Watching how the work of engineers around me is changing, I find the picture quite interesting.
An engineer’s value depends less and less on how fast they write code, and more on how they frame the task and verify the result. An agent with good context closes routine tasks in minutes. An agent without context produces plausible garbage, and only a person who understands the domain can tell it apart from decent code.
Second thing: skills. Expertise is slowly moving out of individual heads and getting packaged into reusable instructions: how we do integrations here, how we write ADRs, how we ship a release. A good skill is a senior colleague’s experience written down so that both a junior and an agent can use it. Teams that do this systematically speed up noticeably.
What follows for anyone growing in the profession:
- Learn to read other people’s code faster than you write your own. The job is increasingly about review.
- Learn to write specifications. This is the new main interface between a human and a machine.
- Understand the domain. Code gets cheaper, domain understanding gets more expensive.
- Keep the whole system in your head. An agent sees its slice, the engineer has to see the landscape.
Pessimists say the profession is dying. I think it is finally getting rid of the routine, and that is good news for people who like to think.