The Year Software Broke: Graydon Hoare on LLM's Sudden Transformation of Programming

The software industry has experienced what may be its most dramatic transformation in decades, according to Graydon Hoare, creator of the Rust programming language, who describes the past year as witnessing changes more severe than the arrival of the internet or cloud computing.

The Sudden Shift

Writing from March 2026, Hoare characterizes late 2025 and early 2026 as "extremely unusual times" where large language models "turned a corner" in capabilities that caught even experienced developers off guard. The progression he describes is striking: from developers "playing around seeing what I can do" to teams now facing "hundreds of new security vulnerabilities" they can barely keep up with.

The coding capabilities first became apparent around the release of major model versions like Opus 4.5, Gemini 3, and GPT 5.2. But the more troubling development was discovering that LLMs excel even more at finding vulnerabilities than writing code. "They can break software even better than they can write it," Hoare notes, explaining that vulnerability hunting requires being "right sometimes" rather than consistently correct.

The New Reality of Development

For Hoare personally, the change has been profound. He still writes code but "less and less," reserving it for "touchups, sketches of APIs and data structures, subtle stuff it's easy to be subtly-wrong about." The bulk of main logic is now handled by LLMs that "really does often write the main logic as well as I would at this point, and faster, and more persistently."

This isn't just about convenience—it's about survival in what he describes as an "arms race." The unit economics of software development have shifted completely. Teams face pressure to deliver 100x more code while simultaneously defending against AI-generated attacks. The velocity is unsustainable, leading to burnout and a sense that the entire profession has been upended.

Community Fracture

The transformation has torn the software community apart. Projects are responding in contradictory ways: some embracing AI tools, others rejecting them entirely. Issue trackers have closed to new submissions deemed "slop." Maintainers have quit, contributors banned, dependencies rolled back or severed, and forks emerging.

Hoare describes "even more fury" than during the drafting of GPLv3 as people debate licenses and rebuild entire software stacks. The anger he witnesses is "unlike anything I've ever seen before." People are making lists of who's "infected by the menace and who's still clean." Professional relationships, friendships, and collaborations have ended as battle lines form.

The Human Cost

The emotional toll is evident throughout Hoare's reflection. He acknowledges thinkpieces about "mourning their loss of identity as programmers" and fear about job losses. There's personal disgust with the quality of AI-generated code, fury at the billionaires funding this transformation, and a sense that it's all part of "the fascist turn of America."

Yet Hoare positions himself as caught in the middle—someone who uses LLMs "day-to-day, because they seem to work" while feeling "reluctant but fatalistic" about it. He's tired of being heckled and has turned off comments on his post, acknowledging he's likely to be seen as "infected" or "facilitating the fascists."

Uncertain Futures

The uncertainty about where this leads is palpable. Hoare lists possible outcomes: everyone out of work, nuclear war, even more burnout from sustained high-velocity development, or the whole thing collapsing as "all just a bubble." He admits he couldn't have predicted this year, so he makes no predictions about next year.

What's clear is that 2025-2026 represents a discontinuity in software development history. The changes feel "very sudden and very severe" to those experiencing them directly. For developers still operating outside this transformation, Hoare suggests they may be "lucky, perhaps unlucky" but are definitely "operating in some level of ignorance of what has occurred."

A Profession in Transition

The most striking aspect of Hoare's account is the speed and completeness of the transformation. Nothing in his decades of experience—not the internet, open source, distributed version control, or cloud computing—compares to this shift. The software industry has moved from debating whether AI will replace programmers to grappling with a world where AI both writes and breaks code at unprecedented scales.

As Hoare concludes, he's "just trying to adapt" while missing the past and hoping people can see him as human again someday. His journal entry captures not just a technological shift but a moment of collective trauma for a profession suddenly confronting its own obsolescence—or at least its fundamental transformation into something unrecognizable from what came before.