Are You Lying to Me Now, Paul? - OmniBook Ultra 16, ProArt P16, & Lenovo Pro 9n Impressions

Summary

Windows, Updates, and Edge

• Microsoft is continuing a broad effort to address Windows 11 usability and performance complaints, including changes to the Start menu, taskbar, File Explorer, and Search.
• Further work is aimed at improving performance on lower-memory PCs, including systems with 8 GB of RAM.
• The Windows Photos app is being rewritten with the newer Windows app framework. Microsoft says it should launch and run faster while retaining separate gallery, photo-viewing, and editing experiences.
• Search is moving toward a simpler interface and natural-language actions, such as changing system settings. Some of these features are in the Windows Insider program.
• A notable issue for new RTX Spark machines: Paul says they arrived on Windows 11 26H1, which initially lacks some of the newer 26H2 features. Microsoft has not clearly explained the version’s upgrade path. The hosts argue that security updates should be separated from feature updates so organizations can install urgent patches without also taking new features.
• Microsoft Edge is experimenting with “restart to update on idle.” It installs an update when the browser is not in use, then restarts and restores the session, reducing interruptions. The feature must currently be enabled manually.

RTX Spark Hands-On Impressions and Local AI

Paul discusses three laptops: the Lenovo Pro 9i, HP OmniBook Ultra 16, and ASUS ProArt P16. The Lenovo is the lower-end configuration he tested, with 32 GB of RAM; the HP also has 32 GB, while the ASUS configuration has 128 GB. The machines pair Arm-based processors with NVIDIA Blackwell graphics hardware.

His central criticism is that these are powerful machines without a clear, ready-to-use local-AI experience. Unlike Copilot+ PCs, they do not provide an obvious set of built-in AI features for ordinary users. Buyers must choose applications and models, configure them, and learn a fast-changing ecosystem of tools and settings. Paul found that the manufacturers’ usual laptop review materials did little to explain the local-AI setup, while NVIDIA’s more useful documentation was difficult to discover and arrived late in his review process.

Other impressions:

• Performance and cooling: The laptops can be loud and hot under heavy AI workloads. Performance drops substantially on battery, making them less suitable for unplugged use than typical thin-and-light PCs.
• Memory matters: 128 GB of RAM makes larger models and more demanding workloads possible, but Paul views it as an entry point for serious local AI rather than an unlimited resource. RAM allocation between the GPU and system may require manual adjustment and a reboot.
• Setup and usability: Local models and AI “harnesses” vary in their requirements and behavior. Paul’s attempts to use a large model to create a simple Notepad-style app stalled or failed, despite lengthy runs and considerable resource use. He contrasts this with cloud coding services, which he says can often complete similar tasks more quickly and with less effort.
• What the hardware may suit: Paul sees the laptops as potentially useful for creators using tools such as Photoshop or Premiere, and for developers, engineers, or AI enthusiasts prepared to configure their environments. He questions whether high-performance AI workloads would be better served by always-on desktop or headless systems rather than laptop designs.
• Gaming: The machines appear to run a selection of games well, with NVIDIA saying games can use the graphics hardware natively even though other parts of the system involve emulation. Paul reports good results with titles including Doom: The Dark Ages and Microsoft Flight Simulator 2025. He says the RTX hardware offers a stronger gaming experience than Snapdragon PCs, though the local-AI setup remains the biggest practical challenge.

Microsoft 365, OneDrive, and AI Security

• Microsoft’s OneDrive event introduced Copilot-related plans for files and photos, including AI-generated reports or visual summaries, photo tools, map views, and a command-line interface. OneDrive is also being optimized to work with as little as 8 GB of RAM. The hosts note that Microsoft has discussed AI over users’ files for years, so they are waiting to see how broadly and when the announced capabilities arrive.
• Microsoft outlined a broader security strategy for business AI agents. The plan involves identity, permissions, monitoring, and containment across Windows, Microsoft 365, and enterprise services, rather than relying on containers alone. The goal is to keep agents within approved boundaries and make their actions visible to administrators.
• OpenAI introduced a coding model trained for older languages such as Fortran, COBOL, and Visual Basic. The hosts see potential in helping maintain or modernize long-lived software systems that have fewer developers and less current documentation.

Xbox and Gaming

• A report says the next Xbox project, Project Helix, may begin with a handheld gaming device in 2027, followed by a living-room console in 2028. The reported price range of $1,200–$1,800 is discussed skeptically. Xbox head Asha Sharma disputed aspects of the report while indicating that Helix is being built on Windows technology.
• The hosts think a Windows-based Xbox platform could make sense for compatibility with PC games and existing Xbox titles. They also discuss how upscaling and related rendering technologies may help future consoles deliver attractive visuals without relying solely on more powerful hardware.
• Microsoft announced standalone cloud-gaming plans for Xbox Insiders: 10 hours for $3.99, 25 hours for $7.99, or 50 hours for $11.99.
• The ASUS ROG Xbox Ally handheld was discussed as an expensive option, starting around $1,300. AMD is also bringing a newer upscaling technology to certain Xbox handhelds, Xbox Series X|S systems, and supported Windows PCs.
• The Final Fantasy VII remake is among the games coming to Game Pass. The show also mentions a remaster of Telltale’s The Wolf Among Us and the closure of BioWare, known for Mass Effect and other franchises.

Apps, Guide, and Tutorial-Related Items

• Tolaria is presented as a free, open-source, Markdown-based alternative to Notion that stores content as files and can sync through cloud services. Mobile versions are in early testing for iOS and Android.
• Llama Windows is a Windows app for running local Llama-family models. Paul says it offers a comparatively approachable setup and model recommendations, though his own lengthy model run did not produce the app he wanted.
• RunAs Radio: Richard Campbell recommends an episode featuring Microsoft security specialist Will Valita. It covers the OWASP top risks for AI agents—including memory poisoning, goal hijacking, and privilege abuse—and discusses what administrators should consider before deploying agents.
• Paul’s RTX Spark segment is a hands-on impressions report, not a step-by-step setup guide. His main takeaway is that the hardware is promising, but buyers currently need substantial technical knowledge and better documentation to make local AI practical.

Main Speakers and Sources

• Micah Sergeant — host, filling in for Leo Laporte.
• Paul Thurrott — Windows journalist and reviewer; provided hands-on impressions of the three RTX Spark laptops and commentary on Windows and Xbox.
• Richard Campbell — co-host; contributed analysis on Windows, AI-agent security, gaming, and the RunAs Radio recommendation.
• Other sources discussed: Microsoft, NVIDIA, OpenAI, Xbox, and the manufacturers of the reviewed laptops.