You're more than welcome to link these studies!

I would say that I had my concerns before Opus 4.5 and my Claude Code usage was on the decline as I only found it to be suitable for very small isolated tasks. I started to think that agentic coding might be further away than we might have thought. The agents needed so much babysitting. And then Opus 4.5 launched on Nov 24th and it has changed the game for me at least. I gave it a try again and so far (about a month in, I didn't immediately jump onboard) I haven't caught it doing anything "weird" or "wonky" like previous models do.

It's still not the 100x productivity boost all AI companies are talking about though. We are still very much limited by things like code reviews and how people need to understand the changes before they are merged.

And we are what... 3 years into this era? People are still figuring out how to work with agents. You might still be skeptic but the speed which we are advancing at should be worrying to anyone writing code for a living.