A Quiet Revolution in Development
The software development workflow is shifting from writing code to supervising it, as AI-generated code becomes more common.
“It’s the difference between being the site foreman of a skyscraper and laying every brick by hand; the job still requires judgment and oversight, but not manual effort at every step.”
Verification as the New Frontier
As more code generation is delegated to AI, verification becomes the new center of gravity.
Unit tests serve as the first line, confirming that outputs behave as expected. But checking outputs alone isn’t enough:
- What if the AI introduces vulnerabilities, such as unauthorized file access or unexpected network calls?
- What if the overall design is flawed—unclear, inconsistent, or difficult to maintain?
- What if the code passes all tests but violates architectural principles or causes performance issues?
Addressing these challenges requires new tools and processes, representing an evolution of current static analysis. Interestingly, other AI systems can help perform these verifications, creating a feedback loop between builders and inspectors.
Supervision Defines Tomorrow’s Engineering
Is this the direction in which software development is heading? What skills will separate the most effective developers from the rest? And how can we cultivate those skills?
Perhaps the future of development lies not in writing more code, but in learning to supervise, verify, and guide what machines create.