AnalysisAI original

My F-You To Sam Altman

Linus Tech Tips · 2026-11-02 · 31 min

https://www.youtube.com/watch?v=vDmjw-c_R2k
#Linus Sebastian#ASUS#local AI#data privacy#LM Studio#Hermes#Home Assistant#RTX 5090#SilverStone Alta T1#power supply
Share:

Video summary

Linus builds a local AI workstation with two RTX 5090s, then runs Hermes and LM Studio without recurring cloud fees.

Linus explains why he wants local AI: it can keep data inside his network, work during internet outages, and avoid recurring token costs for most everyday tasks. He rebuilds the workstation in SilverStone’s multi-GPU Alta T1, pairing two RTX 5090s with an ASUS W790E Sage SE motherboard and a 56-core Xeon W9-3495X. The case supports extensive water cooling, but a 3,000-watt SilverStone power supply proves incompatible with the available 120-volt supply, so he switches to a 1,650-watt unit. After wrestling with motherboard clearance, cabling, and a misplaced fan, Linus and the Labs team configure LM Studio and the Hermes agent. The dual GPUs run a local model and connect to Discord, though responses are slow. Linus sees the system as useful for scheduled research and, eventually, Home Assistant automations such as identifying visitors and adapting lights to rainy weather.

Chapters

  1. 0:00Intro: Linus targets local AI for over 90% of his needs
  2. 2:35SilverStone Alta T1: Multi-GPU case with a 3,000-watt PSU and 56-core Xeon
  3. 6:30RTX 5090s and registered memory: two cards and a questionable connector mod
  4. 8:17Installing the CPU and cooler: 8-inch-pound mounting on the Xeon platform
  5. 10:05Case fit, DGX Spark, and ZMT loop: accommodating the workstation build
  6. 15:05RTX 5090 and Optane storage: reused silver fittings in a clean build
  7. 16:47PSU setback: the 3,000-watt unit needs 220–240 volts
  8. 18:19Trying again with a 1,650-watt PSU and planning for 525 watts of slot power
  9. 19:40Wiring the 5090: Linus’s custom cable signals a 600-watt limit
  10. 20:46Fan-fit mishap and loop build: two RTX 5090s still leave room
  11. 22:16Local AI setup: Hermes, LM Studio, and overnight work on two RTX 5090s
  12. 28:00Sventlana works: scheduled news digests and Home Assistant ideas

What the full analysis includes

12 chapters · 53 key points · 12 section summaries · mind map

Preview of the first chapter · Intro: Linus targets local AI for over 90% of his needs

  • •Cloud AI helps with code, analysis, and smart-home questions0:16
  • •Local LLMs address privacy, subscription costs, and internet outages0:49
  • •Linus expects local hardware to handle over 90% of his needs1:12

The remaining 50 key points, section summaries and mind map are in the full analysis. Analyze this video to see everything →

This video is 31 min. The Free plan analyzes videos up to 30 min; plans from Plus cover videos up to 2 hours.

This is a Tier 1 public summary

Whether the chapter key points, section summaries and mind map are public is up to the person who shared it. Want the full analysis?Submit one yourself.

More from this channel

Related analyses