Data dump repository and browser extensions for Open LLMs

Explain Like I'm Normal

A new community proposal suggests building browser extensions to scrape and centralize user conversations with frontier models like GPT-4. The goal is to create a massive, public dataset for distillation, allowing open-source models to learn from the reasoning traces of their paid competitors. While privacy and PII scrubbing remain the biggest hurdles, it represents a decentralized attempt to bypass restrictive terms of service.