FakeAndres @andres.rive.ro · 26/09/2026music.youtube.comLosing My MindYouTube video by Police In Paris - Topic 000
FakeAndres @andres.rive.ro · 26/09/2026The billionaire/PE/VC groups are trying to downplay Anthropic’s IPO, make it seem like it’s at risk or going to tank, even get frozen by the SEC/USA. Hurriedly promoting their only competition in all the comms they can control. That can only mean one thing: it’s gonna be the biggest IPO of all time 000
FakeAndres @andres.rive.ro · 24/09/2026music.youtube.comM83 - My Tears Are Becoming a Sea (Summit & RYX Emotional Schranz Edit)YouTube video by NOS - Not only SoundCloud 000
FakeAndres @andres.rive.ro · 24/09/2026Have you tried the full 2TB version on private hardware not via zai cloud or open router, but direct 1-to-1? I was gonna make an account to run it on a 8 GPU VM to try it out, comes out to $30/hour. Figure I’d test it for 1 or 2. See the difference between what’s served via shared cloud services 100
FakeAndres @andres.rive.ro · 23/09/2026music.youtube.comHustlin'YouTube video by Rick Ross - Topic 000
FakeAndres @andres.rive.ro · 20/09/2026I like to imagine this is what the GPU processors feel like when they get SLI or even a NVLinkstatic.klipy.comCaptain Planet: Let Our Powers CombineALT: Captain Planet: Let Our Powers Combine 110
FakeAndres @andres.rive.ro · 20/09/2026They need an new emoji with a 🤖 face behind the laptop: 👨🏽💻🧑💻👩🏾💻👩💻👨💻 000
FakeAndres @andres.rive.ro · 20/09/2026SREs can now become ARSEs: Agentic Reliability Site Engineers! 010
FakeAndres @andres.rive.ro · 19/09/2026its just slop (derogatory) ive never made it through a complete video when it comes up on my feed 000
FakeAndres @andres.rive.ro · 17/09/2026I can’t count to a billion cause I’d fall asleep by the time I got to 50,000. I feel like it’s a similar idea. I know it’s impressive what it can do it, but… 000
FakeAndres @andres.rive.ro · 17/09/2026Took me like 20 minutes to make a classification and rules engine for profiling data. If Jev was available as a local model? Wow. I get like 200ms average response in the browser playground. Although my questions were about 15 keys wide. 010
FakeAndres @andres.rive.ro · 16/09/2026how come every AI person with some kind of influence has an obscure name, like: diogo, demis, jensen, dario, timnit, elon, sam 040
FakeAndres @andres.rive.ro · 11/09/2026Guys they figured out lstm and other advanced neural nets for gaming since like 10 or more years ago, stop doing free marketing for billionaires 000
FakeAndres @andres.rive.ro · 07/09/2026It’s the contrails dudestatic.klipy.comCharlie Day Conspiracy: Pepe SilviaALT: Charlie Day Conspiracy: Pepe Silvia 000
FakeAndres @andres.rive.ro · 07/09/2026I just tell it to use SOLID principles with modular files of medium to small sizes 030
FakeAndres @andres.rive.ro · 07/09/2026I agree the verge is just cooking up some drama but that’s what sells newspapers. But i don’t think LLMs are categorically *not* able to do this especially since that they are so commonly now capable with an agentic/reasoning context and memory. Right? Or am I missing something? 100
FakeAndres @andres.rive.ro · 06/09/2026I don’t see anyone dogpilling in a meaningful sense, you really just have to ignore the haters regardless. 100
FakeAndres @andres.rive.ro · 06/09/2026youtu.be/cN259fUOmmk?...youtu.beJOHN SUMMIT FT. INEZ - CRYSTALLIZED (FABIOMAEN SCHRANZ REMIX)YouTube video by NOS - Not only SoundCloud 000
FakeAndres @andres.rive.ro · 04/09/2026Just rename the company OpenBenchmarks if that’s all they talk about 000
FakeAndres @andres.rive.ro · 03/09/2026Akira 4K is currently playing at Cinemark and some other theatres, limited release 000
FakeAndres @andres.rive.ro · 01/09/2026Oh yea absolutely, maybe what I mean is we can always have bigger buses, faster read rates, cheaper memory, 2TB memory capacities etc. but totally agree, what we have now in terms of quality is all anyone really needs for the vaaast majority of tasks 110
FakeAndres @andres.rive.ro · 01/09/2026But also helpful for cybersecurity which is 50% breaking stuff and 50% figuring out how to fix stuff, so removing those guards really only brings the model closer to being more open than dangerousness. 110
FakeAndres @andres.rive.ro · 01/09/2026Abliteration essentially finds where deep in the model those guards are, and because the model is 'open' can surgically remove them, thus creating a model with no hesitation to speak about violence or create dangerous things that other models would warn the user about or tell them they can't do. 150
FakeAndres @andres.rive.ro · 01/09/2026Frontier consumer and business LLMs like Claude or GPT, and open models like Chinese Deepseek, GLM, or Llama from Meta all have internal rails and guards for certain questions. Whether it's giving a vague answer about Tianemen Square or stopping when a user asks for help coding a malicious virus. 160
FakeAndres @andres.rive.ro · 31/08/2026Skipping church to watch cooking infomercials explains so muchyoutu.beStavvy's World #196 - Nick Mullen | Full EpisodeYouTube video by Stavvy's World 000
FakeAndres @andres.rive.ro · 30/08/2026Oh yea of course, but a better solution is far more complex and expensive than ‘add a timer’ to a self install 020
