The End of Coding Is the Wrong Question to Ask

The End of Coding? Wrong Question

TLDR: Oskar Dudycz challenges the tech industry's shallow "end of coding" debate, arguing that generating code with LLMs is just a PoC-level trick, not real engineering. He draws historical parallels to Java's introduction and Joel Spolsky's JavaSchools critique to show that every new abstraction layer triggers the same panic, and what we actually need is a mature conversation about restructuring the entire SDLC.

What is missing from this argument: Dudycz does not engage with the possibility that LLMs might improve faster than his "transition phase" framing suggests. He also sidesteps the economic pressure angle -- even if chat-based development is suboptimal from an engineering perspective, if it is cheap enough and fast enough, businesses may not care about the quality gap. The essay also does not address the middle ground seriously: hybrid workflows where LLMs handle boilerplate and well-understood patterns while humans focus on novel architecture and business logic. This is arguably already happening at scale, and it is neither the "end of coding" nor business as usual. Finally, the Roman Empire analogy, while evocative, conflates two very different knowledge preservation contexts -- we have version control, documentation tooling, and searchable codebases that the Romans did not have.

The End of Coding? Wrong Question

Referenced:

• The Java Language Environment - Sun Microsystems (1995)
• The Perils of JavaSchools - Joel Spolsky