I wonder how much smaller it could get with some compression. You could probably encode "This website hijacks the scrollbar and I don't like it" comments into just a few bits.

▲

Rendello 14 hours ago | parent | next [-]

The hard-coded dictionary wouldn't be much stranger than Brotli's:

https://news.ycombinator.com/item?id=27160590

	▲	maxbond 7 hours ago \| parent [-]
		You can use a BPE variant like SentencePiece to identify these patterns rather than hard coding them.

▲

jacquesm 14 hours ago | parent | prev | next [-]

That's at least 45%, then you can leave out all of my comments and you're left with only 5!

▲

hamburglar 10 hours ago | parent | prev | next [-]

It might be a neat experiment to use ai to produce canonicalized paraphrasings of HN arguments so they could be compared directly and compress well.

▲

rossant 6 hours ago | parent | prev [-]

Guilty.