Pinned Tweet
it blows my mind that docker downloads 1.17 GB to run a numpy script that opens 210 files
so i built a container runtime that only fetches a file when the app actually opens it. the median python app i tested reads only 7.8% of its image
A 1.17 GB image: 28,542 files. A simple numpy app only reads 210 of those 28,542 files. Only 6.9% of the image.
Look how big translations and docs is!
when I first read how Traffic Control predicts a query's CPU before it runs, I thought: no way that works.
turns out, it does work. I implemented their model and ran the same query 50 times. Here’s predicted vs. actual CPU.
I wrote up how the model works and benchmarked it on TPC-H and the Join Order Benchmark, including where it breaks: swayne.dk/blog/postgres-quer…
(Independent project, not affiliated with PlanetScale.)
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
github.com/openai/math
