preview

Karthinks and Tecosaur rewrote Org mode's LaTeX preview feature from the ground up, enabling asynchronous near-WYSIWYG previews with live updates, theme matching, and 200-800 snippets per second bulk generation, along with a deep-dive explanation of how it works.

Microsoft is extending Defender for Endpoint to Linux desktops, bringing EDR, next-gen antivirus, vulnerability management, and threat analytics to workstations using the same agent and deployment methods already used for Linux servers. The feature enters public preview for distributions like RHEL, Ubuntu, Debian, SLES, and Fedora, requiring Defender for Endpoint Plan 2 or Microsoft 365 E5/E7.

Mistral AI team member announces that Mistral Large 4 is still in preview, with continued reinforcement learning runs yielding further improvements, and the full release is expected by the end of the month.

The ChatGPT desktop app for Linux is now available in preview, supporting distributions like Ubuntu, Debian, Fedora, and Arch Linux with instructions for installation and setup.

An early preview of GEFS, a crash-safe and snapshotting file system, has been ported to OpenBSD. The port is experimental with several issues but is open for community testing.

Tencent has released Hy4 Preview, an open-weight text-only large language model with 770B total parameters, 49B active parameters, and a 1M token context window, featuring reasoning capabilities with 'high' and 'no_think' effort levels.

Tencent AI has announced the Hy4 preview model with 770B total parameters, 49B active parameters, 1M context, and open weights, marking a significant update in the competitive Chinese AI landscape.

The article announces the upcoming release of Qwen3.8-Flash, a preview of the Qwen4 architecture, on Hugging Face with a 22-hour countdown.

Unitree has previewed a humanoid robot called Superman that can jump higher than any human and has a faster top speed than Usain Bolt.

Mistral AI announces OCR 4.1, an updated OCR service with native paragraph-level bounding box extraction, structural block labels, and block-level confidence scores, now in public preview.

OpenAI is previewing Ultrafast mode for GPT-5.6 Sol, offering up to 14x speed, initially in the OpenAI API to select customers with expanded access over time.

A hands-on preview of Marvel's Wolverine for PS5, highlighting its visceral combat and the developer's take on Logan as a vulnerable, reluctant hero.

Kijai shares a quickly trained 2D tiny VAE for MiniMax-H3, intended to improve latent previews in ComfyUI. A better alternative trained by madebyollin is also linked.

SenseNova released a preview of its U1.5 Lite model, showing benchmark gains in image generation and editing, with native 4K output and improved Chinese/English text rendering, though acknowledged weaknesses remain.

Nous Research has opened free access to FLUX 3 Preview for all Nous Portal users, including free tier, and is running a short film contest with prizes.

JEP 401: Value Objects (Preview) has been merged into OpenJDK master, marking a significant milestone for Project Valhalla's value types in Java.

NASA flight directors preview three spacewalks at the International Space Station in August.

Alibaba's Qwen team is releasing daily checkpoints of Qwen3.8-Max-Preview, showing broad improvements especially in web frontend, with plans for an open-weight official version.

Qwen3.8, a 2.4-trillion-parameter language model, is being released as open-weight by Alibaba, with a preview available on Alibaba's Token Plan, Qoder, and QoderWork.

Alibaba launches Qwen3.8, a 2.4 trillion parameter model, with an open-weight release planned soon. A preview version, Qwen3.8-Max-Preview, is now available on Alibaba's Token Plan platform, Qoder, and QoderWork.