Roman soldiers were sometimes paid in salt, which is where the word “salary” comes from. Salt was so scarce that it worked as currency. Then we learned to extract it efficiently, prices collapsed, and salt merchants didn’t disappear. Abundance opened uses no one had imagined: food preservation at scale, chemistry, de-icing roads.
Code is going through the same transition. Writing clean, working code used to be a rare skill that gave you leverage. Today coding agents produce more code per minute than any human could, and most developers use AI assistants regularly. Like electricity, nobody cares where it comes from anymore. Only whether it works, ships fast and costs little. Chris Messina, who watches launches on Product Hunt every day, recently saw more than 500 products launch in a single day, roughly double the usual volume. Most were forgettable. That’s what abundance looks like.
So the real question isn’t “will developers be replaced?” but “what becomes valuable when code is cheap?” Not lines of code, commits or velocity. The developer’s role is shifting:
- From reviewing output to verifying outcomes. Instead of reading every line in a code review, the focus moves to the intent of a change and to building the tools, starting with tests, that prove it delivers the expected result.
- Upstream, to requirements and specs. More time goes into the early phase, writing the Product Requirements Document and the tech spec, because that’s the context everything else is built from.
- From the change to the system. Every change has to fit the overall architecture, so the real work is assessing how it affects the system as a whole.
- Toward quality assurance. Developers increasingly act as QA engineers for each change, evaluating its impact and risks before it ships.
There’s a catch, though. Cheaper code doesn’t mean less work. It means more code: more services, more tools, more integrations, all still to be run, secured and maintained by the same number of people. And a growing share of it was accepted because the tests passed, not because someone read it. Codebases are quietly turning into black boxes.
This is why knowing how to code still matters, perhaps more than before. Not for nostalgia, but for safety. If you can’t read code, secure code and code that merely runs look exactly the same to you. Having salt in your pocket doesn’t make you a chef. The safety net has to be structural: solid test coverage, clear review standards, and pipelines that actually block bad changes.
For engineering leaders, the takeaway is clear. Stop measuring value by output. Invest in the skills that remain scarce: context, taste, architectural thinking and the discipline to review what we ship. Code is a commodity. Judgment isn’t.
References