Frustrated by AI assistants that can only suggest actions but never take them, I built Merlio (
https://merlio.app) - an AI that generates charts, searches the web, creates images, and analyzes YouTube videos directly within conversations. Built with React, it functions as an AI hub providing access to multiple models (Claude, GPT, Gemini) through a unified interface. The custom LLM orchestration layer enables the AI to execute tools with proper parameters and process results while maintaining conversation flow. Users are visualizing data, generating designs, and summarizing hour-long videos all in a simple chat interface. Would appreciate feedback on this approach to making AI assistants more useful.