I am using dirac with Kimi 2.6 for refactoring a rust codebase. I have a Clean Architecture design which is being reinforced. The scope of work is laid out in a Beads epic with sub-issues. The planning was done with gpt5.5, and gpt5.5 is checking the work is complete. I have found that dirac is more productive on large codebase refactoring than OpenCode which actually trashed the .rs file and had to revert the code.

For gpt-5.5, do you use Dirac, too?