I‘ve gotten a lot done today, made a huge amount of progress on this React front end I've been developing. I am basically down to the last few features. And now I‘ve gotten this new backlog based workflow working really well with Gemini. I have all the features planned out in quite a lot of detail. It occurred to me earlier that why wouldn‘t I be able to turn on Gemini Agent mode, so it could run commands and query MCP servers etc, and see how far Gemini could get implementing the entire feature. That would be kind of wild.
Thinking about it some, and after doing a bunch of research online, there are quite a few things that still need figuring out. The biggest for me is what‘s the safest environment to configure. Obviously some things could turn out to be very bad if say it had full access to the Github repo. But perhaps there is an easy way to set things up so it does everything locally and then I check it all before pushing to remote and merging in the code.
Hmmm, it‘s very tempting. And they even have a setting called Yolomode, where it no longer needs to ask you for permission to do things, which is both amusing and a bit worrying at the same time. #