generated multi-modal embeddings of my bookmarks using Google DeepMind’s new Gemini Embedding 2, allowing low-dim spatialisation using UMAP and high-dim similarity querying via different media types.

@GoogleDeepMind Looks great 👍 doing something similar in the Bookmarx AI app

@GoogleDeepMind love this! will probably write something up about learnings r.e. embedding media vs text vs media & text together.

@GoogleDeepMind Gemini 2's multi-modal depth may win-yet human curiosity

@GoogleDeepMind UMAP spatialization is such a clean way to audit how well those Gemini embeddings cluster.

@GoogleDeepMind i'm building a fence with 150 4 by 4 inch 8 foot posts. you appear childlike and whimsical to me.

@GoogleDeepMind Very cool- would love to have something like this for monitoring info flows across thousands of cloudfare hosted domains