A small experiment in learning from people
Learn to fold a paper airplane
Your folds teach the model.
Use rectangular Letter paper. Photograph the whole sheet on a contrasting surface before each action. Keep the nose pointing away from you. Allow 5–10 minutes.
Take a photo before each action. We pair the image with what you do next.
Photos are used for this demonstration and retained until the presenter manually deletes them. Hands and backgrounds may appear; keep faces and sensitive information out of frame. The presenter can view your photos and display name.
Photo of the blank paper
Take a photo of the paper as it looks now, before folding. Include the whole sheet and leave a little space around it.
Now photograph the paper after this fold. This will be the starting state for the next fold.
Airplane complete.
Your photos are now demonstrations for the model. Thanks for contributing!
Try a prediction
Take a photo of a paper airplane state to see which fold the active model predicts next. Your photo is used for this prediction and is not added to the demonstrations.
Presenter access
Experiment dashboard
Learn from demonstrations
Hold out a whole participant to measure generalization. Every action needs at least two training photos after the split.
Model history
Rehearse with real photos, train a model, then mark the verified version as your fallback. Activating it is always manual.
What would you do next?
Show the model an unfinished airplane. It only knows the configured actions; it cannot detect that you are finished.
These are model probabilities, not guarantees of correctness. Try a new background or an unfamiliar fold.
Demonstration gallery
After the presentation
Delete participant records, runs and photos manually. Trained models are retained for reuse. Collection closes during cleanup.
Previously issued uploads expire after five minutes. Repeat cleanup after that interval to remove any late uploads.