: asked it to build a transparent proxy, while it actually had to build a protocol mapping due to the wrong testing endpoint I provided :D

Time for a break, I guess #claude

That seems to be roughly the fair use limit for Opus 4 on the Max Plan - but not sure what the exact time window for the rate limiting is. Back to Sonnet 4 for a couple of hours I guess

That's an interesting MCP idea for agentic coding > A comprehensive bridge between Language Server Protocol (LSP) servers and Model Context Protocol (MCP) that provides Claude Code with advanced semantic code understanding capabilities.

Yep, this 💯

Anyone remembers Rails Rumble? Something like that, but with coding agents 🤔

1 hour to kill - what am I going to build? :D

The otel data from ClaudeCode is super interesting to track. Have a little app up and running, which provides ingestion endpoints and a dashboard like this. Anyone interested?

Had so much fun, I’ll do the same talk again tonight at @hamburgonruby.bsky.social

At the JS meetup in Hamburg - gonna talk about agentic coding #claudecode