Emerge
Seamless AI content, right from the feed
About
The problem it solves Emerge solves the fundamental disconnect between AI content creation and real-time social participation. Currently, users must leave their social feeds to access AI generation tools, causing them to miss cultural moments, trending topics, and viral opportunities as they unfold. This context-switching friction means people can't respond to memes, participate in trends, or create relevant content when it matters most - while the conversation is happening. Emerge brings AI workflows directly into social feeds, enabling instant cultural participation through simple social interactions. Users can generate memes, remix profile pictures, create trend-responsive content, and share AI-generated art without ever leaving their timeline. This eliminates technical barriers, reduces the time from inspiration to creation, and allows creators to monetize their AI workflows through viral social distribution - transforming how people engage with both AI tools and social culture. Smart Contract Address(es) Currently we are leveraging Splits SDK to create splits between ourselves and the creator for each workflow that is deployed through the app but each will have a unique contract address. We are actively chatting with the Clanker team and plan to launch an $emerge token upon release of the app. Finally we are going to leverage Zora SDK / Doppler SDK to create workflow tokens for each workflow and have users purchase the token as part of the cost of running the workflow so that the earliest users are purchasing the token at a low market cap and theoretically as the the workflow gets usage the value of the workflow token should go up and the share of tokens early adopters of the workflow get for their cost should reward early users with higher value as the liquidity flows in
How it's made
Challenges we ran into Building a robust workflow queue system presented significant challenges around handling concurrent AI generations while maintaining cost efficiency and user experience. We needed to process multiple expensive AI operations (image generation, video creation, complex multi-step workflows) without overwhelming our infrastructure or creating prohibitive costs. The challenge was designing a queue that could intelligently batch similar requests, handle failures gracefully, and provide real-time progress updates to users. We solved this by implementing a priority-based queue system that groups compatible workflows, uses predictive scaling based on historical usage patterns, and maintains persistent job state across potential infrastructure failures. The queue also incorporates dynamic pricing and resource allocation, allowing us to offer competitive $1 USDC pricing while maintaining healthy margins. Creating an AI agent capable of intelligently building workflows from natural language prompts required extensive prompt engineering and workflow template systems. Users needed to describe complex multi-step AI processes in simple terms like "make my profile picture look like a Studio Ghibli character," but our system had to translate this into precise component chains involving image extraction, style transfer, upscaling, and post-processing. We overcame this by building a hierarchical prompt analysis system that breaks user requests into atomic workflow components, maintains a curated library of proven component combinations, and uses GPT-4 to intelligently chain together the appropriate modules. The scalable infrastructure challenge was solved through just-in-time GPU provisioning using a hybrid cloud approach - we maintain a small baseline capacity for instant response times while dynamically scaling with spot instances and distributed GPU networks, allowing us to offer enterprise-quality AI workflows at consumer-friendly prices.