All published AI Radar articles.

Mnemo Cortex is an open-source memory coprocessor designed for AI agents, offering persistent recall, semantic search, and crash-safe memory capture without requiring hooks. It includes features like cross-agent overnight synthesis, an auto-compiled knowledge base (WikAI), agent-to-agent messaging, and a structured facts store with confidence tracking.

Friday is an open-source, local-first AI personal operations platform designed for supervised, evidence-backed automation. It allows users to set goals for AI agents, which then use configured tools, seek human approval for critical steps, and provide verifiable results and memory.

GitHub Copilot has announced the general availability of Claude Opus 4.8 for its service, enhancing its AI-powered coding assistance.

GeniePod has released GenieClaw, a Rust-based AI agent layer designed for smart homes. It focuses on low-latency, privacy-first operation with limited context windows, optimized for devices like the Jetson Orin Nano Super. The latest alpha release, v1.0.0-alpha.9, introduces Qwen3-4B as the default model, improves voice interaction, and enhances CI/CD and supply chain security.

OpenAI has unveiled its Frontier Governance Framework, detailing its approach to AI safety, security, and risk management. This framework aims to align OpenAI's practices with upcoming regulations in the EU and California.

Anthropic has announced Claude Opus 4.8, an upgraded version of its Opus class of models, designed to deliver enhanced performance in coding, agentic tasks, and professional applications, while maintaining consistency for extended operations.

Sutando is an open-source AI agent designed for macOS that operates in real-time and can autonomously rewrite its own code. It aims to provide a personal AI superpower, capable of handling tasks via voice, vision, and screen interaction, and can integrate with various communication channels and even scale across multiple Mac machines.

Core-Mate has released OpenGUI, an Android GUI agent framework that enables AI to perceive, plan, and interact with real mobile applications through their graphical user interfaces. The framework supports long-running tasks and offers various integration methods for AI agents.

A new study investigates whether vision-language models (VLMs) offer a global advantage over large language models (LLMs) in aligning with human language processing during natural reading. The research, conducted under text-only conditions, suggests that multimodal pretraining does not uniformly improve human alignment. Instead, any VLM advantage appears selectively, particularly when sentences have strong visual semantic content.

Researchers have introduced NEO-ov, a native foundation model designed to learn cross-frame and pixel-word correspondence end-to-end. Unlike traditional vision-language models (VLMs) that combine separate encoders and decoders, NEO-ov eliminates module boundaries to enable unified spatiotemporal modeling, aiming to improve fine-grained visual perception across multiple images and videos.