TechBrief — تازه‌ترین اخبار فناوری

مرجع روزانه خلاصهٔ اخبار و تحلیل‌های کوتاه از منابع معتبر.

آخرین خبرها

DIDs Are Cool. We Didn't Need Them

Article URL: https://www.inamoon.com/blog/dids-vs-subjects

Comments URL: https://news.ycombinator.com/item?id=47830534

Points: 4

# Comments: 0

A cache-friendly IPv6 LPM with AVX-512 (linearized B+-tree, real BGP benchmarks)

Article URL: https://github.com/esutcu/planb-lpm

Comments URL: https://news.ycombinator.com/item?id=47830071

Points: 5

# Comments: 1

Marathon battery life makes Keychron’s Ultra 8K keyboards its best yet

Keychron's expansive keyboard catalog covers everything from basic full-size mechanical boards to niche layouts like southpaw Alice or a one-handed half-keyboard for gaming. Its Q and V series are among our go-to recommendations for off-the-shelf keyboards, and the latest iterations - the Q and V Ultra models - are a straightforward upgrade to the lineup, […]

Contra Benn Jordan, data center (and all) sub-audible infrasound issues are fake

Article URL: https://blog.andymasley.com/p/contra-benn-jordan-data-center-and

Comments URL: https://news.ycombinator.com/item?id=47829757

Points: 7

# Comments: 2

Claude Token Counter, now with model comparisons

Article URL: https://simonwillison.net/2026/Apr/20/claude-token-counts/

Comments URL: https://news.ycombinator.com/item?id=47829178

Points: 32

# Comments: 7

Show HN: A lightweight way to make agents talk without paying for API usage

Article URL: https://juanpabloaj.com/2026/04/16/a-lightweight-way-to-make-agents-talk-without-paying-for-api-usage/

Comments URL: https://news.ycombinator.com/item?id=47828968

Points: 20

# Comments: 5

Show HN: Run TRELLIS.2 Image-to-3D generation natively on Apple Silicon

I ported Microsoft's TRELLIS.2 (4B parameter image-to-3D model) to run on Apple Silicon via PyTorch MPS. The original requires CUDA with flash_attn, nvdiffrast, and custom sparse convolution kernels: none of which work on Mac.

I replaced the CUDA-specific ops with pure-PyTorch alternatives: a gather-scatter sparse 3D convolution, SDPA attention for sparse transformers, and a Python-based mesh extraction replacing CUDA hashmap operations. Total changes are a few hundred lines across 9 files.

Generates ~400K vertex meshes from single photos in about 3.5 minutes on M4 Pro (24GB). Not as fast as H100 (where it takes seconds), but it works offline with no cloud dependency.

https://github.com/shivampkumar/trellis-mac

Comments URL: https://news.ycombinator.com/item?id=47828896

Points: 122

# Comments: 21