I tried local AI on HP's Omnibook Ultra

Key Points:

• PCs built on Nvidia's RTX Spark architecture, like the HP OmniBook Ultra, deliver exceptional AI performance with 1 petaflop of FP4 AI power, 128 GB unified memory, and 2 TB Gen5 NVMe storage, making them ideal for intensive AI tasks.
• Running local open-weight AI models eliminates per-token charges and usage limits, allowing projects to be completed cost-free while maintaining privacy by keeping data off cloud servers.
• The author successfully used a local AI setup with open-source tools to index 20 years of ZDNET posts in one day, demonstrating efficient human-in-the-loop coding and real-time AI reasoning without overheating the laptop.
• Commercial AI subscriptions can become costly and restrictive due to token usage limits and unpredictable billing, with top users spending thousands annually, highlighting the financial advantage of local AI models for heavy coding workloads.
• While the $7,000 HP OmniBook Ultra is currently expensive and not a practical investment for all users, ongoing improvements in computing power and price reductions are expected to make such powerful AI-capable PCs mainstream and more affordable in the near future.