Falcon: reading the Grasshopper definition
Our working research prototype explores how an assistant can understand the definition a designer is already using.
Writing a script is only one part of computational design. A Grasshopper definition also carries relationships between components, data structure and project assumptions. Assistance becomes more useful when it can address that working context.
Work with the existing definition
With Falcon, we are exploring how to understand definitions, review parametric logic and check a proposed change. The question is how the assistant can help a designer continue the work already on the canvas.
A useful response should identify the inputs it understood, explain the operation it proposes and make uncertainty visible. A suggestion to change spacing, for example, should account for the guide geometry and the grouping that downstream components expect.
Make a change inspectable
For a small trial, we would keep the original definition and choose one bounded adjustment. Compare the outputs before and after, including an input near the permitted limit. The designer should be able to explain why the change works and undo it when it does not.
General models can assist with code and reasoning; a useful connection to a design environment also needs explicit tools, context and validation. We treat that connection as the research problem.
Falcon’s place at Fraktal
Falcon remains ongoing research. FraktalMCP is our separate published Grasshopper plugin, and Archly is our active visualization product. The current Falcon page describes its scope. This note records the question guiding the prototype, rather than promising a release schedule.