Hacker News Reader

With how fast everything is moving I wouldn't be surprised to see "run a codex like LLM model via torrent" in reaction to this.

That and comparable performance running locally. On r/MachineLearning there was just a post about running 7B Alpaca on Android mobile.