In early August 2026, Linux kernel maintainer Greg Kroah-Hartman removed a 27-year-old driver for discontinued industrial hardware — not because it was broken, but because AI scanning systems had begun generating a steady stream of false findings from its dormant code. The decision had already been sanctioned by the hardware's manufacturer, Moxa, months prior; what AI provided was not wisdom, but friction — enough low-level noise to transform a deferred obligation into an immediate one. It is a quiet parable about how automated systems may reshape the economics of neglect, making inaction more
AI Scrutiny Accelerates Removal of 27-Year-Old Linux Driver
AI made keeping it around more expensive than removing it
So AI didn't actually find a bug in the Moxa driver?
Not a real one. It flagged things that looked suspicious to an automated system, but in code nobody was using anymore, against hardware that doesn't exist. The findings were technically interesting but practically meaningless.
Then why did Kroah-Hartman care?
Because the findings kept coming. AI agents don't stop scanning just because you tell them the code is dead. They keep poking at it, keep flagging things, and every flag requires a human to look at it and say "this doesn't matter." That's the friction.
Moxa had already agreed to removal though?
Yes, back in May. They said the hardware was discontinued, the software support was over, and they didn't need the driver anymore. But the removal didn't happen until August, when the AI noise became loud enough that acting became easier than waiting.
Is this a story about AI being useful, then?
It's a story about AI being useful in a way nobody planned for. The driver needed to be removed anyway. AI didn't discover that. But AI made the removal urgent by creating a cost to keeping it around.
What happens next?
That's the real question. If AI starts making maintenance decisions across the kernel just by being annoying enough, that changes how open source works. Not because AI is smarter, but because it's louder.
O Pulso
- AI agents scanning the Linux kernel began flagging potential bugs in a driver tied to hardware that ceased to exist in 2017, creating a stream of false positives that demanded real human attention.
- Each automated finding required a maintainer to evaluate and dismiss it — a mounting tax on human time paid against code that served no living user or machine.
- The removal had already been agreed upon: Moxa confirmed in May 2026 that both hardware and software support were fully retired, and the company even offered to help write the deletion patch.
- Yet three months passed before Kroah-Hartman acted, suggesting that consensus alone was not enough — it took the ongoing friction of AI noise to convert a known task into an urgent one.
- With 2,148 lines now gone, the kernel is lighter, but the episode raises a larger question: will AI code analysis become the force that finally clears open source's vast accumulation of technical debt?
In early August 2026, Linux kernel maintainer Greg Kroah-Hartman removed a 27-year-old driver for discontinued industrial hardware — not because it was broken, but because AI scanning systems had begun generating a steady stream of false findings from its dormant code. The decision had already been sanctioned by the hardware's manufacturer, Moxa, months prior; what AI provided was not wisdom, but friction — enough low-level noise to transform a deferred obligation into an immediate one. It is a quiet parable about how automated systems may reshape the economics of neglect, making inaction more costly than resolution.
Greg Kroah-Hartman, the Linux kernel's chief maintainer, deleted the Moxa Intellio serial driver in early August 2026 — 2,137 lines of code that had handled multiport serial connections for industrial hardware and had gone untouched for nearly three decades. The hardware it served was discontinued in 2017, and by 2021 Moxa had formally told kernel maintainers that software support was no longer needed. For five years, the code simply sat there, inert.
What finally forced the issue was not a bug or a security flaw, but noise. As AI systems began scanning the kernel more aggressively, they started surfacing "interesting" findings in the Moxa driver — edge cases and apparent anomalies that looked meaningful to an automated tool but were irrelevant in practice, since no hardware existed to test against. Each finding still required a human to evaluate and dismiss. Kroah-Hartman's response was pragmatic: remove the code entirely, eliminating the source of the false positives.
The removal had a long paper trail. Jiri Slaby had proposed dropping the driver as early as 2021. In May 2026, during a tty subsystem reorganization, Moxa's Crescent Hsieh confirmed the hardware and its support were fully retired — the company even offered to help write the patch. But nothing moved until August, three months after Moxa had already signed off.
The episode illuminates something true about how open source maintenance actually functions. The decision was already made; what AI did was make delay more expensive than action, converting a someday task into an urgent one through sheer accumulated friction. Kroah-Hartman left a theoretical door open for a modern rewrite, but realistically, the Intellio hardware is gone and the company has moved on. What lingers is the broader question: whether AI code analysis will begin forcing similar reckonings across the open source ecosystem — not by being right, but by being relentless.
Greg Kroah-Hartman, the Linux kernel's chief maintainer, deleted a driver in early August that had been sitting in the codebase untouched for nearly three decades. The Moxa Intellio serial driver—2,137 lines of code that handled multiport serial connections for industrial hardware—was removed not because anyone was using it, but because AI agents had started scanning the kernel and flagging potential problems in dead code that nobody could actually test or fix.
The driver itself was never the problem. Moxa, a Taiwanese company that manufactures industrial networking equipment for factories and railway systems, had discontinued the Intellio hardware back in 2017. By 2021, the company had formally told kernel maintainers they no longer needed software support for the line. For five years after that, the code just sat there, inert, part of the kernel's accumulated technical debt.
What changed wasn't the code. It was the noise. As AI systems began scanning the kernel codebase more aggressively, they started finding "interesting" things in the Moxa driver—potential issues, edge cases, things that looked like bugs to an automated system. None of it mattered because the hardware was gone and nobody was using the driver. But the findings kept coming, and they kept requiring human attention to evaluate and dismiss. Kroah-Hartman's solution was straightforward: remove the whole thing. If there's no hardware, no users, and no maintainer willing to modernize the code, why keep it around just to feed AI scanners false positives?
The removal itself was not sudden. Jiri Slaby, who has maintained parts of the kernel's serial driver subsystem for years, had proposed dropping the Moxa code as far back as 2021. In May 2026, when a separate reorganization of the tty subsystem brought the topic back up, Crescent Hsieh from Moxa confirmed that both the hardware and its software support were completely phased out. The company even offered to help write the removal patch. But nothing happened until August, when Kroah-Hartman finally acted—three months after Moxa had already signed off.
The timing matters because it reveals something about how open source maintenance actually works. The decision to remove the driver was already made. The manufacturer had agreed. The maintainers knew it was dead weight. What AI did was create enough friction—enough false findings, enough scanning activity, enough low-level annoyance—that the removal became urgent rather than someday. It wasn't that AI discovered the driver should be removed. It was that AI made keeping it around more expensive than removing it.
Kroah-Hartman left the door slightly open, noting that the driver could theoretically return in the future if someone rewrote it in a more modern way. But realistically, that's not happening. The Intellio hardware is gone. The company that made it has moved on. And now the kernel is 2,148 lines lighter, freed from the burden of maintaining code for hardware that stopped existing nearly a decade ago. What remains to be seen is whether this becomes a pattern—whether AI code analysis will start forcing maintenance decisions across the entire open source ecosystem, not because the decisions are wrong, but because the noise becomes unbearable.
Citações Notáveis
This is a very old driver, and no known hardware is still around for it, and the company says they do not need it anymore, so let's remove it as the LLMs are starting to poke at it and find interesting things which will just waste everyone's time given it's not actually used.— Greg Kroah-Hartman, Linux kernel maintainer