Will AI Make Software Engineers Obsolete?
Five questions about AI and the profession
The role, vibe coding, writing code, why code is still taught, and how uncertain it all is. Each with a one-line take-away.
Will AI make software engineers obsolete?
Probably not, but the role may change substantially.
- Custom software is expensive today, so many organizations settle for similar off-the-shelf systems.
- If AI makes specialized software cheap to build, many will prefer systems designed for their exact needs, raising overall demand for software.
- Engineers then spend less time hand-implementing and more on operating AI tools, defining requirements, making architecture decisions, validating results, and shipping custom systems fast.
Take-away: the job does not vanish, it moves up the stack, from writing code to directing and validating it.
Can everyone build software through vibe coding?
Sometimes.
- Non-programmers can now build simple apps and prototypes, and developers can reach roughly 80% of the functionality very quickly.
- The last 20% is much harder: a reliable, secure, maintainable, complex system still needs broad knowledge of engineering principles, architectures, tools, testing, deployment, security, and the application domain.
Take-away: AI lowers the barrier to creating software, not the difficulty of engineering it well.
Will software engineers still need to write code?
Probably much less than today.
- One likely model: each engineer becomes a mix of senior engineer, architect, product designer, and technical supervisor, while AI does much of the work once assigned to junior developers.
- You still have to understand AI-generated code fast, because you direct the AI, catch mistakes, weigh design choices, and ensure the system does what was intended.
- Open question: can students learn to understand complex code without first building complex systems by hand? There is no clear answer yet.
Take-away: you will write less code, but you are more responsible for understanding it.
Why do I still teach using code?
As a precise formalism for explaining systems and processes.
- Like a physics formula for a shell fired from a cannon: it makes the underlying forces and relationships exact, yet nobody derives and solves it by hand every time.
- Code likewise makes algorithms, architectures, data flows, assumptions, and system behavior clear, even when AI eventually performs most of the implementation.
Take-away: code is taught to build understanding, even when AI writes most of it.
How uncertain is the future?
Very. It may look nothing like today.
- AI might stop generating code in languages meant for both humans and machines, and instead use internal representations optimized for machine understanding, not readable by humans or bound to today’s paradigms.
- Or AI stops producing traditional software at all: the system itself becomes the software, a general system like ChatGPT interpreting requests in natural language, images, speech, or examples and doing the task with no conventional app in between.
Take-away: nobody really knows how this plays out, so stay current, track the news and trends, and place your own bet on the most probable future.
