Project Glasswing: An Initial Update | Hacker News

Project Glasswing’s early update says AI security scanners are finding many real vulnerabilities, including high-severity issues.

The tools are showing strong true-positive rates and producing useful fix suggestions across monitored repositories.

The report highlights progress from Codex Security and related systems, with models from Anthropic, OpenAI, gpt-5.5-cyber, and deepseek-v4-flash mentioned.

If the results hold, AI-assisted scanning could become a standard part of code review and continuous maintenance.