When we're talking about AI products, you mean it's everything from like OpenClaw agents to people running like Claude Code on their desktop, right? That, that's the kind of things we're talking about.

Yeah.

Yeah.

So, so a lot of the connectors, um, I, I kinda listed this out and, and, and, um, yeah, like, uh, uh, you know, a lot of things that you would connect with Claude Code or OpenAI.

Um, uh, a combination of like AI gateways, AI agents, AI workflow automation, workload scaling, um, model building.

Um, and a lot of these tend to be open source products, right? So, um, some of, some of the products, uh, ex-as example, LangFlow, uh, LangLLM, OpenClaw, uh, N8-NAN.

I probably said that wrong.

Yeah

You know, walk us through how you separated like, you know, AI-driven discovery from the early flood of what you, you know, would call AI slop, too.

Yeah, yeah.

Two, twofold, right? Which is like mostly the AI discovery thing, I was tracking Anthropic Glasswing.

You know, 'cause like naturally it's like, okay, things that have Anthropic Glasswing or the researchers are clearly associated with AI-assisted.

There's a larger bucket of AI-assisted which Berkeley has done a really good job at like collecting.

There's a little bit of mix in there of like AI products in AI-discovered things.

Those are two different things, you know.

... you know, I know you all at VulnCheck have disclosed something like 1,600 or more vulnerabilities across hundreds of open source projects, and only a small fraction of those have been patched, which seems similar to what, you know, we saw from that Glasswing, um, uh, diagram that we looked at, right? Uh, given that reality of, of the remediation bottleneck, um, and, and that, you know, now AI can generate disclosures at scale, does that 90-day disclosure model still hold up, or does it have to change when exploitation is already happening in the wild? How do you think of that?