Google will allow only apps from verified developers to be installed on Android (9to5google.com)

We already have "sparse" embeddings. Google's Matryoshka embedding schema can scale embeddings from ~150 dimensions to >3k, and it's the same embedding with layers of representational meaning. Imagine decomposing an embedding along principle components, then streaming the embedding vectors in order of their eigenvalue, kind of the idea.

jxmorris12 · 4m ago

Matryoshka embeddings are not sparse. And SPLADE can scale to tens or hundreds of thousands of dimensions.

tkfoss · 26m ago

Wouldn't holy grail then be parallel channels for candidate generation;

  euclidean embedding
  hyperbolic embedding
  sparse BM25 / SPLADE lexical search
  optional multi-vector signatures

  ↓ merge & deduplicate candidates

followed by weight scoring, expansion (graph) & rerank (LLM)?

ArnavAgrawal03 · 1h ago

we used multi-vector models at Morphik, and I can confirm the real-world effectiveness, especially when compared with dense-vector retrieval.

codingjaguar · 9m ago

Curious is that colbert-like ones?

Google will allow only apps from verified developers to be installed on Android (9to5google.com)

Ask HN: The government of my country blocked VPN access. What should I use?

Gemini 2.5 Flash Image (developers.googleblog.com)

FFmpeg 8.0 (ffmpeg.org)

What are OKLCH colors? (jakub.kr)

Dissecting the Apple M1 GPU, the end (rosenzweig.io)

A German ISP changed their DNS to block my website (lina.sh)

Claude for Chrome (anthropic.com)

DeepSeek-v3.1 (api-docs.deepseek.com)

AI tooling must be disclosed for contributions (github.com)

Updates to Consumer Terms and Privacy Policy (anthropic.com)

Show HN: Base, an SQLite database editor for macOS (menial.co.uk)

A visual introduction to big O notation (samwho.dev)

Go is still not good (blog.habets.se)

Comet AI browser can get prompt injected from any site, drain your bank account (twitter.com)

We regret but have to temporary suspend the shipments to USA (olimex.wordpress.com)

U.S. government takes 10% stake in Intel (cnbc.com)

Monodraw (monodraw.helftone.com)

Waymo granted permit to begin testing in New York City (cnbc.com)

Ban me at the IP level if you don't like me (boston.conman.org)

Tesla said it didn't have key data in a fatal crash, then a hacker found it (washingtonpost.com)

Google has eliminated 35% of managers overseeing small teams in past year (cnbc.com)

Michigan Supreme Court: Unrestricted phone searches violate Fourth Amendment (reclaimthenet.org)

Altered states of consciousness induced by breathwork accompanied by music (journals.plos.org)

Scientist exposes anti-wind groups as oil-funded, now they want to silence him (electrek.co)

US Intel (stratechery.com)

Unexpected productivity boost of Rust (lubeno.dev)

Are OpenAI and Anthropic losing money on inference? (martinalderson.com)

Io_uring, kTLS and Rust for zero syscall HTTPS server (blog.habets.se)

Nx compromised: malware uses Claude code CLI to explore the filesystem (semgrep.dev)

Claude Sonnet will ship in Xcode (developer.apple.com)

How to build a coding agent (ghuntley.com)

What makes Claude Code so damn good (minusx.ai)

Framework Laptop 16 (frame.work)

Some users have noticed settings that let Meta analyze and retain phone photos (zdnet.com)

Building the mouse Logitech won't make (samwilkinson.io)

Line scan camera image processing for train photography (daniel.lawrence.lu)

The Therac-25 Incident (2021) (thedailywtf.com)

Proposal to Ban Ghost Jobs (cnbc.com)

The GitHub website is slow on Safari (github.com)

Ask HN: Why hasn't x86 caught up with Apple M series?

I Am An AI Hater (anthonymoser.github.io)

Malicious versions of Nx and some supporting plugins were published (github.com)

Uncertain<T> (nshipster.com)

A teen was suicidal. ChatGPT was the friend he confided in (nytimes.com)

Manim: Animation engine for explanatory math videos (github.com)

Everything I know about good API design (seangoedecke.com)

We put a coding agent in a while loop (github.com)

95% of Companies See 'Zero Return' on $30B Generative AI Spend (thedailyadda.com)

Show HN: A zoomable, searchable archive of BYTE magazine (byte.tsundoku.io)

The Theoretical Limitations of Embedding-Based Retrieval

Comments (6)