OpenAI Codex is OpenAI’s latest GPT-based language based model for GitHub Copilot that generates code. Here we test if Codex can debug code, and even correct it’s own errors. We look at both fixing error messages and identifying silent errors. The results were far from perfect but quite intriguing! Check out my channel for more OpenAI Codex content if you find this interesting.
Previous video on OpenAI Codex: https://youtu.be/7QWVJ5rWy2s
Link to API post: https://openai.com/blog/openai-codex/