The Software Engineer Who Codes on His Commute to Stay Sharp in the Age of AI
AI tools are writing more code at more companies. Some engineers are quietly going back to basics to make sure they still can.
Key points
- Software engineering ranked among the highest-paying US professions in 2022, before widespread AI adoption began reshaping the field.
- One engineer says his role shifted noticeably toward reviewing AI-generated code over the past six months.
- He now writes code by hand on his daily commute, not for work, but to keep skills he fears are fading.
- Engineers across the industry are weighing whether AI tools help them or quietly hollow out their expertise.
Every weekday, a software engineer who goes by Matt boards a train to Pawling, New York. He opens a side project: a browser-based video game he codes entirely himself. Not because his employer asked him to. Because he's worried about what happens if he stops.
"I am actively trying to keep my axe sharp," Matt told The Guardian, asking that his real name not be used to protect his job.
He has good reason. Over the past six months, Matt says his day job has quietly shifted. Less writing code, less solving problems from scratch. More reviewing code written by AI, meaning output that a large language model, the same underlying technology behind ChatGPT and Claude, generated automatically. His role went from builder to checker.
That shift troubles him. Skills, like muscles, weaken without use.
"I am trying not to use AI where I can," he said.
Matt isn't alone. Software engineering was one of the best-paid professions in the United States in 2022. Then AI coding tools arrived at scale. Layoffs followed at major tech firms. Hiring slowed. Engineers who kept their jobs found their roles changing under them.
The pattern is real. What's harder to measure is what it costs an engineer to spend their days reviewing AI output rather than producing their own.
Is this actually a threat to your career?
For many engineers, yes. The honest answer depends on your employer and your role. Engineers whose value sits mainly in writing routine code face the clearest pressure. Those who design systems, catch the subtle errors AI tools routinely miss, or explain technical decisions to non-technical colleagues are in a stronger position.
Survivorship bias deserves a mention here. Engineers who tell confident stories about thriving alongside AI are visible. Those who quietly left the industry, or never broke in, aren't.
Matt's train-commute strategy is low-tech and unglamorous, and it's also sensible. He's treating his core skill like a second job, protecting it from the drift he sees at work.
If you're in any knowledge-based role watching AI absorb pieces of what you do, the move is straightforward: find the parts that still require a human to do them well, and do more of those deliberately, even if no one is asking. Not a course, not a certification. Practice, on purpose, before the skill quietly disappears.
