Coding was always the easiest part of the job, and AI has taken over much of it now.
The tough part still stays with the human mind:
- Designing the system end-to-end (tradeoffs, constraints, scale)
- Breaking vague requirements into clear specs
- Choosing the right abstractions, data models, and boundaries
- Thinking through edge cases before users find them
- Making it reliable: failures, retries, idempotency, fallbacks
- Security privacy by default
- Observability: logs, metrics, tracing, alerts, SLOs
- Performance cost: what matters at P95/P99, whatβs waste
- End: βit worksβ isnβt the finish line
Engineers build systems. You may have an argument, if you write a detailed prompt covering all of it, AI will be able to do it. As I said, coding is the easier part of the job once you know what you up-to.