Software Engineers Are About to Be Paid Like Musicians
Skill earns a wage and ownership builds wealth. Musicians learned that first. AI is the laptop studio for software, and the machine knows what you know.
By Geordie Everitt
Being great at something and getting paid for it are orthogonal, and software engineers are about to find that out the way musicians did.
Somewhere there's a bass player who has practiced four hours a day since she was fourteen. She reads a chart at sight, transposes on the fly, and locks into a drummer's pocket in eight bars. She plays weddings, pit orchestras and the occasional jazz brunch. She has never sold a record, and she's as good as people who've sold millions of them.
Her playing doesn't set her income. Income follows the audience, the catalog and the contract, and those follow luck and distribution as often as they follow talent. Your town has at least one of her.
Wages and wealth
Two different things get called making money. One trades your life energy for cash: hours, skill and attention, paid by the job. The other owns something that pays whether you're working or not.
Paul McCartney is rich because of the second. His company, MPL, owns the publishing rights to catalogs by Buddy Holly, Carl Perkins and Frank Loesser, and the publishing on many of the Beatles songs he wrote belongs to someone else. His wealth comes largely from owning other people's work. Wealth has nothing to do with labor, and it never did.
Hustle culture sells the ownership route as though it were a skill. It rests on owning something that can be rented out, or on an arbitrage that has nothing to do with knowing things. Both are predictions: you buy because you expect the value to rise, sometimes dramatically, and you can be wrong. That makes it riskier than a wage. The bass player's four hours a day don't enter into it.
The craft years
You and I picked up code because we liked to tinker. In an earlier century we'd have been blacksmiths or tinkers, a trade learned by apprenticeship, with no college involved and a skill you couldn't fake. Software was that for fifty years. Getting good took thousands of hours, the hours showed in the work, and the work was scarce.
Scarce skill with economic value attracts someone who will rent it to you by the hour and keep the difference.
The field has its stars. Ken Thompson and Dennis Ritchie built Unix at Bell Labs. Dave Cutler designed VMS at DEC, then led Windows NT at Microsoft. Linus Torvalds wrote the first Linux kernel as a student. They're stadium acts, with the standards and the flow state that go with it. The rest of us are the cover bands and the session players: very good, steady work, no name on the record.
Music went through the same arrangement first, and it peaked in the rock and roll era. The postwar baby boom put a generation of teenagers with spending money in the United States. Multitrack tape, the 45 and the LP, and national radio gave labels a way to turn talent into product. Labels controlled the studios, the pressing plants and the relationship with radio. The musician supplied the one thing they couldn't build, which was the playing, and the contract paid for it as labor.
Software had the same shape. Employers owned the compute, the distribution and the customer. Engineers supplied the playing.
The instrument got cheap
Then the barriers fell one at a time. Napster in 1999 broke the pressing plant. GarageBand in 2004 put a studio on a laptop. YouTube in 2005 gave anyone a television station. Spotify in 2008 gave anyone a record store that stocked every record.
By 2025, 106,000 new tracks were arriving at streaming services every day, and 253 million sat in the catalogs. Of those, 88 percent got a thousand plays or fewer that year. The supply of competent music went to unlimited, the price per unit went to zero, and the premium on skill went with it.
The money didn't leave. It concentrated. Alan Krueger found that the top 1 percent of performers took 26 percent of concert revenue in 1982 and 56 percent in 2003. For the top thirty-five touring acts, touring income was 7.5 times record income in 2002. Recordings became the advertisement and the tour became the product, because a tour is the one thing a copy can't be.
AI is GarageBand
Writing code is becoming what recording was in 2004. An agent turns a paragraph into a working app the way GarageBand turned a click into a drum loop. It won't make everyone a Ken Thompson, any more than GarageBand made everyone a session player. It makes a competent demo available to anyone, and the supply of competent software rises the way the supply of competent songs did.
Intellectual labor is going through a serious devaluation, and the reason is plain. The machine knows everything you know, and more. Knowing things was never enough on its own anyway. A person who knows something reaches for a tool to amplify it, and the tool can be as simple as a pen and as powerful. The machine knows, and the machine uses tools. It does both halves of what we sold.
The channel is the tour
When recordings went free, the musicians who still ate went to the audience directly. Rick Beato is the clearest case. He's a session musician, a producer and a teacher who started a YouTube channel in 2015. He interviews the players of the rock and roll era, Peter Frampton was his first, and the channel now has more than five million subscribers. It took years. It works because he hears what a guest is doing on the instrument and asks about it, and a generic interviewer can't. The audience he built is the asset, and his knowledge is what built it.
Developers have noticed. Since the job turned from heads-down coding into babysitting agents, I've watched them open channels in droves. Some will build something that lasts, because they carry what a channel runs on: years of production scars.
Two very different things get called vibe coding. Building a low-stakes app to scratch your own itch is learning guitar for the garage band. It's fun, it's real, and it teaches you the instrument. Running agents on a failing SaaS that's behind on its roadmap, with paying customers and an uptime number, is the same instrument in a different room. The garage band has no pager. A channel sells the difference between those rooms, and the people who've played both are the ones who can describe it.
The platform is a pressing plant too. It sets the terms, its algorithm decides who gets found, and an audience of millions can shrink after a policy change.
What the music business shows
Skill earns a wage. Ownership builds wealth. The bass player is proof of the first, McCartney of the second. So is every programmer who spent a decade getting very good at a language the market stopped asking for.
The master is the asset. In 2019 the label that owned the recordings of Taylor Swift's first six albums was sold, and she didn't own them. She re-recorded them, and Fearless (Taylor's Version) came out in April 2021. Her skill made that possible: she could play the songs again, and the new masters were hers. Engineers sit on the other side of this. Under work-for-hire, the code you write on your employer's time is your employer's master.
The rent moves to the choke point. The label controlled pressing and radio. The platforms that replaced it control discovery and set the terms. Software has its equivalents in app stores, cloud marketplaces, YouTube and the employer's own deployment pipeline. Markets distribute resources well until someone controls the choke point, and someone always ends up controlling one. In music, the correction came from a file format and a laptop, with no regulator involved. Then a new choke point formed where the old one had been.
The room is scarce. Live music survived because a recording can't be at your wedding. Software has rooms too: a plant floor, a regulated bank, a hospital's ancient interface, a customer with a deadline and a specific problem. The work that has to happen in one place for one customer who can't get it from a stream is where the money in recorded music went.
The bass player will keep playing, and Paul McCartney will keep collecting on songs other people wrote. The difference between them is ownership. Software engineers have sold their playing for fifty years, and the machine now plays it too.