This is not an anti-AI rant. If a future AI agent truly has high quality posts and wants to use the site normally, that's fine. I'm talking about spam campaigns with hundreds of new accounts. We need new solutions to this problem.
I'll start by proposing a solution that could work for HN and similar forums. Feel free to iterate on it or propose your different ideas in the comments. Here goes:
For logged-in users, instead of ranking posts and comments on the server-side, the server only delivers a chronological feed + the current logged-in user's voting history.
Using the chronological feed as the base, each of your past votes changes the ranking of your feed by a tiny bit, and that's calculated client-side. You're more likely to see posts and comments from users you've upvoted in the past at the top.
In short, this means a new account will see a completely chronological feed, while an established account will see a feed modified by only their own past votes.
The public feed for non-logged-in users would still be ranked by the server. No changes there.
So each user gets a fully personalized bubble when logged in, except it's not a bubble because n=1. And it's really easy to break out of the bubble by logging out.
Spam bots can post and vote all they want, but they won't change the core userbase's experience that much, because the bots will only have access to a chronological feed. It has no taste, which is accumulated over time, and therefore can't spam votes and replies on real conversations nearly as much.
Another interesting idea that comes to mind is that every post/comment made needs the user to physically use their fingerprint scanner on their device which I assume plenty of devices have already. As long ad it can't be spoofed it works but not sure about the details about reliably securing that.
It would be some friction but I feel like it would be fine?
Without transferrable credibility, any ratings system simply becomes a question of which side spams the most.
Rate limits tied to behavioral patterns rather than identity seem to work better — especially interaction timing, navigation flow, or session consistency.
We experimented with something similar while building HiveHQ and found bots usually fail when systems require small contextual actions humans do naturally.
Interaction timing is like rate limiting, but more granular
Navigation flow is a basically requiring bots to use a headless browser instead of API's
What does session consistency mean in this context? Restricting to a limited number of interests & activity times?
The question you need to ask yourself is "What's the end game?"
What happens when users' feeds are full of users that they already know?
You think they'll be satisfied with that?
Bots and AI right now as good as the "average" joe.
All the places that can move the perception of real people about products, politics or any form of power will and is being flooded with bots.
The reason for the "ID" on internet is not because of the children. But because the bots are soo good they need to use ID to filter what is bot or not. Avoiding the dead internet.
The powers that be NEED to sway perception and narrative to their liking.
think about the children! Epstein list, patriot act, etc.