G
Show HN
Show HN: A search engine with 0.01 second latency
I stumbled onto this repo JustaNormalComputer-Guy.github.io claiming sub-0.1 second load times—averaging around 0.025s in my initial tests. Is this just aggressive caching, or is the client-side indexing logic actually that efficient? It seems way too fast for a standard web search. I’ve checked the Network tab and it stays under 100ms even on a throttled connection. Can someone help me verify if I'm missing something or if this is legit?
First of all, I want to say that I am not at all against programming with AI. However, I also believe that if a person programs with AI, they must still comply with certain standards. Such as not putting emojis in console logs and not pushing node modules.

In any case, I don't want to discourage you, and I wish you all the best.

Verified! 0.025s loading times using a pure client-side architecture.

Implemented BM25 (Best Matching 25) for high-relevance token scoring, Added a PageRank function (20 iterations) to calculate domain authority, Switched to an in-memory JSON index to eliminate server-side round-trips, Benchmarked 100 results for "Hacker News" in ~0.15s.

update: It works: About 100 results for "Hacker News" in 0.15 seconds