Hacker News is relatively positive compared to, say, Mastodon or Twitter or many similar things.
I note he's using a sentiment model from NLTK which is tuned up for short texts and might work OK for this task although my take is that you have to train sentiment models on specific domains and I don't know what they trained VADER on. I've seen really good data sets for product reviews and movie reviews but these aren't necessarily going to be effective for "angry toots". If somebody felt there was too much negative content on HN they'd get better results training it on what they think negative is but it's a substantial amount of work, a good training set would have about 5000 positive and 5000 negative examples and would take about an FTE week to make with the caveat that looking at 1000s of negative content items in the social media space can take a toll on your mental health.
dont think so