Overview
AI lab building frontier coding models with massive context windows for autonomous software engineering tasks.
Details
Magic.dev is an AI research lab building frontier-class coding models with extremely long context windows (100M+ tokens in their LTM-2 model) for autonomous software engineering. The company emphasizes reasoning over very large codebases and complex engineering tasks. While Magic has been quieter than some competitors on commercial product details, its models are aimed at autonomous coding agent use cases. Magic is significant for its focus on long-context coding models as a path to better autonomous software engineering.
Tags
coding, foundation-model, autonomous-agent, research