HN Reader

For those of us who operate on site, we have to add back network latency, which negates this win entirely and makes a proprietary cloud solution like this a nonstarter.

Are there vector DBs with 100B vectors in production which work well? There was a paper which showed that there's 12% loss in accuracy at just 1 mln vectors. Maybe some kind of logical sharding is another option, to improve both accuracy and speed.

This is legitimately pretty impressive. I think the rule of thumb is now, go with postgres(pgvector) for vector search until it breaks, then go with turbopuffer.

v cool and impressive!

Fun!

I was curious given the cloud discussion - a quick search suggests default AWS SSD bandwidth is 250 MB/s, and you can pay more for 1 GB/s. Similar for s3, one http connection is < 100 MB/s, and you can pay for more parallel connections. So the hot binary quantized search index is doing a lot of work to minimize these both for the initial hot queries and pruning later fetches. Very cool!