A macOS hyper key utility, voice journaling, and a few AI-powered tools made the cut from today's launches.
Google shipped Gemini 3.1 Pro with the highest-ever ARC-AGI-2 score. Meanwhile, OpenAI closes a massive $100B round, Anthropic cracks down on unauthorized API usage, and the open-source community delivers new TTS, BCI, and sparse training breakthroughs.
Google's latest reasoning model sets a new state-of-the-art on ARC-AGI-2, the benchmark designed to test genuine novel reasoning. Ships with a 1M token context window and improved multimodal capabilities.
Available now in Gemini Advanced and via API. Early benchmarks show it competing with o3-level reasoning while maintaining fast inference speeds.
Quick hits — funding, tooling, and inference updates.
Video generation tools and training workflows — LTX-2 gets easier prompting, and LoRA Gym ships 16 training templates for Wan 2.1/2.2.
A factory-reset tool for Mac apps, a privacy-first drone log analyzer, and a fast offline file converter with 80+ formats.