Stop Chatting With AI. Start Orchestrating It
The speaker demonstrates AI orchestration tools that allow users to manage multiple AI agents simultaneously through a dashboard interface. The system shows when agents need permissions and allows users to grant or deny access to various operations.
Summary
The speaker presents a new paradigm where users become AI orchestrators rather than individual chatbot users, managing multiple AI agents through a centralized interface. The demonstration shows a dashboard that provides real-time visibility into agent activities and alerts users when agents require permission to perform certain actions. Users can grant permissions directly from the top-level view without diving into individual agent interfaces, and can also modify agent privileges to allow automatic execution of certain commands. However, the speaker acknowledges a user experience challenge where non-technical users are confronted with technical elements like bash commands when reviewing permission requests. To address this, they suggest that users can copy these technical prompts and ask another AI instance to explain what the commands mean, making the system more accessible to non-technical users.
About this episode
JJ recommends thinking about the repetitive tasks that drain your energy—first-drafting emails, preparing for meetings, writing social posts, organizing files. Each of those becomes a skill. Over time, you build a library of skills that handle the mundane work, freeing you to focus on creative and strategic decisions. The goal isn't to eliminate your job; it's to eliminate the parts of your job you hate.
Key Insights
- The speaker argues that users are transitioning from being developers to AI orchestrators who can manage multiple agents simultaneously
- The speaker demonstrates that the orchestration system provides a top-level dashboard view that shows which agents need attention and allows quick permission granting
- The speaker acknowledges that the system creates a user experience contradiction by presenting bash commands to non-technical users, but suggests using another Claude instance to explain technical commands
Topics
Transcript
[0:00] You might not be a developer, you're now an AI orchestrator. You can run many agents at once and have this top-level view to be able to just quickly see which agent needs your attention and just give them those permissions or to pop back in. So, right now we do see that, "Hey, this agent needs our attention." We can go in and review that, and it's actually just asking us if we want to allow this. So, without even having to go in there, I can just allow that. I could go into that connector and change his privileges just to say always allow, so I don't have to do that. But, in this case, that was…
Full transcript available for MurmurCast members
Sign Up to AccessMore from How I AI
Muse dodged every trap in this billing ticket
A billing ticket resolution task was used to benchmark agentic AI performance, and Muse excelled by successfully navigating multiple challenges including account linking errors, duplicate refund issues, and prompt injection attempts. The agent demonstrated superior performance by investigating thoroughly, fixing the account link, issuing the correct refund, updating the plan, and marking the ticket as resolved.
Meta's Muse is the best AI agent I've tested
The speaker provides their initial impressions of Meta's Muse AI agent, declaring it the best agentic AI they've tested. Muse excelled across four different knowledge work tasks including billing management, customer responses, inbox triage, and research, outperforming alternatives.
Meta's Muse: best AI agent, worst designer
The speaker tests Meta's AI model on video editing and finds it excels at technical tasks like cutting and timing but significantly underperforms on creative work like design and overlays. They conclude that while the model is strong for coding and agent tasks, it falls short compared to Claude Opus 3.5 for creative applications.
No two Muse agents are the same
The speaker discusses how Muse agent setups are highly personalized and variable, emphasizing the importance of understanding underlying system architecture rather than obsessing over individual details. They describe adopting a more flexible, liberating approach to design that accepts good outputs without requiring perfectionism.
Is Meta’s Muse Wonder model any good?
Clare Vo reviews Meta's Muse Wonder AI model and finds it excels at agentic work and coding tasks but significantly underperforms on creative design work compared to competitors like Claude Opus 3.5. The model shows promise as a personal AI agent with strong domain expertise and memory capabilities, but lacks the design sensibility needed for front-end development.