Your curated collection of saved posts and media
Everyone understands that an AI model generating a happy dancing cat isn’t a cat and holds no internal emotions. A prompt selects “happy” or “dancing” patterns from its training corpus. It’s steering, powerful for prompting, useless if you mistake it for AI cat therapy. An LLM generating “happy” text works exactly the same way. The only difference is the chat interface. It tricks you into treating the output as coming from a speaker. But ChatGPT, Claude, or Gemini aren’t entities, they’re just a system prompt and RLHF training regime that vanishes the moment we change it. Years of instant-messaging family and friends close the loop. That inferred “speaker” is the illusion. Model ↓ Probability distribution ↓ (sampling) Output (text/image/video) ↓ [Dialogue framing → implied interlocutor] ↓ Human cognition (agency detection + narrative completion) ↓ “It thinks/feels/believes” Strip away the dialogue framing, exactly what happens with pure image or video generation, and the illusion vanishes instantly. The underlying process never changes. Don’t confuse your own psychological projections with the technology. Anthropomorphizing is a useful shortcut. But it’s a story, not the mechanism. It’s all next-token sampling. The math never changed. There’s no room for alternative explanations, only the stories we tell ourselves to make sense of data-driven statistical artifacts. Anyone can see it the moment an AI image or video glitches into ghostly shapes. Training data gets thin, the mask slips, and the illusion shatters. That’s the real AI zeitgeist.
Everyone understands that an AI model generating a happy dancing cat isn’t a cat and holds no internal emotions. A prompt selects “happy” or “dancing” patterns from its training corpus. It’s steering, powerful for prompting, useless if you mistake it for AI cat therapy. An LLM generating “happy” text works exactly the same way. The only difference is the chat interface. It tricks you into treating the output as coming from a speaker. But ChatGPT, Claude, or Gemini aren’t entities, they’re just a system prompt and RLHF training regime that vanishes the moment we change it. Years of instant-messaging family and friends close the loop. That inferred “speaker” is the illusion. Model ↓ Probability distribution ↓ (sampling) Output (text/image/video) ↓ [Dialogue framing → implied interlocutor] ↓ Human cognition (agency detection + narrative completion) ↓ “It thinks/feels/believes” Strip away the dialogue framing, exactly what happens with pure image or video generation, and the illusion vanishes instantly. The underlying process never changes. Don’t confuse your own psychological projections with the technology. Anthropomorphizing is a useful shortcut. But it’s a story, not the mechanism. It’s all next-token sampling. The math never changed. There’s no room for alternative explanations, only the stories we tell ourselves to make sense of data-driven statistical artifacts. Anyone can see it the moment an AI image or video glitches into ghostly shapes. Training data gets thin, the mask slips, and the illusion shatters. That’s the real AI zeitgeist.
Had early access to @synclabs_so latest 3.0 model. The accuracy is wild. It captures lip movement, timing, and subtle facial emotion without breaking realism. This feels natural. Well done, team. Animated on @Kling_ai 3.0 model. @therealprady @IamPrajwalKR @RudrabhaM @sama_sanjit @mhadifilms
Google launches the free Google Vids Screen Recorder Chrome extension, allowing users to capture up to 30 minutes of screen time with a single click. https://t.co/KpyUiyMyv5 https://t.co/A7iO4pIlh9
Meet Pocket 📱 Your IM, your MiniMax Agent Desktop, finally connected. Most AI agents still need you at your desk. Pocket doesn't. Sort your local files. Send attachments. Browse the web. All from a chat message. While you're nowhere near your desk. Try today! → https://t.co/VZEkxJ78Sg
Made in NYC with AWS: George Sivulka, Founder & CEO, @hebbia. New York is a central hub for global finance and the ideal home for a startup serving the world’s investors and bankers. Sivulka discusses why being a founder in NY means ‘going against the grain’ and how building on AWS has helped Hebbia scale and embrace generative AI.
The AI race is not one race, but two very different models competing. The U.S. is pushing rapid innovation driven by companies and markets, while China is advancing under strong state direction and control. Each approach has its own strengths and trade-offs. The outcome may depend on adoption, not just technology. The side that wins over more users and ecosystems could shape how AI defines the century. https://t.co/XXLaBAeCXN @bbcnews @MishaGlenny @lukemintz
Time to try out the new Gemma 4 on iPhone. If it can run agents, it might actually become a really useful part of my daily routine. https://t.co/KPG9G1xoko
HIGGS: https://t.co/5tILf5wmE6 A weaker version of HIGGS is now standard in llama.cpp: https://t.co/1iWgdUuNtT Hadamard vs rdm benchmarks (I found the same): https://t.co/ZtYLG3b6sI
QJL hurts performance: https://t.co/ZtYLG3b6sI https://t.co/a3KWuDiMLT https://t.co/MFYCMpMx2K
MiniMax-2.7 open weights confirmed and coming super soon - this is going to be a very big one 🔥 https://t.co/qrRuBN0p4N
MiniMax-2.7 open weights confirmed and coming super soon - this is going to be a very big one 🔥 https://t.co/qrRuBN0p4N
Helion is now a foundation-hosted project within the PyTorch Foundation, writes Michaël Aussems in ITdaily. The project, contributed by @Meta, aims to simplify the development of AI kernels and improve portability across different hardware platforms. Check out the article: https://t.co/OBshM7K1dT
Milla Jovovich has a Github 😏 She's co-developed the highest-scoring AI memory system ever benchmarked with @bensig Totally free and OSS. What a boss. https://t.co/AhVRf4x486
Excited to announce a new open-source, free-to-use memory tool I have been developing with my good friend @MillaJovovich. The project is called MemPalace and it is an agentic memory tool that scored 100% on LongMemEval - the industry standard benchmark for memory… this is higher
Today at #PyTorchCon Europe 2026 we’re excited to share that ExecuTorch is becoming a part of PyTorch Core. ExecuTorch extends PyTorch functionality for efficient AI inference on edge devices, from desktop/laptop to mobile phones and embedded systems. Learn more: https://t.co/REtKoVJ6GZ
At #PyTorchCon EU, Mergen Nachin of @Meta discusses bringing ExecuTorch to the next frontiers of Edge AI https://t.co/jmnrT7U3f1
🇺🇸🇭🇺La visite officielle de la honte. Alors que des preuves montrent que la Hongrie a livré des informations classifiée à la Russie, J.D Vance se rend à Budapest pour soutenir Orban avant les élections. Outre l'ingérence évidente juste avant les élections (mais on n'est plus à ça près), c'est un marqueur important de la politique étrangère des États-Unis vis-à-vis des Européens. C'est-à-dire qu'ils choisissent consciemment de soutenir celui même qui fait tout pour bloquer les institutions européennes. Il faut bien se rendre compte que dans un monde normal, les États-Unis auraient mis un maximum de pression sur la Hongrie et Orban pour le forcer à faire respecter le droit. Sans compter les révélations sur la collaboration avec la Russie. Il est donc officiel (même si on le savait déjà) que Washington soutient ceux qui affaiblissent l'Union Européenne. Washington valide, par cette visite, la collaboration entre la Hongrie et la Russie de Poutine. Il faut s'accrocher parce qu'à ce rythme, il ne restera plus grand chose à la fin du second mandat de Trump. Ronald Reagan doit se retourner dans sa tombe.
🇺🇸🇰🇷Donald Trump s'en prend désormais à la Corée du Sud, au Japon et à l'Australie. Il critique lourdement ses alliés et chante les louanges de Kim Jong Un. Je dis bien Kim Jong Un. À quel moment est-ce que les États-uniens vont-ils se réveiller et se rendre compte que Trump es
Mistral AI 🇫🇷 🤝 Sakana AI 🇯🇵 https://t.co/4fiv9Q2Djv
Mistral AI 🇫🇷 🤝 Sakana AI 🇯🇵 https://t.co/4fiv9Q2Djv
The TorchInductor team dives deep on how to generate state-of-the-art matrix multiply kernels with the new CuteDSL backend. Learn more: https://t.co/xTLltGx6SW https://t.co/b0TRPFrhN8
SNSに広がる誤情報をAIで検知する技術 総務省事業でSakana AIが開発 https://t.co/CCpZ7lRhJe
SNSに広がる誤情報をAIで検知する技術 総務省事業でSakana AIが開発 https://t.co/CCpZ7lRhJe
@thomascygn https://t.co/cnzavfkclR
Measuring how well models and systems perform in agentic tasks is fragile & difficult to scale across different domains. At #PyTorchCon Europe, @besanushi of @NVIDIA discusses the common challenges in reproducing agentic evaluations, including differences in reference implementation, error handling, trajectory post processing, and tooling definitions & best practices that can help alleviate “lightness” and build more consistent measurement pipelines
Live from #PyTorchCon Europe Patrick von Platen of @MistralAI discusses Streaming Everything, how to move from Request input to Streaming input https://t.co/vfFE63KUwc
Open Source AI Unlocks the World's Potential shares Maryam Tahhan and Nicolò Lucchesi of @RedHat at #PyTorchCon Europe 2026 https://t.co/WAmApsqE4z
Milla Jovovich? The AI engineer? https://t.co/w1cqPTTwoS
My friend Milla Jovovich and I spent months creating an AI memory system with Claude. It just posted a perfect score on the standard benchmark - beating every product in the space, free or paid. It's called MemPalace, and it works nothing like anything else out there. Instead o
PyTorch has become the backbone of production AI. From One Node to Distributed Training and Inference, @roaner of @AMD shares how the PyTorch ecosystem changed AI at #PyTorchCon Europe https://t.co/W1ax7UJPHy
The RL open source ecosystem is exploding 💥 At #PyTorchCon EU @joespeez of @reflection_ai, shares how community-led open source RL is reshaping the AI landscape https://t.co/kxP4QQJgMJ
90% of open source ML projects are using PyTorch explains @ezyang of @Meta who takes the stage at #PyTorchCon Europe to share the latest PyTorch updates and what's ahead for the framework from torch.compile to dynamic shapes & beyond https://t.co/1KzrfKeFl2
Kicking off #PyTorchCon Europe with @sparkycollier, Executive Director of @PyTorch Foundation & GM of AI & Infrastructure at @LinuxFoundation! 🎙️ Mark's keynote explores how the open source intelligence stack compounds and what it means for the future of AI infrastructure https://t.co/0FuXd6SaQP
PyTorch Foundation is excited to welcomes Helion as a Foundation-hosted project to standardize open, portable, and accessible AI kernel authoring 🎉 Helion joins the Foundation as AI model development expands from training to an inference boom, elevating the importance of serving models at scale. In this landscape, in which hardware, software, and model architectures are shifting simultaneously, engineering teams face significant hurdles in cross-platform compatibility. Helion eliminates bottlenecks associated with model architectures and execution, providing developers with radically simpler kernels, automated ahead-of-time autotuning, and greater hardware performance portability. Read more: https://t.co/1n2zrtZwCo #PyTorchCon
