Doing lesson 15 of the @fastdotai course; deducing how to rearrange convolutions as a matrix product

How Python's __all__ works. from lib.core import * 1. check if core.python has an __all__ list 2. if it does, then only import everything defined in __all__ 3. if it doesn't, only import those functions that are not prefixed with _

Introducing FrontierCode: a coding eval that raises the bar for difficulty & quality. Each task took 40 hrs of work by leading open-source maintainers. Models write sloppy code that works but isn’t maintainable. Our eval is first to measure: would you actually merge this code?

Omni is now open sourced in apache 2 github.com/hanxiao/omni-maco… few things i fixed & learned since friday - perf of search while indexing: a much more common issue for local use as file changes frequently. - any2any search: since omni embedding is used, one can now do find similar of a selected file and do {text,img,audio,video}<>{text,img,audio,video} - search history is added, so u can revisit a query like mac's "smart folder" ux. - swift ui stalls / "not responding" and compatibility issue, same ui can actually render differently on Tahoe and Sequoia - add a profiling tool so everyone can benchmark against the same dummy test data and upload their performance to our homepage finally, my lack of interests of vectordb over years becomes a debt & backfire while developing this. efficiency of the data structure is quite important.

Close reading @ericries's Incorruptable From chapter 2, I came up with a new concept of my own: structural hypocrisy When an org's structure is perfect in one area, but mistakenly thinks it is sufficient for another goal