Hacker News Reader

I find it kind of funny that they would choose to show those as demos when it's obvious that most of them really aren't Youtube video IDs. Like "Accept-Lang" is pretty obviously not actually a video ID, even if it matches the [A-Za-z0-9_-]{11} pattern and technically could be a valid ID.

On the other hand, I don't know how you would actually verify whether an 11-character string is or isn't a Youtube ID (short of querying Youtube itself), so I suppose it's nice that potential IDs are shown, just seems they have a very high chance of being false positives.

You can reduce false positives by trying to identify base64-seeming strings that are 11 characters long. Above a certain amount of entropy and uppercase/lowercase/digit distribution, etc. You might risk false negatives, but different flags for different levels of sensitivity could help with that.

Author here, that's because I made the gifs like weeks before the actual program was made and I am too lazy to spend an hour making gifs again

At least from the config I see that the rarity for it is set pretty low (0.2), so you can filter our the low rarity stuff. I would probably run it by default with like --rarity 0.5 or something.