The list of assistants (dropdown) has become overwhelming, at least for Ultimate users (40 models). Please add some UX guidance for casual users. Ideas:
- Hide most models behind a "more..." sub-menu.
- Or add a "default assistance" at the top, like there already is a default "code" assistant, so the user can switch back to default up-to-date model without doing research first.
- Or add a star to a few models (say, three at most) that are good default choices.
The long list creates decision paralysis for me, or even FOMO. Is my trusted custom assistant obsolete by now? Probably. Which of the 40 others should I try instead? I can't keep up with the speed of LLM news and model naming.
Most of the time, I start with my old trusted custom assistant. I built it with Claude 3.7 Sonnet with a prompt like "answer the question, then add a new perspective". I like that I can keep using it.
The problem usually comes when it doesn't provide a good answer, and I would like to quickly re-try with the default up-to-date agent. But if there is such a default, I don't see a way to select it (as there is for "Code").
Also, what model should I upgrade my old custom assistant to? I mean, eventually I re-discovered the llm-benchmark (thanks - great work, keep it up!). But something less research-y would have helped, like "usually one of those two or three assistants work well".
