> human.json is a lightweight protocol for humans to assert authorship of their site content and vouch for the humanity of others. It uses URL ownership as identity, and trust propagates through a crawlable web of vouches between sites.
This will not (and shouldn't) be used by more than a handful of people who were likely already friends anyway. I can't see it being helpful for anybody (unless accidentally visiting LLM blogspam melts your face à la Raiders of the Lost Ark) unless it's true intention is signalling you don't like LLMs to other people who don't like LLMs.
Virtue signaling at best; noise at worst… It’s trivial for an AI to add, and will be done so by anyone hoping to get a piece of that attention economy…
I think I saw Gaius Baltar implement this on Battlestar Galactica. It went well. /s Honestly seems more like a protocol for encoding a popularity contest, which is already what social media signalling does. How do you defend against self-reinforcing botnets and bad actors "cancelling" other people? I can dilute your human signal by creating massive amounts of LLM-generated noise.
If nothing else, this at least inspired me to put a disclaimer on my own site declaring my AI policy. It's not so fancy and I think it's a good deal more credible than any formal protocol.
If you have to perform a breadth-first search from your "seed" to verify a website, wouldn't every lookup become expensive relatively quickly? Unless max hops is set really low. Id assume you really need mass adoption for 5 degrees of separation to kick in, and that's still a lot of sites to crawl!
The fact that this won't go "web scale" seems to be its strength. The idea of local/human/authentic trust ecosystems is super powerful. "Proof of personhood" is fraught with issues, but it seems that lightweight trust algos like this do a nice job of treating trust as a human-first emergent thing, rather than trying to be a PKI style "infrastructure". Pretty cool!
13 comments
[ 0.17 ms ] story [ 25.1 ms ] threadThis will not (and shouldn't) be used by more than a handful of people who were likely already friends anyway. I can't see it being helpful for anybody (unless accidentally visiting LLM blogspam melts your face à la Raiders of the Lost Ark) unless it's true intention is signalling you don't like LLMs to other people who don't like LLMs.
it's hilarious that the human.json protocol to fight AI slop is itself AI slop