When factories automated their assembly lines, the repetitive tasks went to machines, but humans stayed to oversee, fix problems, and make decisions. That's exactly what's happening to software development right now.
Assembly Line Coding
In modern factories, humans don't manually assemble products anymore. They operate machines, monitor quality, and intervene when something goes wrong. As a developer today, that's increasingly what the job looks like.
You can build automated pipelines that chain together LLMs and tools. Feed in specifications at one end, get working code at the other. For routine tasks like adding features, updating UIs, or creating API endpoints, you're operating the machinery that codes.
Two Kinds of Factory Workers
In factories, there are two types of workers. Line operators handle routine production. They know their machines, set them up for each job, monitor output. Then there are the specialists who get called in when something complex needs doing: retooling for a new product, diagnosing strange defects, optimizing the production flow.
Software development now has the same split. For getting-things-done tasks, you're a line operator. Set up the AI pipeline, let it run, check the output. For complex problems like architecture changes, performance issues, or tricky bugs, you're the specialist who steps in to work directly with the tools.
Who Will Survive?
If your job is taking clear specifications and turning them into code, AI does that now. This hits junior developers hard, but there are also plenty of developers with 10+ years experience whose main skill is implementing specs cleanly. These developers are replaceable by AI tools. Years of doing assembly-line work doesn't prepare you for the jobs that remain.
The developers who survive will be those who can do what AI can't do alone. They can look at a vague business problem and figure out what should actually be built. They can spot when the AI's solution is technically correct but architecturally wrong. They can make judgment calls about tradeoffs the AI doesn't even know exist.
Future demand will be for developers who can:
- Translate messy business needs into technical solutions
- Make architectural decisions based on incomplete information
- Recognize when AI output looks right but isn't
- Debug problems that span multiple systems
- Optimize for constraints the AI doesn't understand
What I'm building
Delegate tasks. Get software.
Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.
Take a look at vroni.comI respect your privacy. Unsubscribe at any time.