ripple
Ripple turns one blog post into native copy for every channel. After an orchestrator writes a brief, platform writers (TikTok, LinkedIn, Reddit, and the rest) draft at once. A scorer can send a weak draft back.
What it does
Ripple solves the content repurposing bottleneck. Creators and marketing teams write one piece of content, then waste 6+ hours manually reshaping it for TikTok, LinkedIn, Twitter, Instagram, Email, Reddit, SEO, and Video. Ripple takes a single input — text, URL, or PDF — and deploys 17 specialized AI agents that run in true concurrency on the Mozaik v4 runtime to generate platform-native content for all channels simultaneously in ~30 seconds. Each output is quality-scored, brand-checked via a self-correcting loop, and exported as training data through Adaption Labs to continuously improve the system.
How the agents run concurrently
All agents are Mozaik createAgent() participants sharing a single defineRuntime<RippleState>() state object. The pipeline starts sequentially: Signal Extractor parses input, then the Orchestrator runs 5-phase research. Once research completes, the Orchestrator broadcasts a DEPLOY_AGENTS SemanticEvent — and all 8 platform agents (TikTok, LinkedIn, Twitter, Instagram, Email, Reddit, SEO Blog, Video Script) react simultaneously through Mozaik's event-driven SituationHandler. They read from shared state (signals, trends, research) and write their outputs to platform-specific keys concurrently. As each agent completes, it emits AGENT_COMPLETE, which triggers reactive agents (Quality Scoring, Brand Voice Guardian, Consistency, Memory via Mem0, Adaption Exporter) in parallel. If Quality Scoring detects a score below 7, it emits BRAND_VOICE_REVISE targeting the specific platform agent, which self-corrects — forming a closed reactive loop with zero human intervention. All state transitions are streamed to Mozaik Cloud via createCloudExporter() for real-time observability.
