Have you ever felt like your Windows PC is working against you? Between the bloatware, the nagging updates, and the ads creeping into the Start Menu, it’s easy to feel like Microsoft has turned your operating system into a marketing platform. But what if an AI agent could take control, strip away the junk, and even build you a hacking lab — all while being safely locked inside a sandbox? That’s exactly what NetworkChuck explores in his latest video, “Switching back to Windows?!?”, where he puts the newly Windows-native OpenAI Codex to the test.
In this honest, hands-on review, NetworkChuck confronts a question many of us have pondered: can a powerful AI agent like Codex actually make Windows enjoyable again? As someone who runs both a Mac and a Windows PC, he admits he’s been drifting away from Microsoft’s ecosystem. But with OpenAI’s Codex now living natively in PowerShell — and backed by a 15,000-line Rust sandbox — the game might have changed. Let’s dive into what this means for sysadmins, security nerds, and anyone tired of Microsoft cramming junk into their OS.
What “Windows-Native” Really Means for an AI Agent
When OpenAI announced that Codex was going Windows-native, it wasn’t just a port. As NetworkChuck explains, the agent now runs directly within PowerShell, not through a compatibility layer like WSL. This is a big deal. Instead of being a guest in a Linux container, Codex can interact with Windows’ core utilities — winget, icacls, firewall rules, and even Hyper-V — using native Windows commands. This means the AI can perform deep system tasks, like uninstalling bloatware or locking down Windows Update, without the overhead of translation layers.
Building on this, the video dives into the engineering behind the sandbox. OpenAI wrote 15,000 lines of Rust to create a security boundary that prevents the AI from escaping and damaging your machine. The sandbox uses restricted tokens, dedicated SIDs, and ACLs to limit what the agent can see and do. For security professionals, this is a fascinating glimpse into how modern AI agents can be safely deployed on Windows without compromising the host system.
How Codex Cleans Windows Bloat and Stops Update Nags
One of the most practical demonstrations in the video is using Codex to clean up a Windows machine. NetworkChuck shows how the agent can automatically strip out adware, disable telemetry, and even stop Windows Update from reverting your customizations. Instead of hunting through settings or running scripts manually, you simply tell Codex what you want, and it executes the changes using winget and PowerShell commands.
As a result, the AI becomes a powerful tool for power users who want a cleaner, faster Windows experience. The video also highlights how Codex can handle complex tasks like setting up a Hyper-V hacking lab with Kali Linux — spinning up six sub-agents to build the environment from scratch. This isn’t just about convenience; it’s about reclaiming control over your own operating system. However, the real star here is the permission system: you can choose between auto-review mode, where you approve each action, or full-access mode, where the agent runs freely. This flexibility makes Codex suitable for both cautious users and those who trust the sandbox’s security.
Is the Codex Sandbox Safe? A Deep Dive into Security
The biggest fear when giving an AI control of your PC is that it might accidentally — or intentionally — cause chaos. NetworkChuck addresses this head-on by explaining how OpenAI’s Windows sandbox works. The sandbox creates a separate user account with restricted tokens, applies strict firewall rules to block internet access, and uses Windows’ own ACL system to limit file permissions. In essence, the AI is trapped in a secure environment where it can only interact with the system through predefined channels.
This engineering deep-dive is a must-watch for anyone concerned about AI safety. The video references the open-source sandbox code on GitHub, as well as OpenAI’s detailed blog post about building the sandbox. For security nerds, seeing the icacls commands and SID configurations in action is both educational and reassuring. It demonstrates that OpenAI has thought deeply about the risks of letting an AI agent loose on Windows, and they’ve built a robust defense.
The Verdict: Can Codex Win You Back to Windows?
By the end of the video, NetworkChuck delivers a balanced verdict. He admits that Codex is impressive enough to make him reconsider his relationship with Windows. The ability to automate tedious tasks, clean bloat, and even build complex environments with a single prompt is genuinely powerful. But he also acknowledges that the experience isn’t perfect — there are still quirks, and the AI sometimes needs guidance. Nevertheless, for sysadmins, security researchers, and anyone who loves tinkering, Codex represents a new frontier in Windows automation.
If you’re curious about what a Windows-native AI agent can do, or if you’re tired of fighting with your own operating system, this video is a must-watch. NetworkChuck’s honest, hands-on approach makes complex concepts accessible, and his enthusiasm is infectious. So grab a coffee, open the video, and see for yourself whether OpenAI Codex can win you back to Windows.
