Generative AI news and analysis on Web Pulse: 15435 curated articles tracking the latest Generative AI developments, tools, and research, updated continuously from vetted sources.
Researchers proposed Progressive Multistep Data-free Distillation (PMDD), a data-free multistep diffusion distillation method that assigns a dedicated adapted network per student distribution to avoid the biased distribu…
A new arXiv paper (2610.14136v2) introduces a correction step for diffusion models that reduces the training-inference gap caused by exposure bias, where the denoiser is trained on real images interpolated with noise but…
A training-free method that steers diffusion sampling trajectories along pullback-geometry directions of maximal output sensitivity improves sample diversity in prompt-conditioned image generation and inpainting, accordi…
ConicFlow, a new method from researchers publishing on arXiv (2610.14307v1), frames constrained sampling from a flow matching model at inference time as an optimal control problem that balances deviation from the pretrai…
A new arXiv paper (2610.19042v1) reports that input loss curvature — the Frobenius norm of the Hessian of a diffusion model's training loss — can serve as a sampling-free proxy for memorization in diffusion models. The a…
A study of zero-shot drawing on AgentDrawBench, a 200-prompt benchmark, found that general-purpose coding agents construct images through executable programs and excel at preserving structure and satisfying precise visua…
LidarVerse, a unified LiDAR world model presented in arXiv paper 2610.19201v1, achieves state-of-the-art performance on most reported metrics across all three evaluation protocols while running real-time inference at 11 …
NVIDIA shipped a new DLSS 5 AI model it says is ten times as fast as the version that launched with DLSS 5, debuting in NBA 2K27 on October 16, 2026, the same day RTX Spark laptops go on sale, according to NVIDIA's annou…
Runway is licensing its Gen-4.5 and Aleph 2.0 video models directly to enterprise customers for deployment on Baseten's dedicated inference service, adding isolated deployments and customer-specific fine-tuning, accordin…
Black Forest Labs released FLUX.3 Image Klein, an open image generation model that runs in Hugging Face Diffusers from day one and handles text-to-image, multi-reference image-to-image, and bounding-box layout control in…
Anthropic's Claude Motion, a video generation feature inside Claude's artifacts panel, turns text prompts, screenshots, and Wikipedia links into editable animated videos by building the animation in code rather than rend…
Google's Anti-gravity agentic development platform launched a teamwork preview that spins up a lead agent, worker agents and auditor agents to run in parallel for hours on a single prompt, powered by the new Gemini 4 Arg…
Fish Audio released Drama 3 on October 13, 2026, an open-weight text-to-speech model with a 35-billion-parameter mixture-of-experts backbone (about 3 billion parameters active per token) that accepts stage directions in …
A freelance developer in Spain replaced a paid bookkeeping tool with roughly 2,000 lines of Python and 45 tests, using Claude only to extract fields from supplier PDFs while plain code handles numbering, VAT, IRPF, total…
Meta released Muse Spark Open, an open-weight multimodal mixture-of-experts model with roughly 816B total parameters and about 42B active per token, which the company calls the most capable US open-weights model to date.…
Scenario co-founder and CEO Emmanuel de Maistre announced on October 14th an open-source Blender extension that lets AI agents inspect and manipulate 3D scenes, build camera paths and request renders. The plugin places a…
A developer built Review Reply Writer, a local-first tool that uses Ollama and Gemma 3 4B to draft responses to customer reviews for independent cafe and hospitality operators, with an approval boundary that blocks any u…
Kai, Daniel's AI assistant, moved the spoken line at the end of every answer from ElevenLabs to Fish Audio's Drama 3 model, cloning its existing voice from a 30-second sample at 1.25× speed. Drama 3 accepts free-form inl…
Simon Willison built a Three.js game called "Tree Appreciator" by prompting Claude Opus 5.5 Max in the default Claude iPhone app's chat mode to generate two algorithmically created trees and reward players for appreciati…
Developer Brandon Thomas released WordCraft, GridCraft and DeckCraft, free open-source clones of Microsoft Word, Excel and PowerPoint written in Rust with help from Anthropic's Claude, published on GitHub under MIT and A…