I want Opus to be only marginally better, but I do mostly research engineering and its ability to not fuck up my projects is absent. Every time my credits lapse I let kimi and composer2.5 have some play and it’s basically just an excuse for me to keep playing computer because when the oai/ant credits refresh I always need to spend hours recovering from the other models either misconceptions or boneheaded eng practices. Even when I only let it touch my web games…
separately trained experts can surpass performance in their activated regime and DOES result in a smarter model, the Claude system cards talk about this and eg there is https://openreview.net/forum?id=iydmH9boLb to read...
I was somewhat surprised to learn that phi is _merely_ (1 + √5)/2, I didn't have a good conception of what it was at all but I didn't think it was algebraic.
What's stopping you at pasting only a single file? I use the workflow Elon suggests (although I've never used it with Grok) predominately, it's well over 30% of my use of LLMs. I have a small piece of python called "crawlxml" that filters + dumps into <file> tags. And of course the LLM doesn't need your actual code in its context to do its job.
Whoa there boss, extremely tough for you to casually assume that there is a consistent or complete metascience / metaphysics / metamathematics happening in human realm, but then model it with these impoverished machines that have no metatheoretic access.
This is really sloppy work, I'd encourage you to look deeper into how (eg) HOL models "theories" (roughly corresponding to your idea of "frame") and how they can evolve. There is a HOL-in-HOL autoformalization. This provides a sound basis for considering models of science.
Noncomputability is available in the form of Hilbert's choice, or you can add axioms yourself to capture what notion you think is incomputable.
Basically I don't accept that humans _do_ in fact do a frame jump as loosely gestured at, and I think a more careful modeling of what the hell you mean by that will dissolve the confusion.
Of course I accept that humans are subject to the Goedelian curse, and we are often incoherent, and we're never quite surely when we can stop collecting evidence or updating models based on observation. We are computational.
i’ve vaped about 3g of DMT in my time, cumulatively. spotify (and other computer UI) looks very nice under the effects. it can supercharge spatial reasoning/visualization powers for a short time. i mostly haven’t met “entities” but it has happened a notable few times. once, an elf told me to “knock it off” after an afternoon of several trips in a row.