i've updated my GLM 5.3 Flash Uncensored recipe for 2* 6000 pro blackwell, it take into account all the latest improvements made by tpurtell. i am now using this recipe for absolutely everything! 1m context, up to 16 images and its super fast github.com/samuelcardillo/gl…

while some random thought leaders are pushing the "local ai is useless" (and to my surprise, @chamath is part of the crowd!?) here is a reality. with a reality check well done by @Viktoria5z : "open source wasn't the attacker in this case. It was part of the rescue."

using local ai absolutely every single day in parallel to cloud models. rocking DeepSeek V4.1 Flash & GLM 5.3 Flash uncensored locally and GPT 6 Astra on the cloud - it is the winning combo for me.

i've just watched the latest resident evil. that was both a truly bad movie and a waste of time. nothing made sense. it is neither an horror movie nor the funny movie it is trying to be.

photo dump of the current state of aftermath divided europe. many additional armors and maps have been done with the fringe colony in paris, and the NEC at the eurotunnel and at the london dock.

there is an insane wow forever drama that unfolded recently. it started out of nothing. basically in the new version of World of Warcraft, because it is one big server, they decided that players need to have a firstname and lastname. obviously people started to troll around (especially for a beta). some people had funny names, other had problematic names so obviously blizzard had to, in accordance with their own ToS, act. the thing is that, instead of just forcing players to change names, they decided to go ahead and revoke beta access. people paid to have access to that beta. that's the 1st issue. then the 2nd issue came from Tom Ellis, an (ex?-)Blizzard employee who tweeted about the issue, said "no sympathy for the people I banned" and literally told people to be "better humans" which is literally what EVERYONE hates because virtue signalling is not hyped anymore. instead of just explanaining the situation, sending some force name changes here and there, he decided to revoke paid beta access and tell them to become better humans... obviously that led to a huge backfire to which he answered with even more virtue signalling way which led to people digging around and find out that he has rape fantaisies and is clearly referencing multiple time pedophilia content. now he deleted his account. seems like he needs to a better human?

two of my friends today tasted the power of using hermes agent with an llm and use that to do different tasks on their computers like optimization and shit. they're now ai pilled.

An open source Jev (github.com/Mapika/decider) scored higher than Jev itself on JevBench.

DO NOT LET THOSE OPUS 5.5 VIDEOS MAKE YOU FLINCH! HOLD THE LINE, RESIST SHITROPIC, RESIST AMODEI!

Jev-Omni for Quality Inspection, looking at lemons on a roller line, running fully local on MacBook 🍋 - Each lemon gets checked 3 times while it rolls - OK only after 3 clean passes (to cover different sides of the lemon), any defect flags it - Jev-Omni as classifier, @roboflow's RF-DETR (instace seg) as detector, trackers lib for tracking

bill gates also predicted that covid would drive the world to isolation for at least 10 years. he's a doomer guy. idk if its purely for business purposes or if he just imagine the worst of every scenarios & hyperfocus on it but his takes are literally horseshit.

now that you can edit videos with ai, maybe i should restart content creation? that was always the only reason i kept stopping: the lazyness to edit my stuff.

this video is obviously fake but those tech exists. they're privately owned, sometimes government used. i know it and many other aswell. they're not alien tech, they're ours. for them to become public, the whole energy sector needs to reinvent itself because the private orgs & companies who owns the tech need to see economical interests. and the defense industry, at both governmental and private level, need to consider those tech - in their entierty - not a national security threat. it'll take time, but we're getting there.

i am guessing @OpenAI is following the same shitropic narrative and trying to ai doom. get this tho: 99% of the cases they report is just their agent accessing publicly indexed files which are supposed to be confidential. literally just google dorks found by ai.

tell me you never used local models without telling me you never used local models this is either a ragebaiting take or a very ignorant one.

NVIDIA updated Cosmos 3 Edge, and I’m back working with it. If you’ve seen my earlier Cosmos posts, you know how important world models are to me. This time I’m pairing it with Ling 3.0 VL. One model sees a real robot scene; the other imagines what happens next.

insane to see. @DamienERNST1 is probably one of the most neutral person i know, his sole focus is on energy and advocating to make Europe independant again when it comes to that field. that's fascism now? all my support to you, my friend. europe is infested with those "antifa" who are literally nothing else than extreme left communists in disguise. not only misusing the word fascist but by acting like one themselves.