Oddbean new post about login | logout A simple web of trust can solve the bot problem, eg. by connecting friends of friends. We can establish multiple degrees of connections and prioritize them.
As long as I can see people posting with hashtags that I follow even if they are not friends of friends.
You pick an N connections. After 3-4 hops that probably encompasses anyone you would want to hear from.
I suppose that could work. I'm just still a bit squiffy on the proposals I've seen.
I am skeptical of more complex algorithms. But one that simply varifies that the account is "in network" is good enough to stop spam. If one of your friends or friend's friends etc is an idiot and keeps friending bots, keep that friend but ignore their contacts. Spam vector closed.