← BackSomething's Changing

Small teams are shipping faster than ever, and bigger, sometimes bloated teams are laying off / have laid off. Is it just "the economy"? "AI is taking software jobs?" Or is something else happening?

To some extent, of course AI and macro forces will have some effect on the job market. After all, ChatGPT, Copilot, and even AI-first IDEs can generate simple code much faster, and they are getting better. However, we all know LLMs are prone to "hallucinations" and just general confusion, which are unacceptable for shipping production-grade code. AI is prone to writing code with subtle bugs - code that LOOKS right, but with minor issues. This makes it hard for even experienced developers to catch on first-glance and these issues can slow down development and even get into production if engineers aren't careful. Further, unemployment is really low (as of March 2025).

So if the overall labor market is strong, and AI isn't ready to take our jobs - why don't we need as many engineers?

Team Size and Tooling

Disclaimer: Java (replace with your favorite OOP-heavy language) isn't a bad language. It's still a great tool, and even the best tool in many cases, especially for large-scale systems that benefit from a rich, decades-old, battle-tested ecosytem.

2005-2015ish - Boilerplate Era

This was the heyday of Java and enterprise-y codebases. Layers like DAOs, Service classes, Controllers, DTOs, etc. were necessary for the tools of the time, and crucial for scalability. The cloud and the modern DevOps tools we now take for granted were either nonexistent of in their infancy. Deploying was an absolute nightmare. Roles in a team were always clear-cut: backend, frontend, DevOps, QA. To avoid spaghetti code, and to follow the rules of OOP, layers had strict boundaries. Everything had to be swappable and the system should still function if one component was replaced.

One team of engineers would often focus on one layer, and use interfaces and other boilerplate to create "contracts" between layers. This way, you only have to know your team's part of the stack. It took 10 engineers to ship.

2015-2020