Last month, a startup in San Francisco's SoMa district shipped a complete SaaS product — authentication, payments, dashboard, API — in four days. The founding team? Two people and an AI coding agent. Three years ago, the same project would have taken a team of six engineers two months.
This isn't an outlier anymore. It's Tuesday in Silicon Valley. And it's forcing the engineering profession to confront an uncomfortable question: if AI can write the code, what exactly do we need software engineers for?
The productivity explosion
AI coding assistants have moved far beyond autocomplete. Modern agentic tools can take a natural language description, break it into tasks, write the code, run the tests, debug failures, and iterate — all with minimal human intervention. Developer surveys show that engineers using these tools report 3-5x productivity gains on greenfield projects.
The impact on hiring has been immediate. Companies that previously needed 10 engineers for a project now need 3-4. The engineers they do hire are expected to operate at a higher level of abstraction — designing systems, making architectural decisions, and reviewing AI-generated code rather than writing it line by line.
CS enrollment drops for the first time in decades
Perhaps the most telling signal: undergraduate computer science enrollment dropped 8% this academic year, the first decline since the dot-com bust. Students are reading the same headlines as everyone else and questioning whether a CS degree still guarantees the career trajectory it once did.
My students are asking me if they should switch majors. I tell them the same thing I told myself when Stack Overflow launched — the tools change, the thinking doesn't. But I understand their anxiety.
What remains uniquely human
Despite the disruption, experienced engineers are quick to point out what AI still can't do well: understand business context, navigate ambiguous requirements, make tradeoff decisions with incomplete information, and maintain complex systems over years of evolution.
- System design and architecture decisions
- Understanding and translating business requirements
- Debugging complex, multi-system production issues
- Security review and threat modeling
- Technical leadership and mentoring
- Making tradeoffs under uncertainty
The role of the software engineer is evolving from 'person who writes code' to 'person who solves problems using code (and increasingly, AI).' It's a meaningful distinction — and it means the bar for what makes a great engineer has shifted from typing speed and syntax knowledge to judgment, communication, and systems thinking.
The engineers who will thrive aren't the ones racing against AI. They're the ones learning to leverage it — treating it as a force multiplier rather than a threat. The code is the easy part now. The hard part was always everything else.
