These days 8BitDo controllers are superior to first party Xbox controllers and sold for cheaper.

The updated fluid mechanics are a lot more forgiving and basically have infinite throughput. It’s still a whole new layer of complexity but doesn’t have nearly as many confusing limitations as it used to.

Bravely running away is the quintessential FromSoft experience. The ultimate flex on enemies is to not even bother attacking them and just rolling to dodge occasionally while you grab items and run past them to the next checkpoint.

What I haven’t figured out is this…

If we’re all going to LLMs instead of asking each other for help (or providing help to others), then how do the models learn new things? Aren’t we no longer generating the same volume of consumable data?

I suppose we can provide feedback to the models to tell them if their solution worked, but I can’t tell if that sort of feedback is more or less useful than just crawling forums.