Using AI as a Grasshopper Tutor

Rhino AI
1

I’m relatively new to Grasshopper and have a lot to learn. My approach is watching YouTube tutorials, and when I don’t understand something I try to track it down until I get it. Often this leads to side quests and experimentation, which leads to more side quests and more experimentation.

For this, I heavily use Gemini 3 and ChatGPT 5.1 as tutors. They are valuable resources but also far from perfect. Both hallucinate components quite a lot, and when I screenshot my GH canvas they often misread the wiring or give advice that doesn’t make sense or isn’t on point. I usually only find that out after spending a lot of time following their nonsense advice—which is frustrating, yet it can still be somewhat useful.

As input, I mostly use speech (Gemini 3 produces awful transcripts; ChatGPT 5.1 is much better), screenshots, or even little videos. Grasshopper can export an image of the entire canvas, but I found those are too large and get reduced to a blurry, unreadable mess by the AI.

What are your experiences using AI most productively? Which models are best? What are the best methods to feed it information (screenshots, videos) ? Is it possible to export “some code” representing the definition to feed the AI instead of screenshots?

2

In my opinion this is a chicken-egg problem. Answers are often very convincing but can be wrong. Entirely wrong. The easier a problem is, and the better terms you use to describe your problem, the better answers you get. So to get into a new technology, AI can help you to master the initial learning curve. But never blindly trust it, just as you shouldn’t trust anything you read when googling. But this requires some knowledge on the topic beforehand. A funny question you could as is: How many e’s are in awesome? And you get wild answers (It told me one)