I legitimately think @elonmusk has read both the celestine prophecy by James Redfield, and Accelerando by Charles Stross (both great btw) - and is trying to haze us to some kinda weird blend between them.
@koshercockney@dsisme48 I spent hours debating tuckers positions (both for and against) on many things over the years, and am fairly split on a lot of his views, but it was a genuine shock to see him so strongly condemn Israel.
@koshercockney These two sides can't agree on shit normally, you've lost your marbles if you think both sides are working from the same book intentionally. Even in my neck of woods our extreme lefts and rights are starting to have unified views, many chronically offline. השתקפות עצמית maybe.
@bridgebench would be interesting to see what providers your using for glm 5.1 - not all providers are equal, as most devs know. If we are going to compare benchmarks, we should do it across providers as well IMO. also having the prompt used to generate these would be good.
@elonmusk@elonmusk It's all coming together, but the important question for now is - Is it possible to offer flat rate thinking to consumers long term? I feel like the better the models get, the less token pricing is going to be bearable, it's already stretched.
@dtaspire@thdxr@cmuratori Thaler v. Perlmutter (The "Creativity Machine" Case) seems to say, even if a human is involved, it's involvement must be substantial for the output to be copyrightable. Unless the supreme court weighs in, that's pretty much the standard i'd say.
@elonmusk Wasn't most of Claude Code was written by AI? Won't they be SOL - AI generated "stuff" can't be copyrighted right?. I Imagine that's a big reason @elonmusk's focus is hardware and models. Software's value is/has been heading to zero for some time. It's trust that matters now.
@adamdotdev the amusing thing here, is that anything created by ai founder x, can and probably will, be duplicated clean room style by founder y. the value isn't the software anymore, it's the trust associations.
@norpadon I've noticed that quite often those LOC are adding/removing similar functionality because they're changing a parameter, or adding a field. LOC is a terrible metric to measure by IMO.