Felix Rieseberg today released Clippy, a desktop application that combines a retro Windows 98-style interface with local large language model inference. The app uses llama.cpp and node-llama-cpp to run models offline on the user’s machine, automatically selecting the best backend (Metal, CUDA, or Vulkan). It is available for macOS, Windows, and Linux.

The project quickly climbed to the top of Hacker News with over 1100 points and 275 comments. Many commenters saw a natural fit between Clippy’s legacy—a character that was both derided and beloved—and the current wave of AI chatbots. “I’ve been humorously referring to chat agents as next gen Clippy,” wrote user mrandish, who expressed surprise that Microsoft had not already adopted the character for Copilot.

Others debated whether Clippy’s return would be ironic nostalgia or a genuine UX improvement. Some noted the original Clippy was actually helpful for beginners, while others recalled its intrusive popups. The conversation reflected a broader sentiment: LLMs, like Clippy, are often seen as chipper, talkative, and occasionally in the way. Rieseberg frames the project as art rather than utility: “I made it because building it was fun for me.”