Generative AI news and analysis on Web Pulse: 15433 curated articles tracking the latest Generative AI developments, tools, and research, updated continuously from vetted sources.
Apple Intelligence received approval from China's Cyberspace Administration on July 15, 2026, making it the only foreign generative AI service cleared to run on iPhones in the country, with a second sign-off from the Min…
A developer built Listing Writer, an open-source web app that turns a product name and a few details, plus an optional photo, into a ready-to-paste Etsy or eBay title, description and 13 search tags. The roughly 90-line …
Swiss AI released the Apertus 1.5 8B and 70B model weights on Hugging Face on July 24, 2026 under an Apache 2.0 license, adding 4 trillion tokens of text and multimodal training data to the 8B model and 2 trillion tokens…
Google DeepMind released Gemini 4 Argon, a 1.7-trillion-parameter multimodal foundation model that processes text, images, and code in a single forward pass and ships with a hosted REST API supporting streaming, function…
A developer built Receipt & Invoice Sorter Pro, a privacy-first local micro-SaaS utility that parses multi-vendor PDF, CSV, and JSON invoices into structured, itemized tables with automatic tax categorization. The tool r…
Developer Korny Sietsma built rust-mpd-controller, a roughly 2,000-line native Mac application written mostly by Claude Code, to replace two earlier hacks for controlling the Music Player Daemon (MPD) from a menu bar wid…
X_eTaL-extensions added a Blender integration that converts X_eTaL array programs into Python bpy calls, letting a fixed player script build scenes, keyframes, cameras and captions in Blender with no hand-written Python,…
A developer has outlined a documentation-to-video workflow built on NoteGPT's AI Explainer Video Maker, which converts existing written docs, PDFs, or plain text into explainer videos with AI voiceover, synced visuals, a…
Google's Gemini 4 Argon generated working Rust code for a blog-post filename function on the first attempt, producing the output in 15.2 seconds using 1,178 tokens at a cost of $0.014, according to a first-hand account. …
Kunlun Tech released Matrix-Game 4.0 on October 15, an open-source system that combines a game engine with real-time AI video rendering on a single consumer graphics card, delivering 480p gameplay at 60 frames per second…
A developer outlined a five-stage AI photo-to-video pipeline that turns two still photos into a 15-second vertical zombie story, using keyframe generation, beat-based prompt templates, segmented image-to-video generation…
A developer built Chat2Invoice, an open-source tool that converts pasted client chat conversations into ready-to-send invoice PDFs using Google's Gemma 3 model running locally through Ollama. The app extracts invoice fie…
Google DeepMind's published evaluation scores show Gemini 4 Argon beating Anthropic's Claude Opus 5.5 on 14 of 18 head-to-head benchmarks, including a 19.6-to-3.8 rout on Legal Work and an 84.2-to-66.8 lead at one millio…
Acrylic launched as an AI-native editor that works directly inside DOCX, XLSX, and PPTX files, editing paragraphs, tables, cells, formulas, and slides while preserving existing formatting and saving in the original file …
TypeSafe AI released Jev 1.14 under the jev-preview alias, adding experimental image input, a Box bounding-box type and an opt-in 262k-token context to its System One model at $0.042 per million input tokens. The release…
Subframe cofounder Adam Ginzberg reported that after a few thousand submissions to the Real Design Benchmark, 57% of participants preferred AI-generated screen designs over real product designs, while 33% preferred the r…
RevisePulse launched as a Show HN tool that uses Llama-3.1 to semantically analyze architectural and structural IFC models for coordination conflicts, replacing brute-force bounding-box intersection with AI reasoning. Th…
Figma announced Figma Specs, a Markdown-based editor for defining design language, at its first Config event in India, alongside public beta availability for its code layers feature and MCP support for over 20 new tools …
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…