Atlas AI
Chat is live today. Image generation and speech / translation will roll out next. This page tracks what’s planned and what hardware it will land on.
Goal: small, fast image generation for diagrams, thumbnails, and mockups. Targeting Jetson nodes once stable. Output will be gated to members only.
Goal: low-latency ASR + TTS for meetings and media. Results should stream back into Matrix/LiveKit rooms and Pegasus.
Atlas AI chat is running on local GPU hardware at chat.ai.bstein.dev. The chat page streams responses and
reports latency per turn. As larger models come online on the Jetsons, the chat endpoint will be upgraded in-place.