Does anyone know why when I do catalog.load on a versioned PickleDataset it is not loading the latest version but the third one of five?

Dumb question—how do you know it's loading the third one?

I pickle loaded them directly in my notebook one by one

There's no reason this should be happening. That said, I suppose you could try modifying your Kedro installation to spit out the filepath it's loading from, just to confirm. (Not 100% trivial, but if you want to investigate, I can suggest how to do it.)

Did you keep the same catalog while having many kedro runs? Does the result change after you run

%reload_kedro

?

It was a bad mix of different datasets having similar structures + I was reading the name of the function used inside the node (in bold pink) instead of the name of the node, my fault. I would recommend highlighting more the name of the node. Thank you for your patience, I like Kedro a lot!

Gotcha, I have similar thoughts after we introduce the %load_node magic, there are some complexity because not all nodes has a custom name and the auto generated one is not human readable. One more thing for us to consider.