1
00:00:00,180 --> 00:00:04,800
Well, we built Dogwood because we were thinking about an agent policy problem, of how

2
00:00:04,800 --> 00:00:10,320
do you describe, given an agent or an agentic application, so this is an AI agent, this

3
00:00:10,320 --> 00:00:15,760
could be everything from Codex or Claude Code, one of these very open -ended agents,

4
00:00:16,220 --> 00:00:21,320
to a customer service chatbot, to whatever kind of agent you're building, and we want

5
00:00:21,320 --> 00:00:27,940
to be able to describe what that agent can do in a way that is very exact, in a way that

6
00:00:27,940 --> 00:00:34,140
is very crisp, and a way that can be deterministically applied. And so if you think about,

7
00:00:34,260 --> 00:00:38,580
or if you've been in this space, you'll know that folks are doing LLM as a judge type

8
00:00:38,580 --> 00:00:44,200
techniques, they're having models watch agents, having heuristics watch agents, all of

9
00:00:44,200 --> 00:00:49,180
those things are good, valuable techniques, but we wanted to build something that would

10
00:00:49,180 --> 00:00:54,400
allow us to express policy in a way that we can apply it deterministically, in a way

11
00:00:54,400 --> 00:00:59,940
that we can reason about it very keenly, and a way that you can know for sure when you

12
00:00:59,940 --> 00:01:02,940
express something in Dogwood, it has a particular behavior.