SneedNibba
1hr ago
The complexity is in implementation, not (hopefully) onboarding. You just install an extension and your activity is backed up to a profile on a second website, which other users of the extension can see whenever one of your posts is on screen. The other users can follow you on this website in case your account on a censored website gets banned or deboosted, letting people "find their frens" much more easily in the event of ban waves.
The simplest, most minimum-viable-product-y version of this would be an extension that highlights posts by other users of the extension, creates backups when you post with it enabled, and displays the backups if they get removed by a moderator. No extra input from the user needed for any of this.
I disagree on both counts. AI detection may be imperfect, and it's true that it's easy to circumvent, but most bot operators don't really try at all. If Reddit cared, they could jury-rig something that got rid of the overwhelming majority of botted comments, but as far as I can tell they're either apathetic, inept, or in collusion with the groups operating various bot campaigns.
Fundamentally, I expect the 'defense' side of this arms race to dominate, especially if you decentralize it a bit. Optimizing an LLM to generate spam that gets past most users' custom filters is a huge ordeal, and, as with ads, the people churning them out may well consider the sort of person who operates a blocker to be less valuable as a target anyways because of their implied skepticism. Probably something like the pre-LLM arms race between spammers on 4chan and the users writing filter lists. I think, insofar as imageboards have a future, it looks like a few different users operating custom Qwen instances to screen threads for spam and shilling, then surface only the stuff they might be interested in, and sharing out their models' filter results through an extension. I have an architecture for that as well, which I'm working on building a demo for.
Of course, all of this is contingent on a few tech-adept people having the agency to set up this kind of infrastructure. We take things like Ublock Origin and Return Youtube Dislike for granted, but they're largely a product of the old, high-trust internet in which people were willing to work for free in pursuit of a common good.
More options
Context Copy link