I have a different perspective: in my opinion, the main difference is "Claude Code", the harness, not Claude Opus itself. Earlier on I used the same Claude AI in both copilot and in Claude Code, and the harness made all the difference. And for me, looking at code is mostly a waste of time now, it just makes me less efficient for the most part, I find it faster to just ask questions about the code. Claude Code will still occassionally show me fragments of code, of course.
RE: Release of new HAF API stack 1.28.6 next week