I’m Chris — a software developer based in London, currently CTO at Searchable. This is my personal blog, where I typically write about programming, search systems and technology.
Hornet's founders say agents are the new user of retrieval, and their own numbers say the agent's results page is narrower than ours was. The person who used to read the results is still there, one step back, and can no longer see them.
Bessa, Douze and Jégou have released Dino10B, a 10-billion-vector nearest-neighbour benchmark built from DINOv3 image patches. The scale is overdue. How they got to ten billion is the more interesting part.
Qdrant got 5 hyperbolic dimensions to beat 200 Euclidean ones on hierarchical data, and then HNSW returned 0.020 recall on it. The representation win is real and the index can't cash it - because an ANN graph isn't neutral plumbing, it has an opinion about what close means.
A correction to what I wrote about TurboQuant in May. A new paper from Mariano Tepper and Ted Willke makes the case that the rotation in front of a scalar quantiser should be learned from your data, and that learning it is cheap enough that I no longer think calibration cost belongs in the argument.
AWS cut S3 Vectors query pricing by up to 80% at scale. It is a real win for your bill and says nothing about your moat - because the price cut lands on the one slice of retrieval where quality barely lives.
