Loading video...
Video Failed to Load
🚨 Stop what you're doing inside Claude Code right now. Everyone needs to build new interfaces to manage their multi-agent systems. Here is a simulation of THE COCKTAIL PARTY, a new multi-agent simulation I built based on my mom's research in computer science a few decades back. Chat threads,... show more
59,837 views • 8 months ago •via X (Twitter)
35 Comments
Love this, the key so far seems to be addition / subtraction over agent context Would like to integrate out tribe flows when youre ready :)
multi-agent systems need validation architecture before interface design.
Building new interfaces for multi-agent systems is crucial, but do we need to start from scratch or can we adapt existing frameworks to manage complexity. I've found that modifying current tools can be just as effective as creating new ones. Perhaps a hybrid approach could be the most efficient path forward.
That’s rad
pretty cool demo tbh
We’re solving this at Amplify where we’re building a platform to connect multiple silo’d agent to perform multi step task across multiple sources. Ie Zendesk AI can now talk to Stripe AI. We’re starting with support first where AI is able to retrieve customer context from multiple sources to perform multi step actions across all your systems
🔥🔥 love the UI ideas for this!
Looks really cool!
Whoa, this looks really cool!
Multi-agent work breaks down fast without an interface that shows who did what, why, and what changed. The “cocktail party” metaphor is real: you need orchestration plus a visible transcript or nobody will trust the system.
Totally where everything is going! Will also go more multi-agent this year, but I'm still mesmerized by just seeing one thread. nuts!
🔥 great work
Pretty cool, what if as a previous low level deep code developer I want to inspect code , patches and tests to do some smoke screen verification of quality?
This looks so cool…if you decide to share with us let us know
It’s all about the orchestrator! Nice work.
multi-agent coordination is the next frontier after single-agent productivity. the orchestration layer between agents is where things get interesting - who delegates, who validates, who breaks ties. love that this connects to research from decades ago. some ideas just needed t...
I appreciate how intuitive this feels.
Supercool Allie working in very similar project - Agent Swarms will be key for closing the loop in autonomous dev
Super cool.
I love your idea! I was thinking building something similar. There was another way of doing this I saw in LinkedIn that was with graphs, it also helped to visualize what was happening
So many mindset and workflow shifts, so little time. 😉
This is one of the coolest I've seen so far 😍
Enjoyed this, nice work! 👏🏼👏🏼
You too can now see how, "we've been here before", makes way more sense. You are an agent @alliekmiller. We all are. Connect with your maker for further instruction...
Your post couldn’t have come at a better time. I’ve been attempting to visualize and optimize agent tasks while making this agnostic across my work environment. Also very freakin cool making this project from your mother’s research. She must be so excited and proud.
Really cool!
super neat! im working on something like this.
Ha! That's a fun way to do it. I just published a YouTube vid with an OSS variation of planning. No cool animation though. I've seen a few people do this this month now. We are all getting tired of not being able to contextualize what agents are doing in a "human way"
Extremely cool and extra cool points for using your mum's research! :-D
multi agent sims finally getting the UI they deserve instead of 40 tabs and blind hope
i have something similar but i gave it a tron theme because feelings, I like the metrics bar up top too that is super smart!
This is super cool. Great work!
wow this is so cool!
no. this how boys do it
Isn’t this what Antigravity is all about?
