The three things I actually optimize for
There is one question sitting under everything I believe about engineering right now: what does a human engineer bring when the agent can write the code?
Because it can. Agents today write code that is right and does what they set out to do. That is no longer the rare part. So if the code is handled, what an engineer is for has to be something more than writing it. These are the three things I think that is, and they are the three things I optimize for.
First-principle framing
Give me the problem, not the instruction.
When a spec arrives stripped of its reasoning, I act like an AI tool. I take the instruction, I produce the output, and I am exactly as replaceable as the agent that could have done the same thing faster. The moment you hand me only the technical specification, you have asked me to be the cheapest version of myself.
The reasoning is what changes that. When I am given the business logic that derives the spec, the task stops being a task. It becomes a frame I get to lay out: what is right for the org, what is wrong, why things are the way they are right now. Do not tell me to centre a button. Tell me it is the primary CTA and users struggle to find it. Do not tell me to build feature X. Tell me we are losing creators during onboarding. Hand me the problem and let me arrive at the solution on my own terms. One framing makes me a tool. The other makes me an engineer. The difference is entirely the why.
Cultivated intuition
This is the part the agent cannot inherit from you, it has to be earned.
Intuition is not a shortcut. It is the compound interest of every problem you have shipped before. Call it pattern finding or learned practice if you like, but in day to day engineering it comes down to the instinct you build from past experience. I know I have reasoned enough when I have a real read on feasibility and infrastructure. After that I would rather ship a scrappy version and learn whether the thing is even possible than keep reasoning in the dark.
And it compounds. The sharper my intuition becomes, the sharper my assumptions become. Every problem I move through is a new one, so moving through it sharpens the next read. The interesting part is that AI has made this loop faster. The SDLC cycle is quicker and more efficient now, which means more reps, which means the intuition sharpens faster than it used to. My intuition has been wrong plenty, but wrong is just the next rep. Creativity lives in the standard deviation from the already known solution, not from the goal. You earn the right to deviate by having reasoned hard about the why first.
Realistic optimism
The last thing the agent does not have is a sense of what the work is for.
Engineers get lost in the craft. The lowest level, the most elegant abstraction, the deepest rabbit hole. Going deep is genuinely amazing for learning and it is entertaining, I am not against it. But the end goal of an engineer is to extract value for the real world, and I try not to let technical ego blind me to whether the effort actually translates to that. I would rather ship something people actually use than something clever that no one needs.
The same realism shows up in how I work on a team. Realism is just seeing constraints clearly, whether they are technical constraints in a system or organizational constraints in a team. I have spent my time in fast paced startup environments, and there calling out early is a sign of ownership, not avoidance. When something is out of reach, I make the call early and decide tight, rather than going quiet and dropping a no at the deadline when both paths were still open. The clearer you are and the earlier you take initiative, the better the whole outcome looks. Be distinctively sorted.
Optimism gives you the direction. Realism keeps you honest.
None of this is a moat
Engineering will change shape, maybe past the point of looking like what it is now. I am not mourning that. These values were never the engineer's to begin with, they belong to good thinking, and that carries into whatever shape the work takes next, across industries, though I will always be an engineer at heart.
"You can outsource your thinking, but you cannot outsource your understanding." — Andrej Karpathy