🔥 This might be the biggest AI plot twist of 2025. Alibaba just dropped Tongyi DeepResearch a 30B parameter agent that beats GPT-4o and DeepSeek-V3 at deep research using only *3.3B active parameters. 💥 Fully open-source. Trained on 2 H100s for <$500. The secret? A new

We don’t directly perceive the color of objects in the world. Instead, our brains infer color based on local context. [Link below.]

Remains normal

Normality field is weakening...

#47 The boop

Here's a quick 3 min tutorial on the absolute easiest way to use Wan Animate 2.2, Seedream and Elevenlabs voice2voice together to "remote control" any AI character in super high fidelity all inside of a single chat interface

Open sourcing InScene + InScene Annotate - a pair of LoRAs for steering in-scene image generation w/ QwenEdit. You can use a combination of prompts + annotations to generate consistent images inside a scene. In beta but v. powerful. Model + workflow below, training data soon.

Our 7th gen TPU Ironwood is coming to GA!  It’s our most powerful TPU yet: 10X peak performance improvement vs. TPU v5p, and more than 4X better performance per chip for both training + inference workloads vs. TPU v6e (Trillium). We use TPUs to train + serve our own frontier

Why cat so long? LongCat Video = Extra LOOOONG videos in ComfyUI! youtu.be/NMDJSCI_gCM

ChronoEdit - Temporal Image Modification in ComfyUI youtu.be/ZdGpgs2qFWY

Nvidia just released ChronoEdit-14B-Diffusers-Upscaler-LoRA

PSA: Your AI isn’t “smart” ... it’s on holiday with you! :) Playing at the edge of what’s possible when silicon + human hold hands. Joyful co-creation, anyone? 🎄💫 glif.app/chat/b/ICHolid…

Some people say an AI can't keep you warm over winter. To those people I say, try chatting with one locally when you have a 3090... 🧊 (Time to keep my computer room door closed and have a warm room 😄)

Time-to-Move Training-Free Motion Controlled Video Generation via Dual-Clock Denoising

Sort of like a thesaurus, but also not :) What is love? Turns out the answer may not be "baby don't hurt me. no more".... glif.app/agents/Concept…

Depth Anything 3 Recovering the Visual Space from Any Views

🐀 youtube.com/shorts/M1QLmhu…

We present Olmo 3, our next family of fully open, leading language models. This family of 7B and 32B models represents: 1. The best 32B base model. 2. The best 7B Western thinking & instruct models. 3. The first 32B (or larger) fully open reasoning model. This is a big

Power-up your generations with these EPIC Qwen Edit LoRAs! youtu.be/KYKKpVcdrqg

Many updates - now with axioms! The word is now a virus. The symbols point to more symbols, eating themselves like the Ouroboros🐍 glif.app/agents/Kaleido…