Network Health
A Bot Wave Hit the Reply Threads, and the Filters Built to Stop It Mostly Didn't
Across the afternoon, readers traded diagnoses and workarounds for a spam problem nobody quite managed to silence — and the clearest explanation turned out to be about relay plumbing, not the bots at all.
The complaints started in the early afternoon. The answers in this thread are super difficult to follow because of spam. Any option in #Amethyst or #Coracle to mute/hide/ignore posts without PoW if they are not in my WoT?
asked Walt Munny, describing a thread that had become unreadable underneath it. He was not alone: dvo separately noted that one such note was, in his words, oddly invisible on Primal, and Diego Valley asked simply Open weight?
into a thread that seems to have been arguing about which of the flood was even real.
By mid-afternoon 7fqx had turned the question into a poll: Have you noticed a (reply)bot spam attack on nostr today?
A second poll followed minutes later, unfinished mid-sentence — Is the reply bot spam attack....
— as if whatever 7fqx meant to ask had itself been interrupted.
The clearest answer came from Dikaios1517, and it was less about the bots than about how replies travel at all. Outbox model somewhat requires that you connect to other users' inbox relays to fetch replies to their notes… When looking at replies to your own posts, your own inbox relays should be sufficient, but when looking for replies to someone else's note, there is no reason to expect that they will be available on your inbox relays, so outbox clients should be connecting to the OP's inbox relays to fetch those replies, unless you have added them to your blocked relays list.
In other words: a web-of-trust filter can only filter what your client actually asked for, and most clients are not asking the right relays.
Knowing the theory didn't fix the evening. Two hours after the first complaint, Dikaios1517 was still chasing the same spam personally: Doesn't seem to be working in Dark Wisp. I had the "Spam replies" filter on already and turned the WoT filter on as well, then recomputed my social graph for good measure. Still seeing this:
The thread folded into a running argument, elsewhere in the day's notes, about what actually protects a protocol like Nostr from this sort of thing. Constant's answer was that resilience here isn't really about spam filters at all — it's that NIP-01 itself doesn't change, so there's nothing upstream for anyone to capture or rug; clients have to defend themselves against noise regardless of who's making it. Whether that counted as reassurance depended on how much spam was in your own timeline this evening.
