Code Was Always Commodity. AI Just Made It Obvious.
Chris Messina wrote that code became commodity. Microsoft's AI CEO told Fortune he's giving white-collar jobs 18 months. The apocalypse merchants are back, right on schedule.
Here's the thing: we've heard this exact song before. Delphi. Visual Basic. Dreamweaver. WordPress. Low-code platforms. No-code builders. Every single one promised the same outcome: one developer doing the work of an entire team.
Twenty years later, we're still shipping code. Just with better tooling.
The Salt Analogy Actually Works
Salt used to start wars and build empires. Now it's $2 at Woolies. Classic commodity story.
But here's what the doomers miss: cheap salt didn't commoditise cooking. Knowing what to do with ingredients, that's the skill that matters.
Code was always basic. First-year CS students write algorithms. I coded at thirteen, my younger brother at eight. Engineers have been writing disposable software in MATLAB for decades. People built entire systems in MS Access and Delphi in the '90s.
What changed isn't that code became a commodity. Code was always commodity. What changed is software got absurdly complex, mostly unnecessarily so.
AI reduced the cost of syntactic code production (boilerplate, CRUD, basic patterns). The cost of architectural decisions? Same as before. Possibly higher.
The Complexity Merchants
I've never written Ruby, but DHH's RubyConf keynote introduced me to a perfect term: "merchants of complexity".
We did this to ourselves. Companies think they need separate frontend and backend teams because everyone else does. Suddenly you need an architect, two backend devs, a frontend dev, DevOps, two QA cycles, and a PM to ship a login screen in three months for 47 concurrent users.
Will LLMs change this? Absolutely. Is it the end? Not even close.
The same companies that sold unnecessary complexity are now selling the solution. While Suleyman predicts full automation in 18 months, the same Fortune article cites a METR study showing AI is making devs 20% slower. Classic executive product pitch. Borland did it with Delphi. Microsoft with VB. Now it's Microsoft again with Copilot.
What Isn't Commodity
If code was always commodity, what actually matters?
Knowing what to build, why to build it, and how to keep it alive in production long-term.
AI can generate perfect code for the wrong problem. It'll happily create an implementation that should've been deleted, or build a feature that'll bottleneck in production at scale.
Even senior devs using AI need judgment. The LLM doesn't know your business rules, your SLAs, or why that "simple refactor" breaks three other services. It's never been code vs no-code. It's being a developer vs being a typist.
A New Cycle Begins
Generalist devs are ascending. A fullstack dev with Claude or Cursor can prototype in hours what used to take a squad weeks.
Backend devs who don't understand the apps their APIs serve? Frontend devs who don't grasp basic backend concepts? They're losing ground. Fast.
I see this as a new cycle starting, where writing code isn't enough. But for me and many developers in the market, it never was. Our profession was always more intellectual than manual. That, friends, isn't commodity.
I use LLMs daily. I see AI as powerful tooling for developers who take the craft seriously. I also see it as a bomb in the hands of people who don't understand software fundamentals. When you hit the limit of what AI can do, the first group knows what to do next. The second group opens a support ticket on Lovable.
Same Principles, Better Tools
What's happening isn't a collapse in the value of code or the developer profession. It's a new cycle following the same patterns we've lived before. This tool is more powerful than previous ones, but so was every other tool in its era.
If you survived Delphi, jQuery, Angular 1 vs 2 (or React), blockchain hype, you'll survive this too.
The skill isn't typing code. It's knowing what code to type, when not to type it, and how to keep it running when everyone's asleep.
AI just made that distinction impossible to ignore.