keeganpoppen

it is kinda mystifying bc from what i understand their engineering ethos is very much "if it's not working, just regenerate it" (which i completely understand).

sometimes compaction is a death knell and other times it's totally fine. it really just depends on whether the thing you were doing at the very end has high correlation with stuff from early on in the session, or if there was an obvious ~break s.t. you were just on some specific subtask. the situation has clearly gotten a lot better than even 6 months ago-- i think openai handles compaction better in my own experience. but it's exactly the kind of thing that yegge was trying to get at with gastown-- i used to run every session s.t. i could re-animate it every time the compaction boundary came... now, when the compaction boundary comes i am so pot committed that i just hold my breath.

it does feel like quite the coincidence that 2 seems like the magic number and we do, in fact, have two hemispheres. i doubt that the processing for any of these tasks falls strictly down the left vs right hemisphere, but i do kind of feel like our consciousness is kinda built on having a sort of internal tick-tock cycle, and we do certain kinds of thinking on tick and others on tock. but that is just armchair introspective dilettante neuroscience-- not to be confused with the actual thing.

i can't talk and listen at the same time to save my life, but as an intern at the Idaho State Historical Society a million years ago i did learn how to transcribe in roughly real time typing and listening. funny how those boundaries work-- feels like a "choose two of these three things" sort of... thing. i also don't find that i have the sort of internal narrative/voice in my head in the way that others do, so i would have to imagine that plays a role on some level. not that basically anyone with sufficient time can't learn to transcribe in near real-time-- it's not at all some sort of profound skill-- i was just struck by how quickly i could pick that up compared to other things.

yeah, doesn't the multitasking in this thread feel a little bit like what i imagine playing the drums is like? i've done some low-key drum-playing sorts of musical things in my life, and it was the disconnection point when you genuinely had 2-3 different things to focus on at once that i found my brain to be lacking. i observed the same with piano, but i had much more of a knack for piano, so it was only the rare song that genuinely required tracking three things at once... and those kinds of songs are hard enough for most mortals that people just pin them all as "dw those songs are super hard", when in reality it always felt like that specific skill was the thing that was lacking, and besides rote memorization i never found some way to lick it all the time. when i was really in flow, though, it would be completely effortless. it's bizarre.

yeah, cool: "sometimes when i feel hungry it makes just makes me more grateful for food"... like, seriously? come off it... what a fucking stunting-qua-realitas stance

oh man-- that seems peak, except for the part that my reading level is dynamic between "i wasnt actually paying attention" to "my brain is @ 5x and i am flying through the material"... very interesting idea, though-- kinda like a pace setter at a marathon or bike race or whatever. i endeavor to try this for myself.

no, clearly he is a "dumbass" disney villain that only IDIOTS listen to in even the SLIGHTEST! xD

wow, this was a fantastic, fascinating read

really? you cannot _fathom_ the idea of paying for a robot to do something you, yourself are already capable of doing? someone should tell all these car manufacturers and the like that their cost-benefit analyses of using robots for work humans can do are completely off!

this is the best distillation of what various models are like that i've ever heard... it's wild to me that people view LLMs as this monolithic entity, like "how do i get the best prompts to do <X>?", when it is such a clearly interactive medium, but the returns to engaging with the various models and understanding their "vibes" are very, very high.

it actually is not that impossible, and in fact can be done in a form factor not that different from Starlink satellites: https://www.youtube.com/watch?v=FlQYU3m1e80

they are right. bad output is user error. there, am i suiting the role appropriately? i do like 65% believe that, fwiw.

the output is definitely better. and i find it crazy how every time a new model comes out people trip over themselves to say how much worse it is than previous models, when in fact that is basically an impossibility. like, they've got the numbers, man-- you only release a new model when the numbers get gooder. the burden of proof is on the "didn't get better" side, not the "prove that it's better" side, because the architecture itself (1) only works because of how giant the training data / eval / etc. sets are and (2) has a fractal property of becoming strictly deeper and more thoughtful when you just click and drag the edge up and to the right (obviously AI research is harder than this, but that doesn't make the general point untrue). i say this especially because the scuttlebut is that this model genuinely is a shift-click-expand moreso than any sort of architectural "new science" or anything.

this is exactly why hypotheses come before the experiment in the scientific method.

this reads like "throw everything at the wall and see what sticks" reactionary-ism... i'm guessing that it's not particularly easy to use claude to help you make bioweapons, and we all know that they have neutered Fable vis à vis security research because people have already been complaining about it. and the funny thing about hate speech is that there is absolutely no need for ai-- it tends to come out the best when spoken directly "from the heart", as it were, anyway.

suuper high-value comment

whoa that game looks very cool— love it. also loved SC3k… the soundtrack was amazing. game was kinda lowkey hard, though. or maybe it was because i was just a kid haha.

so maybe we'll get to the right place by accident when all companies are effectively replaced by ai ha ha. (not putting a high %age likelihood on that one, obviously, just being cheeky)

well it’s especially not hard when you turn it into a completely different problem

i… lowkey agree with this, by the way. unpopular opinion though it may be/seem.

there is absolutely no line of demarcation between human reasoning and what you described

no, brah, humans are TOTALLY different. just don’t think about it too hard. we are just special.

why do people insist on claiming that they don’t reason, when they clearly, for all intents and purposes, do. you can be vague; you can express your idea a thousand different ways, and you will get a unique blend of <your input bits> x <hidden reasoning layer> => semi-smoothed output. this is like some Searle Chinese Room bullshit that needs to just die. it is beyond clear that llms can interact with abstract concepts in an extremely meaningful way. this is like the “thought leader” version of the stupid-ass “it’s just smart autocomplete” argument. if you think that, it is user error— either a failure of creativity or a failure of perception or both. just because llms are not a panacea and are problematic for society and “overhyped” and whatever does not make it intellectually honest to claim that there is zero reasoning/creativity/cognition within the box.

oh man my body is ready for any post-beads ideas… i will definitely check this out

i legitimately cannot divine what you are saying at all with this. there are so many dangling antecedents and modifiers that it is completely impossible. and i say this out of a genuine desire to understand what your argument is, knowing full well that i likely disagree with it.

i mean, maybe the golden calf people were right the whole time lol

as someone who is about as llm-forward as anyone out there, this is a brilliant analogy. was equally true of all the “prompt engineer” hype as well from a couple years ago (which i admit i still think does matter)… it kinda makes me feel like an audiophile / hi-fi person talking about how 24bit/192kHz is the one true encoding format and anything less is a willfull (cynical, “Quality”-hating, satisficerist, etc.) compromise. which i freely admit to being one of those people as well.

and in both cases i both “know” that i can tell the difference and “know that i cannot tell the difference”. what anyone takes from that in terms of what it says about me, personally, is a bit of a Rorschack test, but Astrology is about as apt a description as there is… xD

i'm pretty sure all that stuff works w/ Temporal... Temporal is extremely well-designed, in my experience. the js date object, on the other hand, has insane pitfalls, and i say this as someone who thinks not understanding JS ASI is a "skill issue", among other happily-un-"ergonomic" worldviews...

Does aperiodicity have any cool properties that help in specific domains?

well, it helps as the basis for all life on the planet as we know it (DNA, RNA, et al.). which is pithy, but i think actually has fairly deep implications (ie is not _just_ pithy).

the human touch from human curation of the ai output, which tends to be sorely lacking.