It all starter from With the law in US (which was reverted) that you couldn't use all software research as expense so you needed to pay tax on no-profit. Then you cannat hire people remotly if you do that you will loose tax benefits and stuff. then ai picked up . basically we are screwed
One part I think people may find interesting is the media pipeline. It explains how audio flows through the system, how frames are created, why click sounds happen, and how buffering, pacing, and the jitter buffer handle audio that arrives at irregular intervals.
I also open sourced the project. There are some interesting infrastructure issues in the repo that might be worth a look.
The blog traces how search actually evolved. Every algorithm exists because the previous one had a specific flaw and a new one came into existence to fix it.
I went deep over a month to understand how to build a real retrieval pipeline. What algorithms exist, why they exist, and what problem each one was solving when it was invented. Reading this you will come away with a full mental model of how to build a retrieval pipeline for any use case, and how to tune each stage for your specific needs.
The blog traces how search actually evolved. Every algorithm exists because the previous one had a specific flaw and a new one came into existence to fix it.
I went deep over a month to understand how to build a real retrieval pipeline. What algorithms exist, why they exist, and what problem each one was solving when it was invented. Reading this you will come away with a full mental model of how to build a retrieval pipeline for any use case, and how to tune each stage for your specific needs.