Ares was an ambitious attempt by a founder in SF, to recreate the next generation of desktop assistants, inspired by emerging AI-native experiences like Sky.app. In an incredibly aggressive 2-week timeline, we approached the product not as just another AI chat application, but as a true Computer Use Agent capable of understanding and acting within a user's desktop environment.
Before writing a single feature, our team conducted extensive research into the evolution of desktop productivity — studying Spotlight, Alfred, ChatGPT Desktop, Claude Code, Manus, and Sky.app itself. We reverse-engineered user experiences, mapped workflows, and established a structured product development process to maximize velocity without sacrificing engineering discipline.
The resulting prototype featured a native-feeling Electron application with a Liquid Glass-inspired interface, achieving over 60% visual similarity to the original Sky experience. Under the hood, we integrated MCP servers and AppleScript-based desktop automation, allowing the AI to interact with macOS applications, execute multi-step workflows in a single request, create notes, manage files, and request explicit user permissions for privileged actions.
The project also explored a modern AI architecture using bundled MCP servers, local tool execution, and a flexible agent framework capable of expanding beyond simple chat interactions.
Although the project was ultimately shelved before completion, it demonstrated our ability to rapidly research, prototype, and engineer cutting-edge AI experiences at the frontier of human-computer interaction.