Yeah this makes a lot of sense. I don't know why LLM development all seems to have been "one big model that knows everything" rather than more specialized. The former approach just means more wasteful energy use...

I’ve been using Minimax for simple tasks with my claw bot, and opus for advanced coding or building.

small models with external memory and tool usage

I think the goal was to make a model that could keep learning and improving. It's not panning out that way, so with energy constraints we are learning small models make more sense now. I do think there might be a place for large model to discover connections we might not have seen/noticed.

Figures, a Scorpio would say that

Exactly what I've been thinking. Some of the providers in my hospital are using a limited access model that trains on approved data and only gets used by physicians. Apparently it's been pretty good. Still uses a lot of energy, but my guess is a lot less once you compare to the fuckery one would dea...

The only astrology knowledge your tinder agent should need is to stay the fuck away from anyone seriously mentioning it 😆

Also cheaper, which makes sense because these large models are not self-sustainable in my opinion. By the way, what way did you find the most reliable to orchestrate them?

What if your coding an astrology app?