AI startup Recursive has released Flint, its first open-source coding model designed to run on laptops. Developed largely with the help of its automated research system, the Eureka Machine, Flint serves as an early test of recursive self-improvement—where AI builds better AI—under human oversight.
On-device Model
端侧模型
AI models that run directly on local device hardware rather than relying on cloud servers
Curated Intelligence
Google has released an experimental note-taking app, AI Edge Foresight, which uses the on-device EmbeddingGemma 2 model to transcribe meetings and generate summaries entirely offline on macOS. This local-first approach addresses privacy concerns by eliminating cloud dependencies, though it currently supports only Apple Silicon devices.
Google released AI Edge Foresight, a macOS app powered by the on-device multimodal model EmbeddingGemma 2. The tool operates entirely offline to generate polished meeting minutes from audio and notes, while also supporting real-time Q&A based on meeting content.
In response to macOS 27 removing the global toggle for Apple Intelligence, a developer released RemoveMacAI, a command-line tool. It allows users to fully reversibly disable AI features and free up approximately 12GB (up to 30GB) of disk space, addressing the issue of local models occupying storage without easy removal options.
Google introduces a method to retrofit Multi-Token Prediction (MTP) onto existing 'frozen' Gemini Nano models for Pixel devices. This approach addresses mobile autoregressive bottlenecks by improving inference speed and energy efficiency without retraining the base model.