First screen, captured 9 days ago. Open the video.

A thousand fewer paths to root

What HN thinks

A thousand fewer paths to root, and roughly a thousand comments on whose fault the remaining paths are.

Kernel bugs were always there
Many argue local privilege escalation bugs in a huge C kernel have long been assumed, and AI just finds them at a new scale and cost.
Blame language and architecture
Others say memory-unsafe C, a monolithic design and in-kernel drivers make this inevitable, pointing to Rust, microkernels, seL4 or OpenBSD-style minimalism.
Skeptical of OpenAI's numbers
Some want breakdowns by subsystem, bug type and code age, and suspect marketing inflation or reward hacking without published details.

Consensus: Most agree large C codebases like Linux are riddled with exploitable bugs and that relying on user separation alone isn't real security; there's no agreement on the fix.

Recommended in the comments

Rust (tool): Commenters argue it would eliminate whole classes of kernel memory bugs.
seL4 (product): Cited as a small, verified kernel that resists these attacks.
OpenBSD (product): Held up as having a much smaller attack surface, though disputed.
Sashiko ↗ (service): Mentioned as per-commit AI review to catch new vulnerabilities.

Summarised by Claude from the comment thread 8 days ago.

Points: 455
Comments: 331
Best rank: #2
Posted: 2026-11-04 by dangelosaurus
Category: Security

linux-kernel privilege-escalation vulnerability-discovery openai ai-agents exploit-development patch-generation

All captures:

Full page

Captured 9 days ago. Click for the whole screenshot.

Looks similar

Screenshot of blog.zortax.de: A Critique of GPUI

A Critique of GPUI (blog.zortax.de)

117 points by ByteAtATime 4 days ago | 91 comments

A blog post critiquing the state models of Rust GUI frameworks GPUI and egui, arguing with benchmarks that signal-based run-once components scale better for complex desktop apps.

rust gpui egui gui-frameworks

Screenshot of perplexity.ai: Perplexity Decider v1.3: Training, eval, and use cases

Perplexity Decider v1.3: Training, eval, and use cases (perplexity.ai)

94 points by gmays 5 hours ago | 19 comments

Perplexity Research blog post announcing Decider v1.3 and Decider Flash v1.3 open-weight decision/classifier models, explaining their training, evaluation, and internal use.

perplexity decision-models classification model-distillation

Screenshot of lesswrong.com: I struggle to believe that AI R&D is ingenious

I struggle to believe that AI R&D is ingenious (lesswrong.com)

136 points by porridgeraisin 7 days ago | 155 comments

A LessWrong shortform comment by Arjun Panickssery arguing AI agents show poor research judgment in autonomous R&D side projects like a Scrabble engine and philosophy research.

ai-research-agents autonomous-r-and-d llm-capabilities lesswrong

Screenshot of fzakaria.com: A Clock Made of Branches

A Clock Made of Branches (fzakaria.com)

3 points by setheron 11 days ago | 0 comments

Blog post explaining why the Rewind deterministic VM measures guest time by counting retired conditional branches instead of instructions or wall-clock time.

deterministic-replay rewind-vm performance-counters rr-debugger

Screenshot of gyptazy.com: Live-migration support in FreeBSD and bhyve coming

Live-migration support in FreeBSD and bhyve coming (gyptazy.com)

64 points by JohnDohnBohn 5 hours ago | 15 comments

Blog post presenting a work-in-progress RFC implementation of live VM migration for bhyve on FreeBSD 15.1, covering design, test results, and blockers to upstreaming.

freebsd bhyve live-migration virtualization

Screenshot of aworkinglibrary.com: Petered

Petered (aworkinglibrary.com)

156 points by adrianhoward 5 days ago | 90 comments

A reading note by Mandy Brown on what happens to people after the Peter Principle, drawing on Making a Career in Dictatorship and Bullshit Jobs, and on real imposters.

peter-principle management imposter-syndrome organizational-behavior