‘I use AI to do the things that I’m bad at’: Linus Torvalds on why it works for him

Linus Torvalds said AI can make programming more enjoyable for beginners and help identify and fix Linux kernel bugs, including security issues. He cautioned that fabricated reports and low-quality patches can burden maintainers; some subsystem maintainers require Sashiko AI reviews, though there is no universal requirement. The article also recounts Linux’s shift to frequent releases and Git, emphasizing incremental development and a stable process.

Analysis

The investable signal is not that AI can generate more code; it is that review capacity is becoming the scarce input. If automated review finds more genuine defects but also sends maintainers low-quality reports and narrow fixes, the net effect may be slower patch acceptance rather than faster security improvement. That creates a verification tax: projects and vendors with trusted maintainers, triage processes, and enterprise support could gain relative advantage over tools optimized mainly for code output.

For IBM (Red Hat) and Linux-dependent cloud operators such as Amazon, Microsoft, and Alphabet, improved kernel hygiene could modestly reduce long-run security and reliability tail risk. But benefits are diffuse, and the article provides no evidence of changed spending, pricing, or near-term earnings. It does not support a sector-wide AI trade.

Over days, sentiment around AI-assisted development may help coding-tool narratives, but this discussion is unlikely to move fundamentals. Over 1–3 months, watch whether AI review becomes a required gate across more kernel subsystems and whether maintainers report higher accepted-fix rates without longer review queues. Over 6–18 months, successful workflow integration could improve software supply-chain security; persistent reviewer overload could instead discourage contributors and delay fixes. The contrarian point: more detected bugs are not automatically more delivered security. No trade on this item alone; a sustained rise in accepted fixes alongside stable review latency would strengthen the positive thesis, while growing queues or repeated reverts would falsify it.

Key Decisions for Investors

• No immediate position: the signal is operational and long-dated, with no demonstrated revenue or guidance impact.
• Set a 1–3 month watch on kernel maintainer adoption, accepted AI-assisted fixes, review latency, and reverts; treat patch volume alone as a poor success metric.
• For IBM and Linux-dependent cloud operators, view better kernel maintenance as a conditional reduction in operational tail risk, not a near-term earnings catalyst.
• Reassess the thesis if AI-assisted reviews reduce triage time and raise accepted-fix throughput; turn cautious if maintainer backlogs expand or low-quality submissions increasingly delay security fixes.