Chat with Oleg's AI

Runs entirely on your device — no server, no data sent anywhere. Powered by a small on-device LLM via WebGPU. First load downloads the model (~500 MB); subsequent visits are instant.