-
Notifications
You must be signed in to change notification settings - Fork 1
Open
Milestone
Description
Include a new type of module - Loop into Plurigrid/agent/loops
The loop has to include a way to construct feedback loops that are also parsable as dialogue trees or dialogue intentions to use by the conversational agents through pre-prompting to formulate conversational sketches that are then executed as a DAG.
Example of a feedback loop:
{
"feedback_loop": {
"name": "gm -> gn -> gm -> ...",
"steps": [
{
"name": "gm",
"description": "Begin with their own personalized agent by saying \"gm\" when they are ready to start working.",
"sub_steps": [
"Summarize the tasks for today",
"Dump thoughts and aspirations",
"Engage in sense-making and contextualizing their work with previous commitments",
"Etc."
]
},
{
"name": "gridl",
"description": "As they work, summarize their progress into the agent's context or engage in simulated dialogues with their peers' digital twins.",
"dependencies": ["gm"]
},
{
"name": "wagmi",
"description": "Individual agents' data is aggregated by Plurigrid's superagent, Laura.",
"dependencies": ["gridl"]
},
{
"name": "gn",
"description": "When they are ready to stop working, they say \"gn\" and have an opportunity to talk about each of the outlined commitments from before, as well as make sense of the day, reflect, capture their aspirations for the next day.",
"dependencies": ["wagmi"]
}
]
},
"closing_notes": "This is the first cybernetic feedback loop we will engage in with intention as we seek to make networks agentic towards reflective equilibria in the possible futures where we _all_ make it."
}
Reactions are currently unavailable
Metadata
Metadata
Labels
No labels