I started to lose my ability to code : Developers grapple with the real cost of AI programming tools

Maybe it started in February, when programmer and entrepreneur Paul Ford, 51, wrote a guest essay for The New York Times gushing that “the A.I. disruption has arrived.” AI coding tools are getting stronger — and so are the hopes and fears of the programmers they’re displacing. Online, a messy, real-time reckoning is already underway.

In his essay, Ford applauds the possibilities unlocked by faster, cheaper vibecoding: “It’s fun to see old ideas come to life,” he writes, revisiting long-abandoned projects on the subway.

And in mid-March, tech author Clive Thompson saw his interviews with more than 70 software developers at Google, Amazon, Microsoft, and start-ups published in the Times‘ magazine. The consensus was “most were weirdly jazzed about their new powers. I was surprised by how many software developers told me they were happy to no longer write code by hand. Most said they still feel the jolt of success, even with A.I. writing the lines.”

But others continued expressing deeper concerns. Pia Torain, a software engineer for Point Health A.I., told Thompson that after four months of issuing hundreds of prompts a day, she’d “started to lose my ability to code.” Torain now makes a conscious effort to slow down and absorb the program’s entire architecture and flow, warning that this may be the ultimate danger of outsourcing too much of our programming to AI.

“If you don’t use it, you’re going to lose it.”

“If you don’t use it, you’re going to lose it,” Torain told Thompson. And this creates an even bigger dilemma for junior developers, as captured succinctly by a recent comment on Hacker News. “If nobody’s hiring junior devs because LLMs can do junior work faster and cheaper, how is anyone going to become an expert?”

See also: Microsoft execs warn agentic AI is hollowing out the junior developer pipeline

That concern is playing out in multiple ways. “At one point, I thought when I got older it might be nice to teach,” muses a commenter in a Hacker News thread. “Students have infinite teachers on YouTube, and now they have Gemini/Claude/ChatGPT, which are amazing.”

Already, it feels like there’s not much left for a human teacher to contribute, they believed, and with the rate at which tools are improving, that could drop to zero within two years.

Coding faster brings its own problems — including concerns about code quality and questions about whether expertise will be more or less valuable now in a world filled with powerful AI programming agents — as well as a heightened risk of developer burnout.

As new tools transform our world, those conversations are getting louder, with top publications and online discussions capturing disruptions from multiple perspectives, showing an industry now reckoning with new AI coding tools and the transformations they bring — both the bad consequences and the good.

Passion and speed vs. quality and expertise

The discussion erupted in full force earlier this month on Hacker News when a 60-year-old programmer posted that Claude Code had “re-ignited a passion.” That thread drew 1,086 upvotes — and 989 comments with more stories and anecdotes, but also some debate and concern.  (Days later, a kind of rebuttal was posted, arguing that Claude Code had “killed a passion.”)

In the last few months, more powerful AI tools have clearly boosted the output of some programmers — especially those later in their careers. Among the 989 commenters was programmer John Calhoun, describing himself as an old Macintosh shareware author, who started at Apple in 1995, and admitting he’d “vibe-coded a website that I would not have otherwise attempted… And I have already queued up a couple of my abandoned projects…”

The resurrection of abandoned projects seems to be happening around the world. In Dresden, 62-year-old programmer Reini Urban said LLM coding agents “ignited my passion again.  I’ll soon unarchive older projects that were too hard to continue. With Opus, this will finally be doable. ” And 51-year-old Boston electrical engineer/entrepreneur John Reine posted that “It’s given me the guts to be a solo-founder.”

But among those 70 software developers interviewed by Thompson was also a passionate minority who, he said, now “actively avoid” AI tools, including an anonymous Apple engineer who still wanted to do all of their own coding. “I believe that it can be fun and fulfilling and engaging, and having the computer do it for you strips you of that,” they’d said, insisting they didn’t want to “outsource” their passion.

There are also concerns about code quality. As one Hacker News commenter put it, “LLMs are quite good at coding, but terrible at software engineering… At the moment, I am trying to fix a Vibe-coded application, and while each individual function is ok, the overall application is a dog’s breakfast of spaghetti, which is causing many problems.”

And one of the coders commenting was Joel Dare, a 50-year-old software engineer in Utah, who complained that after 40 years in this industry, “I’ve developed a low tolerance for architectural decay.”  So when he’d forgotten to tell Claude not to use frameworks for a Node project, he’d ended up with 89 dependencies.  “In a world where we prioritize ‘velocity’ over maintenance, this is the status quo. For me, it’s unacceptable. I’ll try again, but we NEED to expertly drive these tools, at least right now….”

Several commenters said in the end, AI tools are simply making their expertise more valuable, with one principal SWE believing AI “magnifies the thing I do well” — architecture, debugging, and “making good technical decisions.”

And 63-year-old Chris Marshall, who’s been coding since 1983, said, “I do feel as if my experience… is crucial to using the LLM to develop something that can be shipped. I’ve had to learn how to work with an LLM, but I think I’ve found my stride.”

There was an interesting analysis from software engineer/founder Juan Reyero, who’s been programming for 40 years, and sees this debate diverging into two camps. “I think that the biggest difference is between people who mostly enjoy the act of programming (carefully craft beautiful code….), vs the people who enjoy having the code done, well structured and working, and mostly see the act of writing it as an annoying distraction.”

As Dare put it, AI coding “still requires our expertise to guide it.”

Although he’d also added, “I’m not sure if that will be the case in a year, but it is today.”

Good news, bad news

Perhaps no one typifies the mixed responses to AI coding tools like 57-year-old software engineer Steve Yegge, who commemorated the new enthusiasm that seemed to be spreading as he approached his 57th birthday in January marveling on Medium that he’s “cranking out thousands of lines of production code per day… and generally having an absolute blast.”  (In the Times‘ article Yegge says he’s now 10x, 20x, “even 100 times as productive as I’ve ever been in my career… It’s like we’ve been walking our whole lives.”)

Yegge’s LinkedIn profile now gives his occupation as “AI babysitter,” as he savors his abilities increasing exponentially thanks to powerful swarms of AI agents and orchestrators.

But for Yegge, the thrill of all this productivity is that it even has unintended side effects. Yegge and two of his co-workers decided that “High-end vibe coding is fucking with our sleep cycles,” Yegge writes.  (“This wasn’t happening last year. It wasn’t until we’d started working with a dozen or more agents at once and doing swarming of large piles of work…”)

At his startup’s new AI agent platform Gas Town, “Our hypothesis is that we’re operating at such a high level of decision-making that we’re exhausting some internal buffers, and need to grab some gradient-descent time before we can continue.” But “To me it feels like … We’re finding it to be exhausting.”

A boon for seniors?

Maybe this newfound ability is especially meaningful to senior programmers. Kent Beck, a 64-year-old programming guru, even told Thompson he’d mostly stopped coding 10 years ago, frustrated with languages and tools, but LLMs got him programming again.  (Acknowledging that sense of fun, Beck said AI’s unpredictability “is addictive, in a slot-machine way.”)

Chris Marshall, 63, even sees it addressing a silent ageism. “The thing that most upset me, since retirement, has been the lack of folks willing to work with me,” Marshall posted in the online discussion. “I spent my entire career working in teams, and being forced to work alone reduced my scope. I feel as if LLMs have allowed me to dream big, again.” While he remains a retired coder, he writes, “I’m enjoying having an LLM pair partner.”

As the comments piled up on Hacker News, some senior coders even went into specifics on what draws them to AI-assisted coding. Douglas Tarr, 52, writes that coding “manually” brings back memories of 12-hour days and “makes me tired just thinking about it… I’m too old for that now, my back hurts if I sit too long, and occasionally I get migraines if I look at a screen too much…”

But most heartening of all was the response from Minnesota-based programmer TQ White, who posted that he’d written his first program in 1967 — and that AI coding tools addressed his biggest issue. “The isolation of being a retired programmer is a real bitch…”

White joked that “I’m not allowed to feel like AI is an adequate replacement for fear that the critics will tell me I’m not healthy…”  But then he added that “between you and me, as much as I miss the camaraderie of real humans, being able to brainstorm with an entity that knows pretty much everything and is able to execute my will without complaint is not bad.”

“And, it’s nice to have someone, something, to talk to about technical ideas. It’s a great time to be alive…”

TRENDING STORIES

Group Created with Sketch.