Playing with ChatGPT Code Interpeter - Visualizing HackerNews

This evening I spent some time playing around with ChatGPT code interpeter: https://chat.openai.com/share/35a90a26-1073-492b-b90c-8bdb1710e2e1

I scraped all HN posts from January 1st 2023 to today, ingested them into an sqlite db, and then had code interpeter visualize. They’ve got a neat firebase API endpoint where you can perform this type of scraping: https://github.com/HackerNews/API

I was able to upload a 250MB sqlite file, and it was able to query things with english questions, which is kind of neat. However, the sqlite db ended up being 1.25GB in the end, so I just ran the scripts locally after testing on a subset sqlite file.

Generated some heatmaps to visualize when popular stories were posted (normalized heatmap seems legit), as well as a heatmap to show volume per day