On Prototyping

This is a developing thought so I might come back and change things but fundamentally: the thought is that at the moment Figma prototypes with all their bells and whistles of variables, logic and overlays can only prototype deterministic systems. Turning figma into a turing machine ensures that (I might be sketchy on thie theory. REVISIT) if sufficiently complex, it can give a good enough approximation of most modern software products. This can be a largely design(er) driven exercise.

The case for making only Figma prototypes breaks when you see that Figma's prototyping system is inherently incapable of representing and capturing the behaviours of probablistic systems(like ML models, LLMs, CV models etc). At best it can capture one state out of a myriad. This severely limits the ability of a designer to effectively model and design the behaviours of such a system. That limits their effectiveness as a communication/bridge builder between engineers and designers.

It is easy to create (on paper) great experiences that AI systems wont be able to reliably adhere to. Solving the AI adherence/alignment problem is out of scope for this discussion but the argument is that there is a lot of potential benefit in using quickly codeable "sketches" that give designers and engineers an artefact to anchor around.

HOLY SHIT I JUST SAW LANGGRAPH STUDIO

This changes everything. being able to 'play' with agentic architecture itself opens up avenues for problem solving beyond what a "Figma prototype with API calls" could achieve. Ik, you were not the only one thinking about supercharging figma protoypes with Plugin led API access to AI.