Meta's new structured prompting technique makes LLMs significantly better at code review — boosting accuracy to 93% in some cases

This technique can be used out-of-the-box, requiring no model training or special packaging. It is code-execution free, which ...
InfoWorld

OpenAI launches Aardvark to detect and patch hidden bugs in code

Currently in private beta, the GPT-5-powered security agent scans, reasons, and patches software like a real researcher, aiming to embed AI-driven defense into the development workflow. OpenAI has ...