All published AI Radar articles.

OpenAI collaborated with Thrive and Crete to develop a self-improving tax agent utilizing the Codex model. This agent aims to automate tax filings, enhance accuracy, and streamline workflow processes.

Lim-Code is an AI programming assistant for VS Code that supports multiple AI models, including Google Gemini, OpenAI-compatible APIs (like ChatGPT), and Anthropic's Claude. It features an intelligent tool system for code operations, file management, search, terminal commands, and image generation, along with Model Context Protocol (MCP) support for extending tool capabilities.

RuFlo is an open-source agent orchestration platform designed for Claude AI, enabling the deployment of multi-agent systems, coordination of autonomous workflows, and building conversational AI. It features enterprise-grade architecture, self-learning swarm intelligence, RAG integration, and native Claude Code/Codex integration.

Tibsfox has introduced GSD Skill Creator, an open-source, agent-based framework designed to enhance guided software development (GSD). The project aims to address common challenges in AI-assisted development, such as context degradation, lack of session memory, and difficulties in scaling complex tasks.

Vstorm-co has released a full-stack AI application generator template that integrates FastAPI and Next.js. It supports AI agents, RAG, streaming, authentication, and over 20 integrations out-of-the-box, offering a foundation for building production-ready AI applications.

AdminForth is an open-source framework for building agentic admin panels using TypeScript and Vue.js. It aims to simplify the creation of CRUD pages and offers extensibility with custom Vue components and built-in plugins for features like AI writing and image generation.

Agent Network is a JavaScript-based multi-agent orchestration framework that allows users to run networks of AI agents locally or across servers. It supports various large language models (LLMs) including Claude, GPT, MiniMax, DeepSeek, GLM, Kimi, InternLM, and OpenRouter, and features a web dashboard for management and collaboration.

LocateAnything is a new framework for vision-language grounding and detection that uses Parallel Box Decoding (PBD) to improve both speed and accuracy. Unlike traditional methods that decode 2D boxes token by token, PBD decodes geometric elements as atomic units in a single step, enhancing parallelism and preserving geometric coherence. The framework is supported by LocateAnything-Data, a large dataset with over 138 million training samples.

Researchers have introduced SpatialBench, a new benchmark designed to holistically assess the generalization capabilities of spatial foundation models across diverse tasks, viewpoints, scene domains, and input densities. The benchmark evaluates 41 models across 19 datasets and 546 scenes, revealing that current models are not yet "all-round players" and highlighting the importance of domain alignment and data quality over simple dataset scaling.

AINYC has released Canonry, an open-source, self-hosted platform designed for Answer Engine Optimization (AEO) monitoring and operation. It allows users to track citations across various AI engines like Gemini, ChatGPT, Claude, and Perplexity, diagnose traffic issues, and execute fixes.