I have been using chatgpt a lot and they actually have an option where you can provide “custom instructions” that are injected into each new conversation. It looks like this:
With custom instructions I can pre-train the AI to respond how I want it to. This allows circumventing a lot of the boilerplate that common AI models always use, and makes it a lot more helpful. Typing full sentences re-instructing it on each new page, on each new conversation, becomes prohibitive very quickly if you’re not a user who’s perfectly content with Leo’s out-of-the-box behavior. ChatGPT has become 10,000 times more useful once I got custom instructions where I wanted them to be, and Leo could be as well. Ideally these instructions would be very overriding, though I don’t know how much control the Brave team has over the models.
Probably just this would be sufficient for me to switch over to using Leo instead, since chatgpt cannot interface with a page in realtime, but Leo can. It may be more of a power-user feature but I think using Leo at all is probably something more technical people would be the ones doing, so I think its worth the investment.