Not an author but the answer seems to involve their espoused focus on ease-of-use and “it-just-works” levels of modularity. See their list of plans, which include a few LLMs and a variety of adjacent tools/features.

It’s a basic value prop and I’m way behind the times on local inference (so barely a few months lol), but it seems different enough that I’m excited to try it. Other tools have tried this path but I’ve seen none with a demo page quite this convincing…