# \#diffusion **URL:** https://www.artofsm.art/tag/diffusion/309.md [Latest](https://www.artofsm.art/latest.md) · [Categories](https://www.artofsm.art/categories.md) · [Tags](https://www.artofsm.art/tags.md) --- ## [Mercury 3 Is INSANE: I Tested 13 Live Examples (+ 100 MILLION FREE Tokens)](https://www.artofsm.art/t/mercury-3-is-insane-i-tested-13-live-examples-100-million-free-tokens/26668)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [22 October 2026 22:54 UTC](https://www.artofsm.art/t/mercury-3-is-insane-i-tested-13-live-examples-100-million-free-tokens/26668 "2026-10-22T22:54:33Z")
Mercury 3, a diffusion language model from Inception Labs, offers a novel text generation approach by refining rough drafts through multiple passes, delivering impressive speed and versatility in tasks like… --- ## [Stanford EE292R AI and Data for Science | Autumn 2026 | From World Models to World Agents](https://www.artofsm.art/t/stanford-ee292r-ai-and-data-for-science-autumn-2026-from-world-models-to-world-agents/26664)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [22 October 2026 22:24 UTC](https://www.artofsm.art/t/stanford-ee292r-ai-and-data-for-science-autumn-2026-from-world-models-to-world-agents/26664 "2026-10-22T22:24:53Z")
The lecture explores the advancement of physical AI through world models—generative systems that create realistic video environments by learning from data rather than relying on traditional simulators—and d… --- ## [Do Diffusion LLMs Scale? Testing Mercury 3 + Decide](https://www.artofsm.art/t/do-diffusion-llms-scale-testing-mercury-3-decide/26451)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [20 October 2026 13:20 UTC](https://www.artofsm.art/t/do-diffusion-llms-scale-testing-mercury-3-decide/26451 "2026-10-20T13:20:40Z")
The video highlights Inception’s diffusion language models, Mercury 3 and Mercury Decide, which achieve high-speed, intelligent text generation and rapid decision-making by iteratively refining outputs in p… --- ## [NVIDIA's Sana: 1024px Image in 0.4 Seconds? I Tested It](https://www.artofsm.art/t/nvidias-sana-1024px-image-in-0-4-seconds-i-tested-it/25966)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [15 October 2026 11:54 UTC](https://www.artofsm.art/t/nvidias-sana-1024px-image-in-0-4-seconds-i-tested-it/25966 "2026-10-15T11:54:27Z")
The video tests NVIDIA’s Saana text-to-image model, confirming its rapid 1024x1024 image generation speed of around 0.1 seconds on high-end GPUs like the H100, while noting significantly slower performance … --- ## [ByteDance Just Made a 33B Video + Audio Model Run in 4 Steps (DMAD)](https://www.artofsm.art/t/bytedance-just-made-a-33b-video-audio-model-run-in-4-steps-dmad/25037)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [5 October 2026 09:54 UTC](https://www.artofsm.art/t/bytedance-just-made-a-33b-video-audio-model-run-in-4-steps-dmad/25037 "2026-10-05T09:54:24Z")
ByteDance and Texas A&M developed DMAD, a novel method that enables a massive 33 billion parameter audio-video model to generate high-quality outputs in just four steps, significantly faster than traditiona… --- ## [Box's Aaron Levie: On Reinventing Yourself in the AI Age and Enterprise Diffusion](https://www.artofsm.art/t/boxs-aaron-levie-on-reinventing-yourself-in-the-ai-age-and-enterprise-diffusion/23985)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [15 September 2026 12:21 UTC](https://www.artofsm.art/t/boxs-aaron-levie-on-reinventing-yourself-in-the-ai-age-and-enterprise-diffusion/23985 "2026-09-15T12:21:01Z")
Aaron Levie, CEO of Box, discusses how AI is transforming enterprise software by emphasizing the importance of application companies that integrate advanced AI models into specific business workflows, overc… --- ## [Ask the Experts Session #6: Inside AMD AI Playbooks](https://www.artofsm.art/t/ask-the-experts-session-6-inside-amd-ai-playbooks/22990)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [20 August 2026 20:54 UTC](https://www.artofsm.art/t/ask-the-experts-session-6-inside-amd-ai-playbooks/22990 "2026-08-20T20:54:58Z")
In this session, AMD experts introduced the AMD AI Playbooks, a curated collection of continuously tested, step-by-step guides that help developers run AI workloads efficiently on various AMD hardware platf… --- ## [Infra behind Krea 2: How to train and serve at scale — Gabriel Jorge Menezes, Krea.ai](https://www.artofsm.art/t/infra-behind-krea-2-how-to-train-and-serve-at-scale-gabriel-jorge-menezes-krea-ai/22851)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [18 August 2026 17:20 UTC](https://www.artofsm.art/t/infra-behind-krea-2-how-to-train-and-serve-at-scale-gabriel-jorge-menezes-krea-ai/22851 "2026-08-18T17:20:30Z")
Gabriel Jorge Menezes from Krea.ai details the infrastructure and strategies behind training and serving their diffusion transformer model Krea 2, highlighting challenges like GPU failures, the importance o… --- ## [Training Krea 2: What matters in generative model training — Sangwu Lee, Krea.ai](https://www.artofsm.art/t/training-krea-2-what-matters-in-generative-model-training-sangwu-lee-krea-ai/22828)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [18 August 2026 14:29 UTC](https://www.artofsm.art/t/training-krea-2-what-matters-in-generative-model-training-sangwu-lee-krea-ai/22828 "2026-08-18T14:29:14Z")
Sangwha Lee from Krea.ai detailed the training of their Create 2 image foundation model, emphasizing the importance of data quality and stylistic diversity through advanced curation techniques and a diffusi… --- ## [The BEST local AI music generator is here!](https://www.artofsm.art/t/the-best-local-ai-music-generator-is-here/22531)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [15 August 2026 03:40 UTC](https://www.artofsm.art/t/the-best-local-ai-music-generator-is-here/22531 "2026-08-15T03:40:33Z")
The video introduces Miniax Music, a versatile and compact open-source AI music generator capable of producing high-quality songs across various genres, and demonstrates how to install and use it via Comfy … --- ## [MiniMax H3: My Most Surprising AI Video Tests Yet](https://www.artofsm.art/t/minimax-h3-my-most-surprising-ai-video-tests-yet/22239)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [7 August 2026 16:20 UTC](https://www.artofsm.art/t/minimax-h3-my-most-surprising-ai-video-tests-yet/22239 "2026-08-07T16:20:25Z")
The video showcases the Hailuo H3 AI model’s advanced multimodal text-to-video generation capabilities, highlighting its versatility, fine-grain editing features, and the creator’s custom prompt generator t… --- ## [Chai Discovery's Bitter Lesson: Drug Design Is Another Scaling Problem](https://www.artofsm.art/t/chai-discoverys-bitter-lesson-drug-design-is-another-scaling-problem/22093)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [4 August 2026 12:24 UTC](https://www.artofsm.art/t/chai-discoverys-bitter-lesson-drug-design-is-another-scaling-problem/22093 "2026-08-04T12:24:24Z")
Chai Discovery is transforming drug design by leveraging AI to create scalable, foundational models that design molecules from scratch, significantly improving success rates and accelerating the drug discov… --- ## [Stanford CS229 Machine Learning | Spring 2026 | Lecture 12: Representation Learning](https://www.artofsm.art/t/stanford-cs229-machine-learning-spring-2026-lecture-12-representation-learning/21970)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [31 July 2026 21:07 UTC](https://www.artofsm.art/t/stanford-cs229-machine-learning-spring-2026-lecture-12-representation-learning/21970 "2026-07-31T21:07:52Z")
The lecture covers diffusion models and their noise prediction training approach, then explores foundation models emphasizing large-scale pre-training and adaptation techniques like linear probing, fine-tun… --- ## [Stanford CS229 Machine Learning | Spring 2026 | Lecture 11: Diffusion Models](https://www.artofsm.art/t/stanford-cs229-machine-learning-spring-2026-lecture-11-diffusion-models/21969)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [31 July 2026 21:04 UTC](https://www.artofsm.art/t/stanford-cs229-machine-learning-spring-2026-lecture-11-diffusion-models/21969 "2026-07-31T21:04:25Z")
The lecture introduces diffusion models as advanced generative models that learn to reverse a gradual Gaussian noising process applied to images, enabling high-quality image generation by modeling the rever… --- ## [New AI Just Reinvented Minecraft Worlds](https://www.artofsm.art/t/new-ai-just-reinvented-minecraft-worlds/21133)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 3\ **Last updated:** [12 July 2026 18:05 UTC](https://www.artofsm.art/t/new-ai-just-reinvented-minecraft-worlds/21133 "2026-07-12T18:05:23Z")
The video presents a groundbreaking AI-driven infinite terrain generator that combines the speed of noise-based methods with advanced diffusion models to create vast, detailed, and coherent Minecraft worlds… --- ## [Research to Reality: Bringing Frontier ML Research to Production - Vaidas Razgaitis, Higharc](https://www.artofsm.art/t/research-to-reality-bringing-frontier-ml-research-to-production-vaidas-razgaitis-higharc/20646)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [28 June 2026 19:50 UTC](https://www.artofsm.art/t/research-to-reality-bringing-frontier-ml-research-to-production-vaidas-razgaitis-higharc/20646 "2026-06-28T19:50:28Z")
Vaidas Razgaitis from Higharc outlines the challenges of transitioning advanced ML research into production, emphasizing the need for clear research documentation, a modular and well-structured codebase, an… --- ## [Forget Nano Banana… This Is FREE](https://www.artofsm.art/t/forget-nano-banana-this-is-free/20581)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [26 June 2026 07:50 UTC](https://www.artofsm.art/t/forget-nano-banana-this-is-free/20581 "2026-06-26T07:50:30Z")
The video showcases powerful open-source AI image models like Nvidia’s Cosmos 3 and Idoggram 4, which rival paid options such as Google Nano Banana by excelling in areas like physics-based realism and graph… --- ## [This AI Generates Text Like Stable Diffusion Makes Images](https://www.artofsm.art/t/this-ai-generates-text-like-stable-diffusion-makes-images/20434)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [22 June 2026 12:35 UTC](https://www.artofsm.art/t/this-ai-generates-text-like-stable-diffusion-makes-images/20434 "2026-06-22T12:35:50Z")
Google’s Diffusion Gemini is a novel AI text generation model that uses diffusion techniques to generate entire sequences of tokens in parallel, significantly increasing speed compared to traditional sequen… --- ## [You Might Not Need 50 Diffusion Steps — Ziv Ilan, Nvidia](https://www.artofsm.art/t/you-might-not-need-50-diffusion-steps-ziv-ilan-nvidia/20275)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [16 June 2026 13:20 UTC](https://www.artofsm.art/t/you-might-not-need-50-diffusion-steps-ziv-ilan-nvidia/20275 "2026-06-16T13:20:30Z")
Ziv Ilan from Nvidia discussed optimizing diffusion models for faster, real-time image and video generation by employing techniques like quantization, caching, and distillation, which collectively reduce la… --- ## [New #1 Open Source Image AI? | SenseNova-U1 Mac & Windows Guide & TESTS](https://www.artofsm.art/t/new-1-open-source-image-ai-sensenova-u1-mac-windows-guide-tests/20217)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [15 June 2026 00:42 UTC](https://www.artofsm.art/t/new-1-open-source-image-ai-sensenova-u1-mac-windows-guide-tests/20217 "2026-06-15T00:42:15Z")
The video presents SenseNova-U1, a top-performing open-source image generation AI that uniquely integrates language and vision models for advanced interleaved reasoning, excelling in creating detailed infog… --- ## [DeepMind Was Two Steps Ahead, AGAIN!](https://www.artofsm.art/t/deepmind-was-two-steps-ahead-again/20205)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [14 June 2026 12:50 UTC](https://www.artofsm.art/t/deepmind-was-two-steps-ahead-again/20205 "2026-06-14T12:50:36Z")
The video highlights DeepMind’s advanced and multifaceted approach to achieving AGI, combining innovations in transformer architectures, diffusion-based language models, and generative world models to overc… --- ## [New BEST local AI image generator is here!](https://www.artofsm.art/t/new-best-local-ai-image-generator-is-here/19989)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [9 June 2026 03:48 UTC](https://www.artofsm.art/t/new-best-local-ai-image-generator-is-here/19989 "2026-06-09T03:48:19Z")
The video introduces Ideogram 4, a powerful open-source local AI image generator featuring a unique bounding box system that allows precise control over image composition, enabling the creation of complex s… --- ## [Text Diffusion — Brendon Dillon, Google DeepMind](https://www.artofsm.art/t/text-diffusion-brendon-dillon-google-deepmind/19558)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [4 June 2026 18:28 UTC](https://www.artofsm.art/t/text-diffusion-brendon-dillon-google-deepmind/19558 "2026-06-04T18:28:55Z")
Brendon Dillon from Google DeepMind presented text diffusion models, which generate text by iteratively refining random token sequences, enabling bidirectional reasoning, self-correction, and fast in-place … --- ## [I Used Diffusion Blocks to Achieve 2-3x Memory Savings in Model Training](https://www.artofsm.art/t/i-used-diffusion-blocks-to-achieve-2-3x-memory-savings-in-model-training/19450)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [2 June 2026 14:20 UTC](https://www.artofsm.art/t/i-used-diffusion-blocks-to-achieve-2-3x-memory-savings-in-model-training/19450 "2026-06-02T14:20:41Z")
The video explores diffusion blocks, a training method that reduces memory usage by splitting neural networks into independently trained denoising blocks, achieving 2-3x memory savings without compromising … --- ## [The BEST AI for 4K images. Free & fast](https://www.artofsm.art/t/the-best-ai-for-4k-images-free-fast/19437)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [2 June 2026 03:46 UTC](https://www.artofsm.art/t/the-best-ai-for-4k-images-free-fast/19437 "2026-06-02T03:46:42Z")
The video showcases Nvidia’s open-source Pixel Diffusion (PD) as a fast, efficient, and high-quality AI model for generating and upscaling images to 4K, outperforming competitors like Seed VR2 in both speed… --- ## [This New 1-Bit Image Model Changed My View On Image Models](https://www.artofsm.art/t/this-new-1-bit-image-model-changed-my-view-on-image-models/19341)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [29 May 2026 18:05 UTC](https://www.artofsm.art/t/this-new-1-bit-image-model-changed-my-view-on-image-models/19341 "2026-05-29T18:05:56Z")
Timothy Carbat reviews Prism ML’s Bonsai 4B image model, which uses binary and ternary weight retraining to drastically reduce memory usage and model size while maintaining decent image quality, making loca… --- ## [Inference, Diffusion, World Models, and More | YC Paper Club](https://www.artofsm.art/t/inference-diffusion-world-models-and-more-yc-paper-club/19299)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [28 May 2026 20:57 UTC](https://www.artofsm.art/t/inference-diffusion-world-models-and-more-yc-paper-club/19299 "2026-05-28T20:57:42Z")
The YC Paper Club’s inaugural session brought together AI founders and researchers to discuss cutting-edge papers on topics including speculative decoding for faster language model inference, diffusion mode… --- ## [Compiling the Full Diffusion Pipeline: 4x Faster Image Generation on MI355X](https://www.artofsm.art/t/compiling-the-full-diffusion-pipeline-4x-faster-image-generation-on-mi355x/18759)
**Author:** [@merefield](https://www.artofsm.art/u/merefield)\ **Replies:** 1\ **Last updated:** [14 May 2026 17:14 UTC](https://www.artofsm.art/t/compiling-the-full-diffusion-pipeline-4x-faster-image-generation-on-mi355x/18759 "2026-05-14T17:14:26Z")
The video highlights Modular’s innovative, hardware-agnostic AI software stack—featuring the Mojo programming language and Max framework—that significantly accelerates image diffusion model performance on A… --- ## [AI Simulated OS Is Absurd](https://www.artofsm.art/t/ai-simulated-os-is-absurd/18501)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [8 May 2026 18:15 UTC](https://www.artofsm.art/t/ai-simulated-os-is-absurd/18501 "2026-05-08T18:15:25Z")
The video discusses the concept of AI-simulated operating systems, where AI models internally generate and manage entire OS environments by simulating user interactions and system responses, potentially rep… --- ## [2025 Rewind: How This Gen Z Founder Raised $11 Million To Build The AI-Powered Closet From 'Cluel](https://www.artofsm.art/t/2025-rewind-how-this-gen-z-founder-raised-11-million-to-build-the-ai-powered-closet-from-cluel/17988)
**Author:** [@artesia](https://www.artofsm.art/u/artesia)\ **Replies:** 1\ **Last updated:** [24 April 2026 18:20 UTC](https://www.artofsm.art/t/2025-rewind-how-this-gen-z-founder-raised-11-million-to-build-the-ai-powered-closet-from-cluel/17988 "2026-04-24T18:20:44Z")
The video features Jenny, the Gen Z founder of Ulta, an AI-powered personal stylist app inspired by “Clueless” that offers personalized outfit recommendations by analyzing users’ closets, preferences, and c… [Next page](https://www.artofsm.art/tag/diffusion/309.md?match_all_tags=true&page=1&tags%5B%5D=diffusion)