When you say constitutionally, you mean like… Anthropic-style? Great. Nice. I figured that’s the…

Where does it live?

And sure, text is also coming from a real person, but when you actually hear someone’s voice, that changes your orientation. So that I also…

If this isn’t the case, then you can move forward immediately. I’m interested in pushing even a little more than that. Like, is there value to be able to do video data or audio data? I think that is, the way people react to the Heal Michigan project is so ...

So I think that is very doable. I think it is good to take an assumption on, you know, this is the structure of a CSV that you can import in.

The compass I use for that is, so I wrote my thesis with a group called Democracy, Development and Rule of Law in college. And I was the only technical, like computer science person in my cohort. What would it have taken for everyone else to be able to use ...

So I mean, it’s important that the pipeline can take in many different forms of data and also overlay these different forms of data. I’d be curious to hear your thoughts on, you know, which feature sets would be most valuable.

From there on, a lot of our immediate work will be around testing, rigor and accuracy. So we will be doing a lot of, like we will publish a bunch of our research on if we work with text data versus if we work with voice data. I’d love to ...

I think there’s a way to build a pipeline that is really agnostic of how that part happens. Yeah, I’d love to hear your thoughts on basically what would come next. I think the, I bet the North Star we align on, you know, summarization is one of the core ...

But yeah, I think having the, I mean, one thing we say a lot at the AI Objectives Institute is chat interface is just what we know so far, but we will come up with better interfaces to talk with language models. We just, that needs more innovation too.