Linux Kernel is Now Vibe Coded

Core claim: AI-assisted (“vibe-coded”) kernel development

The speaker argues that Linux kernel development is rapidly shifting toward AI-generated (“vibe-coded”) patches. They estimate that by July/August (or sometime later within the same year), most new patches could be predominantly AI-developed.

How to verify the claim (evidence source)

They point viewers to lore.kernel.org, a Linux kernel mailing list archive, and suggest searching for patches/commits that include an “assisted-by” tag, such as:

• “assisted by Claude/Codex/Grok”

They note that a substantial portion of results appear to be individual patches rather than discussions, implying the tag is being used frequently and can be quantified.

Observed trends / rough metrics

• Double-digit AI-assisted share (recent snapshots): In a recent snapshot (around a May 22 recording), they report double-digit percentages of patches over the last week being AI-assisted, rather than just 1–2%.

• Projection to “over 50%” by August: If the current growth continues (while acknowledging month-to-month variability), they project that by around August the share could reach over 50%.

• Very fast patch arrival times: They highlight that patches sometimes land within minutes of each other, describing the ongoing activity as “fast and furious.”

Security concern / quality risk

The speaker claims they’ve seen increasing bug reports generated or facilitated by AI, including zero-day-style vulnerabilities that can lead to local code execution.

They also acknowledge that some issues get fixed with AI assistance, but overall they treat the trend as concerning—presenting it as “red flags.”

Testing AI coding systems (“up to snuff?”)

They describe running small test projects to evaluate AI code generators for:

1. Large-scale comfort
2. Suitability for Linux kernel development

Their experiments often involve generating code for retro/obscure platforms, such as:

• Lotus 1-2-3
• WordStar
• Languages like C or AppleSoft Basic

They report that the AI worked surprisingly well, which both delighted and concerned them—raising worries about long-term impacts on developer skill and understanding.

Industry / organizational incentives

They argue the shift is being accelerated by institutional alignment, including:

• The Linux Foundation partnering with AI companies
• Red Hat reportedly paying employees more if they use AI for Linux development (and possibly penalizing non-use)

Their conclusion: even if individuals disagree, AI-driven development is likely the direction Linux is taking.

Overall takeaway / framing

The speaker is skeptical and raises concerns about:

• Additional security vulnerabilities
• Erosion of developer expertise
• Whether any benefits will outweigh the risks

They state they haven’t seen clear positives yet beyond issues being found and fixed with AI assistance.

Main speaker / source

• Lunduke / The Lunduke Journal (primary speaker; references lunduke.com and mentions Lunduke Journal subscribers)

Source referenced for verification

• lore.kernel.org (Linux kernel mailing list archive)
• The “assisted-by” tag in kernel patch submissions