@shoni.eth

/threads v2 is almost done! my focus was on minimizing token bloat, a new llm friendly formatting technique, low effort comment exclusions, quote hydration fixes + recursive quote hydrations (quotes in quotes in quotes)--- it won't include media (img/site/frame/video) text descriptions until v3. The last improvement is unlisted in the image but it revolves around classification and intent. v1: https://huggingface.co/datasets/shoni/farcaster

2 replies

1 recast

1 reaction