There is also TensorFlow.js for deploying ML models inside the web browser.

https://github.com/tensorflow/tfjs

The figure with ping results from Vermont was where I stopped. The idea that there would be one result for US East and the same result for every other data centre in the world (median) and that nothing odd is happening means that this blog is not for me.

the article is not really about network latency. i think you can safely ignore that part to enjoy the rest of the article

The article is wonderful and in-depth. The majority of future compiler and optimization work is in the ML realm. So it is important that compiler engineers, how few of them there are, are aware of the issues.