Hi Ian. Can you tell me how do you determine the url to download the datasets? I don't see any reference in the docs to *.json files only ndjson (raw and simplified), .bin and .npy. Perhaps you manually processed the ndjson data offline and then you pushed it to the cloud? I just want to be able to generate a N set of json drawings pick randomly from all the categories. I will probably have to build a pipeline offline to process the ndjson and push the results to the cloud and then work off of that from observable. Any advice is welcome. Cheers! P.S: Miss your Coding with Fire videos.

I did do offline ndjson processing to get these sets. you can see a bit more in this notebook: https://observablehq.com/@enjalot/t-sne-grid-selector i didn't process all the words for this though. using the ndjson commandline tool you can pretty quickly come up with a makefile or something to pull out random samples from each word. and you can download the ndjson from the google github repo. would love to see what you come up with, this is my favorite dataset :) I miss streaming too! reach out via twitter DM if you want to chat more!