Why Programmers Still Matter Even When AI Is Powerful

A programmer asked whether AI’s rapid code generation means developers will soon be obsolete. The author replies that AI is merely a "problem‑solver," while programmers are "problem‑creators" who add essential context.

Story 1 – Reading the Air : After a system upgrade, payment success rate fell by 0.3 %. AI suggested checking the database connection pool, which was correct, but the real cause was a temporary finance field that added an extra join without an index—information only known from a requirements meeting that AI never attended.

Story 2 – Taking the Blame : Before a major sales event, AI analyzed logs and proposed a complete gateway rewrite with a new reactive framework. The team rejected the merge because the deadline was too close; instead they added two machines to isolate non‑core traffic—a "ugly but certain" fix that ensured responsibility, something AI cannot guarantee.

Story 3 – Archaeology : An old module contained a comment:

// TODO: 这里先hardcode，等XX接口稳定后改成动态配置

. The comment reflected a deliberate compromise to maintain user experience when an external API was unstable. AI would automatically refactor the code to a modern micro‑service call, discarding the historical rationale.

In conclusion, a programmer’s delivery consists of three guarantees: accurate interpretation of requirements, awareness of system constraints, and willingness to own any failures. AI currently provides none of these. While AI lowers the cost of routine "brick‑moving" tasks, developers who only perform such tasks lose value, whereas those who can judge, balance risk, and ensure certainty become even more valuable.