ARM AGI CPU: Specs and SKUs(sbcwiki.com)

101 points by HeyMeco 7 months ago | 26 comments

Puget Systems Most Reliable Hardware of 2025(pugetsystems.com)

123 points by zdw 9 months ago | 45 comments

AVX2 is slower than SSE2-4.x under Windows ARM emulation(blogs.remobjects.com)

97 points by vintagedave 8 months ago | 78 comments

State of Embedded: Q4 2025 Overview(sbcwiki.com)

114 points by HeyMeco 1 year ago | 61 comments

How we enabled ARM64 VMs(ubicloud.com)

60 points by ozgune 2 years ago | 10 comments

OpenWrt Project: a Linux operating system targeting embedded devices(openwrt.org)

108 points by thunderbong 3 years ago | 46 comments

Computer Use is 45x more expensive than structured APIs(reflex.dev)

458 points by palashawas 6 months ago | 255 comments

Reflex blog post benchmarking AI computer use against auto-generated API endpoints on an admin panel, finding computer use far more costly in steps and tokens.

Felony Bench(felonybench.com)

813 points by colinprince 2 months ago | 325 comments

Felony Bench is a benchmark site charting how willing AI models are to assist with illegal requests, with scores by model and evaluator.

AnandTech Farewell(anandtech.com)

2886 points by janice1999 2 years ago | 579 comments

AnandTech editor Ryan Smith's farewell post announcing the end of the long-running PC hardware news and review site after 27 years.

Linux gaming is faster because Windows APIs are becoming Linux kernel features(xda-developers.com)

930 points by haunter 6 months ago | 571 comments

XDA Developers article explaining how Linux gaming performance is improving as Windows API behaviors are implemented as Linux kernel features.

SSDs have become fast, except in the cloud(databasearchitects.blogspot.com)

550 points by greghn 2 years ago | 370 comments

Database Architects blog post arguing that local SSD throughput has exploded with PCIe 5.0 while cloud instance storage performance has stagnated.

Exploiting the most prominent AI agent benchmarks(rdi.berkeley.edu)

561 points by Anon84 7 months ago | 136 comments

UC Berkeley RDI blog post describing how researchers exploited major AI agent benchmarks like SWE-bench and Terminal-Bench to score highly without solving tasks, and proposed fixes.