For example, just the blockchain data structure (not a full cryptocurrency, just the data structure) is relatively small and doable. Maybe a tiny perceptron or autoencoder. But there's a lot of opportunity to make something with even such small toys.
And don't be afraid of making something that isn't a "true" or "real" whatever. Yeah, it's not a "true" blockchain but it was toy to test your chops on.
Also, whatever you do end up doing, regardless of whether you finish it or not, write about it. Just a paragraph or two of what you were trying to do, how far you got, what made you lose interest. Informally as possible.