From 02a410de6a67478951669fa9995c8cf6cfbf7224 Mon Sep 17 00:00:00 2001 From: sync Date: Mon, 26 Jan 2026 08:44:22 +0000 Subject: [PATCH] =?UTF-8?q?Edit=20=E2=80=98good=5Fideas=E2=80=99:=20Synced?= =?UTF-8?q?=201769415285204?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- good_ideas.myco | 1 + 1 file changed, 1 insertion(+) diff --git a/good_ideas.myco b/good_ideas.myco index 187c33f..50ccc22 100644 --- a/good_ideas.myco +++ b/good_ideas.myco @@ -102,6 +102,7 @@ Graph-based vector indices do beam search on a graph constructed so that beam se * We can also pack the graphs better (with some cost at retrieval time due to multi-page accesses and more compute): sorting and delta compression of node indices, offload metadata again, maybe general-purpose compressors, maybe QAT/IAA codecs, quantize the graph vectors a bit. * 100μs/node is assuming SSD random latency, and with that assumption these algorithms still aren't useful if we can throw more time at it because we can do more reads. But with really clever pathfinding, HDDs could also work. Maybe? Two orders of magnitude slower, though. * https://arxiv.org/pdf/2501.10479 +* Also see the Jina spherical compression thing. } * SAEs/dictionary learning algorithms for residual quantization. * Minecraft/Factorio crossover modpack via Clusterio-like thing.