Whiplash

Demo

About

The problem it solves Marketing and community teams are overwhelmed, juggling: Scattered dashboards that provide data, but not direction They need to stay on trend, on message, and on time—but the tools they rely on are failing them. Current solutions fall short: Third-party tools only support a limited number of platforms, often focusing on posting or scheduling but not strategy. Analytics dashboards show numbers, but lack business context or campaign direction. Marketing agencies are expensive, slow to respond, and demand sensitive data from clients—introducing privacy and trust concerns. Most importantly, none of these tools are built for the unique culture of Web3, where trends spread differently, communities organize asynchronously, and brand identity is deeply tied to online performance. Whiplash is built to address this exact gap. We offer an AI-powered marketing agent that doesn’t just automate posts — it understands the direction of your business, and curates strategy accordingly. By uploading a short brief or company PDF, Whiplash uses Upstage AI’s solar-mini model to analyze your business profile and reframe how marketing data is gathered and interpreted. It then: Scrapes Reddit and other open platforms for trend data Filters content based on sentiment, keywords, and social signal Suggests ready-to-use marketing actions, personalized to your brand Keeps your data decentralized and private using NEAR and IPFS Through wallet-based login and subscription, users retain full control over their data, while enjoying a high-performance marketing stack that evolves with their business. Unlike traditional tools, Whiplash doesn’t ask users to change for Web3. Instead, it shows them a product that is simply better—faster, cheaper, and safer—and lets that value become the onboarding experience. We’re not building another tool. We’re building a survival kit for modern marketing. And in doing so, we help grow the Web3 ecosystem—one brand at a time.

How it's made

Challenges we ran into 1. Optimizing Input Strategy for Upstage AI We found that passing one post at a time to the Upstage AI model—rather than batching multiple posts—led to significantly higher response quality. This change, while improving accuracy, drastically increased the number of API calls (from 1 to over 100). 2. API Call Overload & Speed Issues The high frequency of calls created serious performance bottlenecks. To solve this, we implemented multithreading. Initially, we spun up 50 threads, which made the process very fast—but caused request errors due to exceeding rate limits. 3. Thread Control with Semaphores We refined our approach using a semaphore, limiting concurrent threads to 10. This balanced speed and stability, resolving both the performance and reliability issues. 4. Finding the Sharp Edge of Our Product While we started with a broad goal—automating the workload of community managers—it took time to identify what made our solution uniquely valuable. Through discussions with judges and real-world marketers, we gradually sharpened our focus, developing a Web3-native solution that goes beyond typical Web2 or non-AI tools. 5. Building a Real-Time Trend-Aware AI Pipeline To deliver personalized marketing suggestions based on live trend data, we had to design a robust data pipeline. This included choosing which models to use, designing the right prompt formats, and sequencing the pipeline efficiently. We experimented with tools like Upstage for document-level understanding and NEAR.AI agents for customized LLM responses. This gave us hands-on experience optimizing multiple models for high performance in a complex, real-time use case.