← All projects

Magic

Building frontier code models to automate software engineering and research.

AI Toolsagicode-generationlarge-language-modelsreinforcement-learningai-researchsoftware-automation
Magic screenshot

About

Magic is an AI research company focused on building frontier code models to automate software engineering and AI research as a path toward safe AGI. Their approach combines frontier-scale pre-training, domain-specific reinforcement learning, ultra-long context, and inference-time compute. The company has raised $515 million and operates a small team working on fundamental research problems.

Problem

Automating software engineering and AI research to accelerate progress toward safe AGI.

For

AI researchers and software engineers

How it works

Combines frontier-scale pre-training, domain-specific reinforcement learning, ultra-long context, and inference-time compute to build code models that automate engineering and research tasks.

Business model

unknown

Status

unknown

Company

Magic

Similar projects