On AI Replacing Software Engineers
There’s been a lot of talk lately about a future where AI could do every job. Even software engineers feel worried after seeing Devin, an AI SWE, that can build apps and fix errors on its own.
But will these AI engineers ever reach the level where they can handle and drive complex projects end-to-end without human supervision?
What I believe is that the AI agents will evolve as assistants to human software engineers. They will take care of boring stuff like basic coding and fixing simple bigs, freeing humans to focus on planning projects, coming up with new ideas, mentoring others, cross-functional collaboration, and going deep, domain-specific tasks. After all, software engineering isn’t just about coding.
It also possible, that as AI gets better at coding and fixing bugs, there will be less demand for entry-level software engineers. But senior engineers will still be needed for complex work I mentioned above.
And, of course, the software engineering jobs will change too — they will require skill of integrating AI features to software systems.
Here is what I suggest to anyone who is thinking about how to adopt to the changes:
- Invest in “human-only” skills such as clear and effective comminication, product vision (what to build instead of how), working with incomplete or conflicting requirements and “unknown unknowns”.
- Get comfortable with AI IDEs. You’ll see how fast you can start a project and understand that coding is only part of being a software engineer, and it loses its weight.
- Master prompt engineering. It will help ship code faster, draft design docs faster, and etc.
In my view, a jobless future for software engineers is still far away. To get there, someone would need to create an AI that doesn’t hallucinate, can do everything a human can, and ensure that regulators don’t impose rules that prevent AI from being used everywhere.