Finally out and officially open! Still behind with practical testing of Claude's equivalent, but Aarvark, based on early results, sounds more interesting. I gave up on Claude a while back and stopped using it in my code-audit LLM based pipeline, mostly because I got tired of setting guardrails for false-positives/negatives in their models. Gemini has been crunching code for me since, and still not exactly there, but it delivers. I'm very curious how OpenAI, with the Aardvark sauce, compares with them. It is announced, similar to Claude Code Security, as a git security monitoring pipeline. But you can of course hack your way around.