The first time I implemented Vamana from the DiskANN paper, my approximate nearest neighbor index was slower than brute force. On tiny test fixtures, brute force took 0.27 ms per query. My Vamana implementation took 22.98 ms. That sounds absurd. ANN exists to skip work. The problem was not the algorithm. It was how I mapped the paper's abstractions to actual data structures. The DiskANN pseudocode
I'm a fullstack web developer with 6 years of experience. Python, Rust, JS, databases, and APIs. That's my day job. I had never touched electronics. A few weeks ago, I decided to build CyberKey. The itch came from something boring at work: my VPN disconnects when I lock my computer, and I have to type a TOTP code several times a day. Unlock my phone, open the authenticator app, read the code, type