Eric Jang Predicts Singularity Stasis at Full Cache Hits
Eric Jang links full AI inference cache hits to permanent singularity stasis.
Eric Jang stated that at 100% cache hit rate all past inference queries become future ones, collapsing the future into the present and driving the universe to permanent singularity stasis. In reply he added that this is why tokens should not go to a single model provider. Nathan Benaich replied to the same post that the outcome is great for margins. The comments appear in replies to a post by PeterJ_Walker discussing cached prompts and agentic tokens.
At 100% cache hit rate, all past inference queries are all future inference queries, the future collapses upon the present, and the universe fooms to permament Singularity Stasis
Eventually 99% of agentic tokens will just be the cached prompt.
Reactions
3 postsEJ 
Eric Jang@ericjang11and that is why you should not send your tokens to a single model provider
- likes: 7
- replies: 2
- bookmarks: 1
- reposts: 0
NB GK Greg Kamradt@GregKamradtThe more of your day-to-day life is cached, the faster it goes by
- likes: 23
- replies: 1
- bookmarks: 5
- reposts: 0