The most common workflow RN is:
1/ Make a planning step with a large context / smart model like Gemini, o3
2/ Implement code with a model that behaves in agent mode like Sonnet 3.5
Right now switching mode (and model) is 6 clicks both mode changes, which totally breaks the flow for me.
Please make it more programmer friendly @deanrie @truell20
Each mode should just keep it’s last model.I imagine one keypress to switch to a specific mode
e.g.:
Cmd-M Cmd-T - architect (preset Gemini 2.5 thinking)
Cmd-M Cmd-A - agent (preset Sonnet 3.5)