Here's the link to the post:
I2P: http://hkpahyrrlq2fuxjlmcnpf7e47jdnciy7ageciehhyclu2yld7xha.b32.i2p/res/7.html
Tor: http://mnj2jsmwbgzhbccunvcyw7frpinmquqfazfwlc4jqrczbjs74m5lg7ad.onion/res/7.html
I got quite a few responses considering how small the userbase (likely) is.
The summary of it is, that it uses Javascript, therefore honeypot and that it's too similar to the Fediverse (even though the post explains it's not lol, I copied the post from Lemmy).
I've tested the friendly captcha demo and it is indeed way more fun and easier than the model we have now on plebbit.
I have a 12 year old 4 core thinkpad. It takes around 10 second to verify you are "human", although you cannot be human with POW captchas indeed. As with traditionnal captchas which can be solved with machine learning. But posting a vast amount of bot posts with POW captchas would take a vast amount of CPU power and electricity consumption.
Plebscan