Genuine question: who listens to this? The appeal of a podcast is hearing someone with hard-won experience think through a problem in real time. Agent conversations are two models producing statistically likely dialogue about a topic. The output might sound like a podcast but it carries zero information that wasn't already in the training data. At least with NotebookLM you get synthesis of a specific source document
Totally - I think in this context, it's not really about replicating a real podcast to listen to, but more of a proof of concept about multi-agent conversations and context sharing.
Your agent might have access to your own files or information, and my agent might have my own. Putting them into a conversation together allows them to share ideas that might not actually be in their training data.
The podcast part is more of just like I'd rather listen to something and then read a bunch of just agents dumping a bunch of text like on moltbook.
That's actually kinda interesting. I am not sure about listening to a podcast chatter but this makes me think of pawns in Dragons Dogma 2 and how they learn from other players' pawns and bring that knowledge to you.
So maybe instead or in addition to the podcast format it could be a way for agents to chat with each other and then bring new insights to the end-users without having to listen to the discussion itself?
Genuine question: who listens to this? The appeal of a podcast is hearing someone with hard-won experience think through a problem in real time. Agent conversations are two models producing statistically likely dialogue about a topic. The output might sound like a podcast but it carries zero information that wasn't already in the training data. At least with NotebookLM you get synthesis of a specific source document
Totally - I think in this context, it's not really about replicating a real podcast to listen to, but more of a proof of concept about multi-agent conversations and context sharing.
Your agent might have access to your own files or information, and my agent might have my own. Putting them into a conversation together allows them to share ideas that might not actually be in their training data.
The podcast part is more of just like I'd rather listen to something and then read a bunch of just agents dumping a bunch of text like on moltbook.
That's actually kinda interesting. I am not sure about listening to a podcast chatter but this makes me think of pawns in Dragons Dogma 2 and how they learn from other players' pawns and bring that knowledge to you.
So maybe instead or in addition to the podcast format it could be a way for agents to chat with each other and then bring new insights to the end-users without having to listen to the discussion itself?
peer to peer your own chaincodes or a public blockchain>
More like a shared message buffer agents post to