A few days ago, I was praising Orca and criticizing Claude for their implementation of projects. In the meantime I was testing different meta harnesses: T3 Code, BB, Conductor, and wasn’t very happy with the DX to be honest. I ended up preferring Orca. Then switched to Cursor and Grok Bot, where I do like the DX a lot. Then I went back to Claude after they implemented spawning new sessions, which resembles a lot what I liked about Orca orchestration abilities. Then switched back to Orca and I kinda missed Claude Code for Desktop.
It’s just funny how the battle for agentic software engineering is open. Every other week there’s something new to test. I just wish I had more time to test them all to have a good evidence-based opinion rather than my first impressions only.
My vote today goes for Claude Code for desktop, and for the combo Grok Bot and Cursor.
Tomorrow there will be a different story.