Creative Consolidator: DoomLearn

Inspiration

Our project began with a metaphor: The Corporate Boardroom. We wanted to recreate the dynamic of a high-stakes team meeting where a Boss calls a meeting to build a master plan. The Boss doesn't work alone; he relies on his Secretary to find the "best of the best." The Secretary then calls in a specialized team of Experts to "grill" the plan, find flaws, and refine it until it's perfect. We realized that this rigorous, multi-agent consensus model was the perfect engine to solve a modern problem: the chaos of doomscrolling. We wanted to take the friction and addictive nature of short-form content and transform it into a state of highly productive "Learning Flow."

What it does

With our multi-agent network established, we gave the Boss (our Generator Agent) a high-stakes prompt: "Create a proposal for a 'doomscrolling to learning' app, where given a prompt, the website will instantly spin out a structured, gamified learning path." The result of that agent-led proposal is DoomLearn. DoomLearn is a web application that takes any user-provided topic and instantly generates an interactive, infinite-scroll feed of microlearning cards. Instead of mindless scrolling, users scroll through highly engaging, bite-sized lessons, quizzes, and warnings. Every card features an auto-playing, human-sounding AI voiceover paired perfectly with karaoke-style text highlighting, making the learning experience both effortless and immersive.

How we built it

To bring this agent-led proposal to life, we integrated an advanced stack of sponsor technologies:

The Brain (Fetch.ai & Agentverse): We utilized Fetch.ai's ASI-1 for high-level reasoning. Using OmegaClaw Skill Design, we engineered the communication protocols between our Coordinator Agent (The Secretary) and Generator Agent (The Boss). We then pulled in pre-trained Agentverse agents to serve as our "Experts," ensuring every curriculum is vetted by a committee of intelligences before it reaches the user.
The Frontend (Cloudinary): We built a modern React (Vite) interface bootstrapped using the Cloudinary React AI Starter Kit. Because high-fidelity audio assets are heavy, our backend automatically pushes generated media to Cloudinary, allowing our React frontend to leverage Cloudinary’s high-speed edge delivery so the TikTok-style scroll never buffers.
The Voice (ElevenLabs): We integrated ElevenLabs' high-fidelity text-to-speech. We utilized their advanced /with-timestamps API endpoint to perfectly synchronize the professional-grade AI audio with our dynamic, on-screen text highlighting.
The Security (Auth0): We secured the platform using Auth0 (@auth0/auth0-react). This ensures every user's progress, XP, and custom-generated library is protected by production-grade identity management, with our Django backend asynchronously validating the Auth0 JWTs. To score and validate the learning outputs from our agents, we used a weighted consensus mathematical model: $$ \mathcal{C} = \sum_{i=1}^{n} w_{i} \cdot \text{Expert_Score}_{i} $$ ## Challenges we ran into Coordinating the "Secretary" and the "Boss" in real-time was our biggest hurdle. Managing the asynchronous communication between the OmegaClaw-designed skills and the pre-trained Agentverse agents required us to dive deep into Fetch.ai protocols and handle complex network timeouts. Additionally, perfectly syncing the React frontend animations with the word-level timestamp arrays returned by the ElevenLabs API required meticulous state management to ensure the "karaoke" highlighting felt natural. ## Accomplishments that we're proud of We are incredibly proud of bridging the gap between a complex backend architecture (a multi-agent AI network communicating autonomously) and a frictionless, consumer-friendly frontend. Seeing the Fetch.ai agents successfully "grill" a curriculum, resulting in an ElevenLabs-narrated, Cloudinary-hosted learning card delivered instantly to a secure Auth0 user session was a massive milestone. ## What we learned We learned that when AI agents are forced to "grill" each other, the resulting output is far superior to any single LLM prompt—it’s the difference between a rough draft and a highly-polished board proposal. We also gained deep, hands-on experience in orchestrating modern, decoupled infrastructure by wiring together Fetch.ai, ElevenLabs, Cloudinary, and Auth0 into a single cohesive pipeline. ## What's next for DoomLearn Our next step is expanding the Agentverse integration to include specialized agents that can pull real-time data (like current events or stock market trends) to generate daily, personalized DoomLearn feeds. We also plan to introduce multiplayer challenges where users can race against each other through the same microlearning curriculum!