I doubt it, why would it?
Apologies, I misunderstood. I thought you meant silent as in the kind that doesn’t send a notification when a post is linked (?silent=true).
So… what does ‘silent’ mean in this case? Thanks.
I am unsure about what the expected behavior is. I thought the way the summary bot posted the link isn’t expected to work.
https://meta.discourse.org/t/configuring-the-default-sidebar-for-your-users/347764?silent=true#p-1690272-handling-plugin-added-links-5
Above is the link it posted. I think it would have been silent if it was
https://meta.discourse.org/t/configuring-the-default-sidebar-for-your-users/347764#p-1690272-handling-plugin-added-links-5?silent=true
instead.
I think there are two possible solutions: the bot learns to use the correct format, or Discourse mutes the format that the bot uses.
Edit: I found the other topic about silent and anchor. When I combine this with my observations, it seems that either silent or anchor works, but never both. Silent link seems can't handle #anchor links
1 like
I can fix silent+anchor, it is a bug, will bookmark for tomorrow
1 like
I think there is now a setting to exclude categories
2 likes
Site feedback > Forum summaries is now in the linkback ignored categories, any others I should add?
1 like
I have recently noticed a couple of issues with Daily Summery Tooic
- I have noticed several summaries have posted as a copy block. Example SS Post
- In Today’s it identifies a Plugin I posted as a Theme Component Screenshot post
1 like
I’ve caught it a few times citing something I didn’t exactly say or do, but I chalk that up to how an LLM works.
The format seems to change frequently – I don’t know if it’s rotating through format options or making them up on the fly. Some are better than others but I agree the code block layout in particular isn’t very useful.
1 like
Gemini 3 Flash hallucinated some random tag here which is super annoying, we need to guide it a bit better, will have a look at the prompt and see if we can coerce it not to make this mistake in future.
1 like
I discovered this feature a few days ago and thought it was great! But today I was very surprised to read this in yesterday’s summary. I may have opened the topic in question (vaguely rings a bell) but it seems a little hallucinatory 
The first time I read through a daily summary I initially thought it had been human-generated for a bit (the fact I arrived there through a notification saying @sam had mentioned me in it did prime me too, though).
(For the record: I am overall rather good at spotting LLM content, in general, so I’m surprised I was « fooled » for so long!)
I do like the feel it gives and can definitely see how this could be super useful for my support community (as long as it doesn’t hallucinate too much…)
Is it useful in some way to keep noting occasions when the AI summary is especially wrong?
This one rather overstates my contribution to the topic:
1 like