Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- AI coding tools act like power tools for programmers.
- Programming jobs will change, but not disappear entirely.
- New tester and AI-wrangler roles will grow alongside coders.
Something terrifying is happening in the world of programming. Demand for coders has collapsed. Up until this year, programming has been considered one of the most secure, predictable, and lucrative career options. But now, we’re seeing reports that employment for programmers has collapsed to its lowest level since 1980.
On the surface, the connection is obvious. AI agents are able to write code and do so much faster and cheaper than professional programmers. Code is structured text, something AIs are particularly well-suited to understand and reproduce.
Also: 10 things I wish I knew before trusting Claude Code to build my iPhone app
Therefore, it seems a foregone conclusion that with programmers being so expensive for companies to hire, and AI coding tools being so much less expensive, companies are going to replace all their coders with AIs.
If you’re a coder, goes the newly conventional wisdom, you better make sure your car works, because next year you’re going to be humping grocery deliveries for Instacart instead of writing code.
But, after vibe coding two fairly impressive projects, I believe the reality is far more nuanced. To understand what’s happening, and to be able to get a better prognostication for the future of programming, we need to look to history, and I need to tell you a story or two.
Growing up in the 1970s
Two of my earliest memories are of my mom with her sewing machine and my dad with his table saw. When I was growing up, my parents were young and far from well off.
Before my mom got her first sewing machine, she did her best to repair and fix all the damage I did to my clothes with a needle and thread. She did each stitch one at a time, and it could take her hours to shorten a pair of pants, sew up the front of a sock, or try to patch the knee of a shredded pair of jeans.
Eventually, she found a used sewing machine at a garage sale and went to town sewing her own clothes. She tried her hand at sewing clothes for me, but that didn’t go well. Kids at school mocked the shirts she did her best to make, and it took far too much work, even with her sewing machine, for her to continue, especially when I was so unhappy with (and, to my everlasting shame, ungrateful for) the results.
Also: How to install and configure Claude Code, step by step
My dad, as a young homeowner, took responsibility for trying to fix and improve the house. He had a variety of saws, ranging from the third-hand clunking and insane safety hazard that was his table saw, to a bunch of traditional hand pull saws.
I remember him making cabinets in the basement that had one redeeming value: they were robust. They weren’t plumb (straight up and down) or level, or complete with wall sheathing that fit properly. But they did hold stuff, so mission accomplished.
My mom eventually got better and better at sewing, to the point where she made part of her living as a sewing teacher. But she never upgraded that old sewing machine, and she never tried making me clothes again. My dad kept making catawampus home repairs, but they all mostly got the job done. He relied on his small collection of barely functional power tools to build and fix what he had to.
The transition to power tools
In 1755, Charles Fredrick Wiesenthal was awarded British patent number 701 for “a double pointed needle with an eye at one end.” This was the first step in the creation of a mechanical device for sewing.
It took until 1829 for an actual practical machine to be invented. It could only sew in a straight line. By the 1860s, industrial sewing machines were in factories, cranking out machine-made clothing by the thousands. It took until the late 1800s, when homes started to get electrical power, for home sewing machines to enter the market.
Even with the availability of sewing machines, sewists like my mom (who sewed both because she liked it and of economic necessity) relied on hand stitching as either their primary sewing practice or to augment what their machines could do.
This continues to today. My wife has a variety of sewing machines, as well as specialized devices (like a serger, which wraps thread around fabric edges to prevent fraying). I’ve seen her do some hand sewing, especially for smaller items like doll clothes, but she uses different types of machine sewing for other work.
Also: Reinventing your career for the AI age? Your technical skill isn’t your most valuable asset
The first power tools can be traced back to Bronze Age times, when early industrial pioneers strapped horses and cattle to central poles and used them to drive cutting and milling tools, including circular saw blades. History isn’t entirely clear about who invented the modern circular saw blade, but they were in fairly common use chopping logs in sawmills by the early 19th century. One reason many sawmills are located along rivers is that the early mills used water power to turn those blades.
The first modern hand-held power tool can be dated back to a trigger-switch drill, modeled on the shape of a gun, introduced back in 1917. Today, power tools are everywhere. They scale up to factory-sized monoliths that press out car bodies in a single stamp, down to the hand-held tools homeowners have in their workshops.
Although my woodworking skill isn’t much better than my dad’s was, all those years ago, I have a much larger collection of much more modern tools than he could ever dream of. I have a band saw, table saw, miter saw, jigsaw, and a couple of circular saws. I have a bunch of other power tools, both in hand-held form and bench-top varieties. And yes, I also have manual hand tools, because sometimes it’s easier to cut an edge with a Japanese pull saw than it is to use a powered jigsaw.
The connection between power tools and AI coding
There is a connection between my stories of sewing machines and workshop power tools, and the burgeoning new world of AI-assisted coding. In fact, after having completed two AI-assisted projects, I think there’s a direct line that can be drawn between the two.
Let’s start with the analogy, and then I’ll explore how that helps us predict what will happen with AI coding.
Programming, dating back to its earliest days, has primarily been the task of writing instructions to guide a computer on a line-by-line basis. One line saves a value to a memory location. Another line acts on that value. Yet another line moves or copies that value. And on and on and on.
The analogy to sewing and woodworking is clear. Coding, by hand, is done line-by-line. Sewing, by hand, is done stitch-by-stitch, each loop of the thread connecting another millimeter or so of cloth.
Also: Claude Code made an astonishing $1B in 6 months – and my own AI-coded iPhone app shows why
Woodworking, by hand, is done cut-by-cut, each pull or push of the saw separating more of the wood’s fibers, each chop of the chisel breaking off another sliver of wood, each turn of the hand drill pulling out another shard of wood from an eventual hole, each stroke of sandpaper shaving down the rough fibers on the surface of the wood.
Our programming languages have gotten higher level and more helpful over the years. One of my first programming tasks was toggling computer instructions into a PDP-8 mini computer by flipping switches on its front panel. My most recent non-AI assisted coding project involves using Python libraries to do graphical transformations, where one line of code accomplishes a Photoshop-level transformation.
But it is still line-by-line. Programmers have built automation capabilities to assemble and test code. And we’ve built enormous libraries of pre-written code (like the Python Pillow library I used to do my photo manipulation). But still, most coding has been sequencing those lines of text.
Over the years, many companies have tried to offer so-called no-code or low-code solutions. These have usually turned out to be form-building tools, where most of the user interface was based on some kind of form, and business logic was either accomplished by stitching together pre-designed blocks or writing small bits of connective code.
To extend the analogy, those no-code and low-code solutions are similar to kits, whether sewing kits for a doll dress or a building kit for wooden models. The kits give some creative freedom in choice of color and design, but provide constraints on what the actual final product is. The same is true of no-code and low-code tools, which are mostly suitable to applications involving data entry and reporting.
Also: The best free AI for coding in 2025 – only 3 make the cut now
Neither no-code nor low-code tools could possibly come close to the series of projects I’ve completed using AI coding. I’ve written WordPress plugins for managing guest access, blocking banks of IP addresses, analyzing visitor access and identifying threat behavior, and for blocking and defending against AI spiders. I’ve also written a powerful iPhone app that scans and writes NFC tags, takes photos, and indexes 3D printing spools, using a variety of fairly sophisticated programming techniques.
The WordPress plugins were written using OpenAI’s Codex, and the iPhone app was written using Claude Code, both agentic AI systems.
AI coding is much more open-ended. The AIs are the power tools, but you can build almost anything, limited by your time and skill.
Yes, skill.
If you really want to understand the “feels” of vibe coding, read this article by ZDNET alum Jason Perlow. While you’re there, go ahead and give his newsletter a subscribe.
As Jason wrote in his piece, and as I’ve discovered in my vibe coding projects, and as countless other developers have discussed in their explorations of vibe coding, there is still skill and work involved. In fact, it’s often harder to clean up a mess created by an AI than a mess in your own code, because you know each line you wrote, where the AIs often write based on their own internal basis for expression.
That brings us to the future. What will happen to coding jobs? Will AI take work from programmers? What does it all mean?
By looking at the evolution of power tool usage, I think we can derive some clues as to the future of coding in a world of generative AI.
Coding jobs will change, but not vanish
Furniture making used to be a small business profession. Many towns and villages had a furniture maker, who would make chairs, tables, and cabinets.
Examination of furniture that remains from the 19th century shows that some craftspeople were extremely detailed in their handiwork, while others used every tool at their disposal to reduce the production time. If you’re at all interested in the way production was done back then, take a look at Rex Krueger’s YouTube channel. He does fascinating dives into the production processes used in making individual pieces of historical furniture.
Also: You should still learn to code, says top Google AI exec – here’s why
Today, except for hobbyist artisan woodworkers and extremely high-end custom craftspeople, most furniture is cranked out in factories. However, those factories employ people who operate the machines, which are large-scale power tools. Many furniture factories still exist in the Southeast United States. Many are located in Asia.
Meanwhile, carpenters who build structural elements of buildings are in almost overwhelming demand. When we put a porch on the front of our house, we hired a team of carpenters who brought their own set of power tools to cut and shape the wood that eventually became our porch.
Sewists have followed a similar path. Sewing factories, both in the US and around the world, still employ people to operate the machines. Unfortunately, in the quest to reduce the cost of clothing, operators across the world have run sweatshops, often with forced labor operating the machines. Many US retailers are fighting back against that practice, but it still exists.
There are also legitimate factories with fair working conditions producing clothing around the world. And there are local tailors and sewists who customize and craft clothing on demand. When I was a mere pup working at a venture-funded startup, my boss insisted I take my suit to a tailor for customization. Some years back, my wife enlisted a local sewist to make her a gown for an important event she was planning to attend.
In both of these industries, the introduction and mainstreaming of power tools did not eliminate jobs in that field. Instead, the jobs changed. There are clearly fewer artisans stitching stitch-by-stitch or woodworking saw-pull-by-saw-pull. But there are a lot more clothes and wooden objects (furniture and otherwise) in circulation than there ever were during the days of unassisted hand work.
Also: Coding with AI? My top 5 tips for vetting its output – and staying out of trouble
Programming will change. There will be fewer professional programmers who make their living coding line-by-line. But programmers will still be needed in order to code line-by-line, either to fill in edge conditions the AIs can’t do or to clean up messes AIs create.
There will also be a need for people who have never programmed, but who have systems experience, who will produce code using AI help. That’s OK. Coding isn’t a priesthood. Even though he’s not a programmer, Jason produced a tool that provides transparency to food quality data in Florida. That tool can assist every Florida resident. Should the data have remained locked up just because no experienced programmers wanted to find the answers Jason did? Of course not.
People will still craft software, but with different levels of quality and skill level. I am barely a mediocre woodworker, but I have still been able to build some of the big projects I wanted to complete.
If I want a new bookcase, the odds are I will order something from Ikea. My woodworking will consist merely of assembling the boards, roughly the equivalent of installing and configuring a piece of software.
Also: 9 programming tasks you shouldn’t hand off to AI – and why
As I see it, programming will break out into the following main trends.
- Fewer entry-level programmers will go directly into coding. Instead, they will go into testing, because all that AI work needs more testing than ever before. Learning the ins and outs of AI coding mistakes will give them powerful skills as they move up in their careers and take on bigger projects.
- People who have never programmed before, but who are willing to do the work of guiding an AI into building a program, will begin to do so. The barrier of entry here isn’t going to be experience as much as it will be persistence. Vibe coding is annoying, troublesome, time-consuming, and doesn’t always work.
- Skilled programmers will do some line-by-line coding and some vibe coding. They may get large chunks of their projects done with AI assistance to save time, but go in and tweak elements where they want to make something work in a certain way.
- Companies will hire all three categories, with the higher-paying jobs going to those who have a proven track record of creating solutions using the methods most chosen by the companies, or who are able to fix problems created by those methods.
- Some coders will still code entirely line-by-line, for a variety of reasons. It might be because that’s what they’ve always done. It might be because they’re maintaining an older codebase. It might be because they or their programmers don’t trust the AIs (privacy and/or competence). Or it might be because line-by-line programming can be more fun than arguing with an AI.
These trends are on the macro scale. What I mean by that is that none of this analysis can predict exactly how your career path (or mine, for that matter) will go. But it does show that overall, programming and programming-related jobs are here to stay. They’ll just change, as did other crafts and skill areas as they evolved from mostly hand crafting to power-augmented.
What do you think about the comparison between AI coding tools and traditional power tools? Have you tried vibe coding or agentic AI systems, and did they make you more productive or more frustrated? Do you see AI changing your role as a programmer, tester, or builder of software? And do you think AI will ultimately expand who gets to create software, or narrow opportunities for professional developers? Share your thoughts in the comments below.
You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.