Stop Targeting Junior Devs

The recurring theme of the disappearing developer … Now it’s the undergraduate students and junior devs who are getting targeted. From low-code, no-code, ADF to outsourcing, from time to time crops up some “innovative” way to replace lots if not almost all of those pesky devs that take just too much time to churn out the stuff we want from them. Didn’t work out so well each and every time. It won’t work this time either.

It has been repeated over and over again that the code that genAI tools produce matches the quality of code that a junior dev produces. Some even go as far as to propose to work with genAI coding tools like you’d work together with a junior dev. Both statements are an unjustified and stupid insult to junior devs in general.

I’ve seen quite some code of junior devs that blows LLM generated code out of the water. Junior devs do not produce utter nonsense with references to stuff that doesn’t exist like genAI tools do frequently (I have written here about why it’s stupid to call this ‘hallucinations’). They need guidance now and then but that’s neglectable compared to the babysitting genAI tools require.

genAI Tools lack some abilities that every dev –no matter whether junior or senior– has, and which are crucial for software-engineering. Here are a few examples:

• knowledge of low-level rules – genAI tools lack explicit knowledge of such rules. The solutions provided by such tools are based on –a rather sophisticated– guesswork.
• reasoning – the ability to apply low-level rules, decide how to design a system, such that it complies with those rules, and explain your decisions. Good luck with cleaning up technical debt in a complex system or migrating it when there is no rationale behind the code.
• creativity – genAI tools infer solutions from training data. Those tools can provide low-quality solutions for simple use cases, based on patterns, but are pretty bad at tailoring those patterns into robust, maintainable solutions for the specific use case.
• logical thinking – A human dev can understand constraints, spot conflicting constraints, and can either resolve or report them. With conflicting constraints, genAI tools produce “solutions” where constraints are either quietly dropped or only a subset of the required functionality is implemented. genAI tools are not capable of resolving constraints.

Whoever comes up with the argument that LLM tools produce code much faster: yes, they do produce shitty code much faster. Producing lots of shitty code in short time has always been the main contributing factor for projects becoming a train wreck that everybody hates working on and that ultimately fail in the long run.

Producing code is not the bottleneck in software-development. Getting the requirements right is. Designing maintainable code is. Weighing and resolving constraints is. Composing complex systems and maintaining an overview of them is. If someone hired a junior dev at the beginning of this hype, she or he would have developed considerable skills in doing all of that by now. genAI tools haven’t. I have the impression that folks that promote genAI tools for building production grade systems view software developers as code monkeys. This is a widespread sickness in our industry and I guess it’s one of the reasons for the proliferation of the current hype.

It’s just plain stupid trying to replace junior devs with genAI tools and expect senior devs fixing the crap that those tools produce.

My cheers go out to all the companies that ain’t copycats. You recognize them by their job listings.

This is an outtake of my post on http://codebase.zone/blog/2025/10/why-i-dont-use-ai-for-coding-or-writing/#the-recurring-theme-of-the-disappearing-developer, with minor adaptions/additions.