Tag
This paper shows that text+image coding agents using sandboxed tool-use can match or outperform native omni-modal models on audio-video benchmarks, converting omni-modal tasks into retrieval and information-processing problems.
Runtime offers sandboxed coding agents for team collaboration, launched on Product Hunt.
An open-source CLI tool for Obsidian that provides sandboxed AI agents for audio transcription, deep research, and mind-mapping, designed to accelerate note-taking without modifying the user's vault.