eps

interesting

Hc

I fucking hate tech bro clowns pretending to be journalists like this.

Reads real journalists coverage, claims it isn't being covered, condenses it and gets it wrong, but you can subscribe to his app $9 a month!

He is a fucking grifter and an idiot - people love him.

thank god this unadulterated genius will be able to stop writing javascript/typescript syntax into an IDE and instead write english syntax into a web browser.

this was worth the economy and the environment and the misinformation and the value of labor and the firings and the suicides and the

I'd recommend it! I switched in Feb of 2025 and it's been great. Please feel free to reach out if you'd like any advice or help in regards to switching or the usual 'how gamedev on linux' conundrums

so no i do not particularly have much hope in projects that will refuse PRs because they also dared to change "he" to "they" in a localisation file

to a degree, i would agree. but unfortunately the founder has a history of denying PRs over something as trivial as "using gender neutral language" on the basis of "advertising personal politics" so apparently, non-males using computers is political :)

for context: SerenityOS PR 6184 on github

it's honestly a bummer, i admired the guy and his work on Serenity but no he just had to decide to be a piece of shit for some reason. i just don't get it

unfortunately iirc ladybird's lead is one of those "le epic based anti-woke anti-dei" crusaders right alongside DHH and similar scum of the earth who believe in the "white replacement" theory :/

so we've got servo at least! that's something :)

hi, the guardian.

no.

big fan of size difference and a frantic “it can’t fit” and a broken “you’re gonna break me” and their tummy bulging when they’re only halfway through and a slurred “s too deep” as the other so softly keeps cooing praises and “you can take it” as they wipe their tears and

heartwarming: company only crushes 17 toddlers instead of 20, hailed as heroes of our generation

Bisexual lighting, anyone?
#blender #3dart #3drender #furryart

In 2000 years, after culture’s long slowdown, when the pre-AI-slop era gets mythologized as the last “real” golden age, people will be quoting Top Gear the way we quote the Romans today.

can confirm! from my recent haul of ~200-ish disks, only one is dark grey!

needs more felicia fun

jarquis i need the steam phone immediatedly

and i do mean "right before switching" the straw that broke the camel's back was that i was testing explorer to see if they fixed it. while it took it's >60 seconds to launch, copilot popped up. :|

i was running arch with kde plasma by end of day. i am now a certified kde whore love me plasma

since switching to linux the biggest issue i've had with my computer(s) was text on icons not having shadows on my laptop. that's literally it.

For contrast, right before switching in Feb 2025 I was waiting for msft to fix the file explorer taking >60s to launch. they gave me copilot instead. bruh

Hi,

no.

best regards

yes for this we've relied on humans for the better part of over half a century for a very good reason

hello again, yes I did say that they are borderline useless even after absorbing every word ever written. And yes, you are correct, even with current LLMs with access to that data, it's by far not worth the energy, environmental and societal costs. As you said, "problem solved" :)

hi. this data could be useful for teaching smaller models (SLMs, instead of LLMs) but unfortunately as we know from early attempts at GPTs, they are genuinely hopeless without a genuinely massive, wide spanning dataset.

Ultimately the one way I find comfort in the "only synthetic data" route is that even right now synthetic data is resulting in massive inbreeding and compounding error. The technology needs to rely on fresh human data or else it'll make itself less effective over time.

it's one of those things that is kind of impossible to explain to someone until it affects them, sadly. No amount of explaining inaccuracies, or societal or environmental impacts will convince many in our world of billionaire pedo-algo induced apathy. Can convince a few but, drops in a bucket :/

honestly you're right in the fact that most people just don't give a shit and this is the path the story will take. personally i still find it heartbreaking that we destroyed what little credibility the internet had over less than negligible benefits, so my soul demands i must rebel against it :P

in my eyes, LLMs have become the asbestos of the tech world, and crying to carve out new reasons to keep using it for it's incredibly negligible benefits at hilariously high costs is sadly just not the way forward. whether we can rely on humanity as a whole to not follow this fool's errand... ehh...

personally I've been on somewhat of a war path trying to find any genuinely useful avenue for the technology, from dataset summarisation, to inferring intent from voice commands for NLP, and so on. ultimately every single example ends with the same answer: we've already done this better without LLMs

imagining the logistics of asking for permission to use data at that scale, let alone paying people for it, sounds like an impossibly sisyphean task simply not worth burning the hundreds of millions needed to create the base model, and then burning more to run it, so it can have a 60% failure rate

hi f4mi, the main problem here overall is that LLMs are effectively impossible to create without stealing data. as a technology they're borderline useless even after ingesting every single word ever written and even words that weren't written