Sounds like it’d be relatively easy to get a loan or investor with that kind of ROI. Seven years is nothing if it’s consistent and safe.

Is there a reason you’d have to go all-in, rather than starting with just a couple dozen panels first?

All at once, absolutely, but it’s going to take a LOT of most things.

Theoretically, anything is poisonous if you have enough of it. We even get oxygen poisoning by breathing air that is too rich in oxygen for too long.

Don’t support nazis or the people that support nazis.

Ah, I’ve only played with the second-smallest one - I guess I should have expected that the RAM req has to scale up for larger models

The online, UI version of deepseek wraps your prompt in their own prompt, just like every other online AI service that exists. Deepseek is unique in that it can be run locally, as well - and is open source - and that doesn’t wrap your prompt in any way.

Locally, you can use a model as small as 9GB(or as big as ~400GB for better responses), it only requires 32GB of RAM (much more for the larger models), and never makes a single request out to the internet.

Interesting how you completely avoid the question 🤔

____pilled is a mindset. Black is depressed, given up, etc.