{"slug": "the-rapture-of-the-programming-languages", "title": "The Rapture of the Programming Languages", "summary": "Generative AI coding assistants are making programming language popularity less relevant, as tools like Claude Code can work with older languages like Perl if sufficient documentation exists. The author's experience with BumpySkies shows that AI can extend legacy codebases, reducing the need to port projects to more popular languages. This shift challenges traditional metrics of language vitality, such as CPAN contributions, which have declined for Perl despite its continued utility.", "body_md": "The pronouncement of certain programming languages as “dead” or “dying”—as has long been the prognosis of Perl, my own home-base language—has become far less relevant in the era of AI-assisted coding. The nonliving and tireless entities that generate an ever-increasing share of today’s new code have become profound mechanical experts in any sufficiently well-documented programming language you could care to ask for—even the “dead” ones, if enough public examples of their code exists.\n\nWhen I started working on BumpySkies again as a genAI-assisted project last year, I gave Claude Code clearance to port it into Python or Node, which tend to be that tool’s default languages for new web-based projects. To my surprise at the time, it balked: sure, it could do that, Claude replied. But the huge amount of Perl I’d hand-written for BumpySkies over the years was perfectly serviceable, and Claude foresaw no issues in simply extending it further using that language, which I was clearly quite comfortable reading. Wouldn’t I prefer that, if I planned to stay abreast of the project’s growth and changes? I responded: yes, I would prefer that.\n\nAnd in all the months since, I’ve had no regrets about this decision. This is true for personal readability reasons, just as predicted. And it’s also true because whenever Claude can’t find a suitable module on CPAN—the public repository of community-maintained Perl modules, and the precursor to other languages’ resources like npm and PyPI—it simply shrugs and codes up the missing library itself, just for the project.\n\nThis ability of LLM-driven coding assistants makes inconsequential one of the most profound markers of Perl’s decline: year over year, community contributions to CPAN have been trending sharply downwards, even as the amount of stuff in the world that people could be writing Perl modules to interact with keeps increasing.\n\nIn the early days of the web, Perl took off in part because CPAN existed at all; if you wanted to program something back then, you could probably find a bunch of CPAN modules that would get you most of the way there, and maybe you could contribute one yourself based on your finished project. This is still true, to an extent. CPAN remains a rich source of stable legacy tools and extensions for Perl, and it continues to receive thousands of new releases every year. But if you run the cpan-release-counts script to display a graph counting per-year CPAN releases from 1995 through the present, you get a tidily triangular image, with its peak in 2014. It’s hard to look at the graph and not see hard evidence that, after the 2010s’ various philosophical transformations about the way that both websites are made and ops scripting is done, Perl plays a significantly diminished role in the digital landscape.\n\nBut another transformation has occurred since then, of course: a very strange one. And, strangely, it feels like the choice of language matters less now. Or, perhaps better phrased: you no longer need to hold a language’s current global popularity as a primary factor when choosing a new project’s language, to say nothing of choosing whether to port a legacy project to a different language. If you use a generative AI coding assistant, and if your preferred programming language has enough public documentation and examples to have thoroughly trained the assistant’s underlying LLM, then the language’s readability suddenly becomes at least as important a criterion as popularity. (This assumes that you intend to read, review, and actively steer all of the code that the bot generates, which I see as the key behavior that separates genAI-using engineers from the cynical vision of “vibe coders”.) A well-trained LLM paired with a well-exercised language can generate the software that you want in that language, even if relatively few human coders would have chosen to work with it today.\n\nAnd so, strangest of all, Perl—a living fossil among programming languages, sidelined by the myriad revolutions of React, Docker, Ansible, and so on—has transubstantiated into something eternal, entirely through the intercession of a new, weird, world-eating technology. Whether Perl is “dying” has become moot. Like an uploaded-mind character in a millennial science-fiction novel, it now resides forever—or at least for as long as generative AI remains widely accessible—in the proto-noosphere summoned into existence over all those roaring, thirsty data centers. And the uncanny entities that haunt it can write all the brand-new, expert-quality Perl code that you want, based on an eerily unchanging concept of the language and its practices.\n\nI suspect that Perl, among all obscure or faded programming languages, is uniquely positioned here because of its past dominance as “the duct tape of the internet”, during all those years on the left side of that triangular graph. The Perl project—and all the projects built on it—generated a vast amount of publicly available code, books, articles, and high-quality documentation that still rivals any of the technologies that have largely supplanted it.\n\nClaude Code and its colleagues, in imitation of contemporary human coding practices, will generally choose a currently popular language for a new project—unless you ask it to use something else, in which case it will comply as best it can, brimming with artificial cheer and optimism. I don’t suggest that this is always a good idea, but in my case—a single coder whose long professional history has given me strongest affinity with one particular archaic language—it has become, unexpectedly, weirdly, a viable one.", "url": "https://wpnews.pro/news/the-rapture-of-the-programming-languages", "canonical_source": "https://fogknife.com/2026-06-19-the-rapture-of-the-programming-languages.html", "published_at": "2026-06-19 17:05:23+00:00", "updated_at": "2026-06-19 17:07:47.426067+00:00", "lang": "en", "topics": ["large-language-models", "generative-ai", "ai-tools", "developer-tools"], "entities": ["Perl", "Claude Code", "CPAN", "BumpySkies", "npm", "PyPI", "LLM"], "alternates": {"html": "https://wpnews.pro/news/the-rapture-of-the-programming-languages", "markdown": "https://wpnews.pro/news/the-rapture-of-the-programming-languages.md", "text": "https://wpnews.pro/news/the-rapture-of-the-programming-languages.txt", "jsonld": "https://wpnews.pro/news/the-rapture-of-the-programming-languages.jsonld"}}