ashvardanian

6 ms·

1.
by ashvardanian 4d ago
One fairly revealing microbenchmark for WASM runtimes is `int8` dot products & angular/cosine distances. (My) NumKong [1] has implementations targeting both vanilla AVX2/Haswell and AVX2-VNNI/Alder Lake, which makes it ea
2.
Bridging Models' Internal States(mostik.ai)
3 points by ashvardanian 14d ago|0 comments
3.
by ashvardanian 1mo ago
If only we had a myriad of heavily funded, founder-led AI-for-coding startups that could afford to store a few source files and need all that code for training anyway :) PS: Saw Cursor’s Origin announcement a second later.
4.
Using AI to build your own software(lemire.me)
6 points by ashvardanian 2mo ago|0 comments
5.
Show HN: ForkUnion v3 – Faster OpenMP-style thread-pool for C, C++, Rust, & Zig(github.com)
3 points by ashvardanian 2mo ago|0 comments
6.
AI Model Co-Design: Hardware-Friendly LLM Design(developer.nvidia.com)
2 points by ashvardanian 2mo ago|0 comments
7.
by ashvardanian 3mo ago
Got really excited for this model and asked my Opus planners in 3 pretty different projects to use Sonnets instead of Opus subagents to help me experiment on HPC kernels faster. Not one of them ended up writing a single line of code... Sonn
8.
KinetIQ Ascend: Toward 100% Reliable Manipulation and Superhuman Speed(thehumanoid.ai)
5 points by ashvardanian 3mo ago|2 comments
9.
Nvidia CUDA Python 1.0 and CUDA 13.3 Release(developer.nvidia.com)
2 points by ashvardanian 3mo ago|0 comments
10.
by ashvardanian 4mo ago
I really like the speed at which Cloudflare is executing toward becoming a critical infrastructure player with all of those new product offerings. That said, not everything needs to be serverless. Their Gen 13 hardware looks impressive, and
11.
Laguna XS.2 and Laguna M.1 by Poolside(poolside.ai)
5 points by ashvardanian 5mo ago|0 comments
12.
FP8 Search and KV-Caching in USearch(unum.cloud)
1 points by ashvardanian 5mo ago|0 comments
13.
Escaping the Fork: How Meta Modernized WebRTC Across 50 Use Cases(engineering.fb.com)
3 points by ashvardanian 5mo ago|0 comments
14.
Porting Go's io package to C(antonz.org)
6 points by ashvardanian 6mo ago|0 comments
15.
Schema as the Core of Reliability in AI Memory(xmemory.ai)
6 points by ashvardanian 6mo ago|0 comments
16.
by ashvardanian 6mo ago
I'm not aware of that, but it would likely be a great application area for SME!
17.
by ashvardanian 6mo ago
The README was written by a human. I’ve used models extensively to refine the content, but never accepted more than a couple of lines of edits at a time.
18.
NumKong: 2'000 Mixed Precision Kernels for All(ashvardanian.com)
47 points by ashvardanian 6mo ago|6 comments
19.
The State of Allocators in 2026(cetra3.github.io)
2 points by ashvardanian 6mo ago|0 comments
20.
by ashvardanian 6mo ago
I don't have the inside scoop on Intel's current mess, but they definitely have a habit of killing off their coolest projects.
21.
by ashvardanian 6mo ago
Would it be accurate to say that Meta currently produces more RISC-V chips than other vendors? The specs for those chips look extremely interesting and seem much more programmable than Google's TPUs. It would be cool to see Meta making
22.
Recraft V4(recraft.ai)
2 points by ashvardanian 7mo ago|0 comments
23.
Nebius to buy AI agent search company Tavily for 275M(nebius.com)
2 points by ashvardanian 7mo ago|1 comments
24.
by ashvardanian 7mo ago
https://www.bloomberg.com/news/articles/2026-02-10/nebius-ag... https://www.tavily.com/blog/tavily-is-joining-nebius
25.
by ashvardanian 7mo ago
8K QPS is probably quite trivial on their setup and a 10M dataset. I rarely use comparably small instances & datasets in my benchmarks, but on 100M-1B datasets on a larger dual-socket server, 100K QPS was easily achievable in 2023: htt
26.
Running Async WebAssembly on Seastar's Reactor(rockwotj.com)
3 points by ashvardanian 7mo ago|0 comments
27.
Open source USearch library jumpstarts ScyllaDB vector search(thenewstack.io)
2 points by ashvardanian 7mo ago|0 comments
28.
Mixedbread: How We Built Multimodal Late-Interaction at Billion Scale(mixedbread.com)
3 points by ashvardanian 8mo ago|0 comments
29.
Nvidia: Using Context as Training Data Unlocks Models That Learn at Test-Time(developer.nvidia.com)
6 points by ashvardanian 8mo ago|0 comments
30.
by ashvardanian 8mo ago
Cool project! And thanks for mentioning "unum-cloud/USearch" among repo examples :)

More ›