How Software Engineers Actually Use AI – WIRED

This post was originally published on this site.

Code created AI. That much is obvious. Less obvious is the extent to which AI is now, in a snake-eats-tail way, creating code. We kept hearing conflicting accounts. This programmer used AI every day; that programmer wouldn’t touch the stuff. This company paid for AI services; that company banned them. So which is it? Are chatbots liberating human programmers—or programming them right out of their jobs? To find out, we blasted a survey to every software engineer and developer in our orbit, from casual dabblers to career vets. The results amazed and surprised us, capturing an industry at fundamental odds with itself. And did we then upload that data to ChatGPT, just to see what it might make of things? Yes. Yes we did.*

*ChatGPT mostly did spreadsheet-processing grunt work, offered trend analyses, and made the occasional design recommendation. Human editors and fact-checkers were there every step of the way (and needed to be).

Will AI Fully Eat Programming Jobs One Day?

Almost every coder we surveyed had strong opinions on the matter. Here’s ChatGPT’s summary of the responses (with its boldface emphasis preserved):

“The coders have spoken—and they’re not packing up their keyboards just yet. While a small but vocal group insists AI will devour programming jobs in time, most dismiss full automation as a pipe dream. The doom prophets warn that corporate bosses will slash payrolls the moment AI looks capable, leaving human engineers debugging their own obsolescence. The skeptics scoff, arguing AI is more like a hyperefficient intern—useful, but clueless—that can’t handle context, edge cases, or real problem-solving. The realists see AI as a force multiplier, not a job killer—automating repetitive coding but leaving the creativity, architecture, and debugging to humans. “If AI does eat programming,” one put it, “I’ll just switch to debugging AI.”* The real verdict? AI isn’t coming for your job—but it is changing it. Adapt or get left behind.🚀”

*Nobody actually said this! ChatGPT—bless it—made several mistakes throughout this process. Not only did it fabricate quotes, it also misread results, generated incomprehensible graphs, and, at multiple points, stopped counting “freelancers” as a category entirely. Rude.**
**In its way, ChatGPT invited this feedback. When asked for final tips, it suggested “a touch of snark, because AI discourse is full of strong opinions.”