
Building a Low-Power Proxmox Server (It Didn't Go Smoothly)
The video documents a hands‑on effort to convert a decade‑old i5 workstation into a low‑power Proxmox server. The creator strips the original components, replaces DDR3 with DDR4, installs NVMe storage, and opts for an eighth‑generation Intel mobile CPU board to keep power draw minimal while supporting Plex media encoding. Key technical decisions include reusing the chassis, selecting a board that integrates a low‑TDP processor, and forgoing 3.5‑inch drives in favor of faster, more efficient NVMe units. The build process is hampered by a flood of mismatched screws, obscure front‑panel pin‑outs, and the need to adapt legacy hardware to modern standards. Throughout the stream the host jokes, “I like fixing broken stuff,” and mentions switching back to the Zed editor for coding. He also highlights YouTube’s improved live‑streaming performance, noting that past streams were problematic but recent updates have made them viable for hobbyist content. The project demonstrates how modest hardware can be repurposed for home‑lab workloads, offering a cost‑effective, energy‑efficient alternative to commercial servers. It underscores the sustainability angle of extending the life of legacy equipment while delivering functional services like Plex or Jellyfin for small‑scale media needs.

EEVblog 1749 - Werewolf VFLEX USB-C Power Supply Adapters - Very Cool!
EEVblog reviews the Werewolf VLEX system, a modular USB-C Power Delivery (PD) adapter ecosystem that lets users program small adapters via a smartphone app to negotiate any PD-supported voltage (up to 48V) from a single USB-C power source. The kit...

HM - FMFM _ Fleetscale Memory Fault Management - Workstream - (2026-01-13)
The FMFM workstream convened to review progress on Fleetscale Memory Fault Management, focusing on logging requirements, standards adoption, and recent research presented by Roy. Participants debated whether to prioritize DDR5 or LPDDR logging, noting DDR5’s easier integration with existing specifications...

Networking - ESUN _ Ethernet for Scale-Up Networking - Workstream - (2026-03-26)
Participants debated whether LR and CBFC flow-control features should be mandatory in the Ethernet for Scale-up Networking (ESON/ENS) spec. The group noted LR/CBFC are optional for endpoints today and making them mandatory for switches could shrink the initial ecosystem because...

Can a Robot Make World-Class Coffee? Meet Artly’s AI Barista
At Muji in Portland, Artly showcased Jarvis, an AI-driven robotic barista trained by world-champion barista and co-founder Joe Yang to reproduce craft coffee at scale. The system encodes Yang’s techniques to deliver consistent, high-quality espresso and latte art, aiming to...

New OLED Tech...For Gamers
LG’s latest tandem white OLED gaming display combines stacked R/G/B sublayers to deliver deeper blacks with industry-leading low reflectance, reducing washout in brightly lit rooms. The panel preserves OLED contrast in dark environments while minimizing ambient reflections thanks to its...

Was Killing the Headphone Jack Wrong?
Apple’s decision a decade ago to remove the iPhone headphone jack—debuting with the iPhone 7—helped accelerate rapid improvement in Bluetooth audio and the rise of wireless earbuds like AirPods. The move pushed the market toward wireless and USB-C solutions that...

Physical Data Transmission - Part 5: Resistance and Interference
The video explains how electrical signals travel beyond wires and how that leakage both enables wireless transmission and causes electromagnetic interference (EMI). It outlines three forms of resistance affecting signal transmission: ohmic resistance (heat and the skin effect), capacitive reactance...

Scaling Intelligence Through the Memory Hierarchy with Solidigm
Kapil Kirkra, senior principal engineer at Solidigm, argued that scaling AI intelligence requires a third, often overlooked axis: memory capacity. While larger models and more compute dominate headlines, the talk demonstrated how the memory hierarchy—from high‑bandwidth HBM to NVMe SSD...

PCB Layout Finished 10x Faster with AI? Here’s How...
The video introduces Quilter, a startup applying artificial intelligence to the PCB layout stage of hardware design, and explains how its founders aim to shrink the traditionally slow layout process. Quilter deliberately avoids LLMs, treating layout as a geometry‑and‑physics problem solved...

Chad Roberts-York: Can a Single Tap Eliminate Checkout Friction and Power Retail Branding?
The video introduces the A920 Pro, a point‑of‑sale device that leverages card‑emulation and NFC to push a digital receipt directly to a shopper’s phone with a single tap. By broadcasting the receipt from the terminal, the system removes the traditional...

Jensen Huang on U.S.-China Relations Following His Trip with Trump
Jensen Huang, NVIDIA’s CEO, addressed U.S.-China relations after returning from a trip with President Trump, focusing on the stalled approval of the H200 AI accelerator for Chinese customers. He emphasized that neither he nor the company can predict when Chinese...

Nvidia Earnings in Focus; SpaceX Eyes Historic IPO | Bloomberg Tech 5/20/2026
Bloomberg Tech opened with a dual focus: Nvidia’s upcoming earnings report, which could cement its role as the AI‑driven engine of the market, and SpaceX’s anticipated historic initial public offering. Analysts expect Nvidia to post roughly 80% top‑line growth and...

The iPhone Ultra Is Coming...
Apple’s rumored first foldable—nicknamed the iPhone Ultra—appears headed for a passport-style design with a 5.5‑inch cover screen and a 7.8‑inch inner display, prioritizing pocketability and a wider tablet-like interior. Leaked specs and mockups point to a thin chassis, dual 48MP...

Cleaning up the Mess - ISPASS'26 Talk by Nisa Bostanci and Haocong Luo
The ISPASS 2026 presentation by Nisa Bostanci and Haocong Luo challenges a recent best‑paper that claimed the open‑source DRAM simulator Emulator 2 dramatically under‑performed real hardware. The speakers identified three fatal flaws: missing source code prevented replication; the simulation used only eight DDR5 channels while the...