DEV Community

Cover image for PS5 Linux Is Dead: Andy Nguyen Quits After Blaming LLM-Assisted Hackers for Its Final Hypervisor Bug
Sanjay Singh
Sanjay Singh

Posted on Originally published at zyvop.com

PS5 Linux Is Dead: Andy Nguyen Quits After Blaming LLM-Assisted Hackers for Its Final Hypervisor Bug

The announcement

Andy Nguyen, the security researcher known online as theflow0, said this week that he's stepping away from the PlayStation 5 hacking scene and shutting down his work on ps5-linux, the project he led to turn retail PS5 consoles into bootable Linux machines.

In a post on X, Nguyen described months of planned work, including PS5 Pro support slated for 2027, as lost. He blamed the decision on the state of the surrounding PS5 research scene rather than a technical dead end.

His criticism, as stated in the post, is aimed at contributors who rely heavily on LLMs to produce hacks they cannot explain or debug themselves. That framing is why the story has spread beyond PlayStation homebrew circles and into the broader argument about what unsupervised AI-assisted development means for volunteer-run software projects.

What ps5-linux actually did

Nguyen's project was never a toy demo. It uses patched hypervisor vulnerabilities to boot Linux on supported PS5 hardware, exposing the console's CPU and GPU for use outside Sony's normal software environment. The project's documentation says ps5-linux supports PS5 Phat and Slim consoles on firmware versions 3.00 through 7.61, with features including HDMI 4K60 output, M.2 SSD support, and access to the console's hardware for Linux workloads. (ps5-linux GitHub)

TechSpot covered Nguyen's early demonstrations in March 2026, when he showed demanding software running under Linux on PS5 hardware, including Grand Theft Auto V Enhanced with ray tracing. Follow-up demonstrations showed the project continuing to improve as the hardware stack was tuned.

By late April, the project had become a reproducible public effort on GitHub, with tooling for building the environment and installing Linux on supported consoles. Tom's Hardware covered the public release and its support for features including M.2 storage.

The project credited other researchers too, including c0w, resulknad, and flatz, alongside researchers associated with fail0verflow and ps5-payload-dev. Its current public documentation lists support for PS5 Phat and Slim systems running firmware 3.00 to 7.61.

The bug that ended it

According to Nguyen, the exploit behind ps5-linux was also the last viable hypervisor bug available for the project. That is Nguyen's characterization of the vulnerability rather than an independently established fact.

Nguyen said other researchers found the same bug and reported it to Sony for a bounty, rather than coordinating with the ps5-linux project. FRVR reported that the researchers involved were using AI during their research. The exact role AI played in discovering the vulnerability is therefore best described as reported rather than independently verified.

Nguyen also said he had asked the researchers to delay disclosure until Grand Theft Auto VI shipped, so users could buy the game and still have the option of dual-booting Linux on vulnerable consoles afterward. According to his account, they agreed and then reported the vulnerability to Sony less than a day later.

That dispute is different from a normal bad code submission. The core issue was vulnerability research and disclosure, but it became another example in Nguyen's broader complaint about contributors moving quickly with AI-assisted work without fully understanding the consequences of what they found.

Other maintainers are saying the same thing

Nguyen isn't the only maintainer dealing with an increase in AI-assisted contributions.

The RPCS3 project now explicitly permits AI tools for research and reverse-engineering purposes, but says contributors must fully understand and take responsibility for the code they submit. The project also requires PRs opened by AI agents or automated tools to disclose the scope of AI involvement and what human testing or review was performed. Undisclosed submissions may be closed without review.

The same pressure is visible in Godot. In June 2026, the Godot Foundation officially acknowledged that AI-generated contributions were increasing the review burden and said the project's contribution rules would become stricter. Its current contribution policy prohibits autonomous AI agents from submitting PRs, restricts the use of AI to generate substantial code, requires disclosure of AI assistance, and states that contributors must understand and take responsibility for everything they submit.

That makes the comparison more concrete than simply saying several maintainers are frustrated. RPCS3 and Godot have both responded with formal contribution rules, not just complaints on social media.

The timeline

Mermaid Diagram

The full run, from Nguyen's first public demonstrations to his decision to step away, took a little over six months.

What this means if you maintain an open-source project

RPCS3, Godot, and Nguyen's experience point toward the same practical lesson: AI contribution rules work better when they are written down before a problem happens, rather than improvised after a maintainer is already frustrated.

GamesRadar+ has the fuller discussion of the RPCS3 rules, while the RPCS3 repository itself documents its current requirements. The project permits AI for research and reverse engineering but expects human contributors to understand submitted code, and it requires disclosure when AI agents or automated tools are used to open pull requests.

Godot has taken a stricter approach. Its current rules prohibit autonomous AI-agent submissions and substantial AI-generated code while allowing limited assistance for tasks such as code completion. It also requires contributors to understand and take responsibility for what they submit.

None of those policies could have prevented a researcher from independently finding and disclosing a security vulnerability. That is what makes Nguyen's case different from a low-quality pull request.

What written policies can do is set expectations before maintainers spend time reviewing work, make contributors accountable for their submissions, and give projects a clear basis for handling AI-generated contributions.

Where this leaves ps5-linux

For now, ps5-linux's public repository remains available, with documentation covering PS5 Phat and Slim consoles running firmware 3.00 through 7.61.

Nguyen is no longer working on the project, and PS5 Pro support that he had planned for 2027 is no longer being developed by him. The public codebase could still be forked or extended by others, while Sony could also patch the underlying vulnerability and further restrict what is possible on affected firmware.

Whatever happens next, the episode adds another data point to a question more open-source projects are now being forced to answer: how much AI-assisted work can volunteer maintainers reasonably review, trust, and maintain?

For projects already dealing with limited reviewer time, that question is becoming harder to ignore.


Published via ZyVOP — Write once in Markdown, auto-backup to GitHub, and syndicate to Dev.to, Medium & Hashnode in 1 click.

Top comments (0)