It is still difficult to tell what exactly she's replying to, and if she really is or isn't skipping prioritized ones
#Show chat messages + sender on screen
lean oasis
young basalt
maybe give a little pop up when she answering chat messages
rain rivet
This assumes that the system randomly picks a single comment and presents it to the LLM.
Perhaps the system samples 20 comments every generation cycle, and the LLM might decide to refer to those comments or not. Showing 1 comment would be helpful, but 20 comments that might not even be used, would just be confusing.
In order to substantiate this theory, we'd need to look for when Neuro clearly refers to multiple messages in the same generation
prisma elm
This assumes that the system randomly picks a single comment and presents it to ...
it does not
rain rivet
it does not
Ok i guess you could still manage to show what Neuro is responding to, regardless of how the chat is inputted.
Maybe have a fast model (something something vector cosine similarity?) that quickly identifies which chat message is probably related to Neuro's comment. If there's precisely one comment, show it on screen. If multiple or none appear to be related, give up.
prisma elm
I could
rain rivet
ah i misinterpreted your first message
yeah i'm not sure if the ROI on that feature is worth it
🤔 maybe it could be good for a devstream? It's AI related but still relatively separate from the rest of Neuro, especially if you create a mock environment thing that hides Neuro's interfaces.
Though it doesn't seem like there are any gaps that need to be filled in dev streams....
prisma elm
it doesn't seem like there are any gaps that need to be filled in dev streams....
is this like
actually how it seems or are u trolling me
because every dev stream I am scrambling for content
rain rivet
i am easily fooled
prisma elm
ill take it
hot tangle
I'm learning a lot...
that's what I meant
hot tangle
oh so you meant highlights instead of normal msgs 