Whereas perhaps documenting a pipeline is not how do you feel about those two different axes? How much of that is a challenge?

Yeah, I think all of them tend to be challenges, right? Let's take the sort of the depth of complexity.

So look over the last, since the time that we launched SnapGPT, many new capabilities have emerged on the scene.

Let's just talk about MCP, right? An enabler, if you will, within the agent's capabilities.

So within our platform, we have the ability not only to build agents, but also to be able to take existing pipelines and expose them via MCP as a callable tool.

So now bringing it back full circle into SnapGPT and the creation capability.

So we've added far more capabilities on the platform that we now also need to be able to generate through natural language.

Is this fast mode? Is it deep thinking mode, et cetera? How did the team experiment with working those? And what unexpected consequences did that create within the deeper design? Because I think there are some pretty fundamental architectural changes that go along with those two specific needs to improve generation quality, isn't it?