I was at an event modeling conference last week, and over lunch the conversation kept drifting to the same place: how work life is going to be disrupted by AI. Everyone in that room builds software for a living, so it was not an abstract worry. It was the “what happens to us” kind.
Here is where I landed. The people who make it in this industry are not going to be the ones who think their job is typing out code once the specs are ready and handed to them. That part, the part where the requirements are already understood and you just translate them into code, is exactly the part AI is good at now. If your whole value is turning a finished spec into an implementation, that value is shrinking.
That sounds bleak if you stop there. I don’t think it is.
What the job becomes
The job doesn’t disappear. It moves.
You now have to be involved in the domain. You have to actually want to understand the problem, not just the ticket, because that understanding is what lets you steer the AI toward a solution that is correct instead of merely plausible. Before, being curious about the business was optional. Plenty of good engineers got by without it. Now it is the thing that makes you effective, because steering an agent well requires you to know what “right” looks like before the agent shows you anything.
The other half of the job is review. This came up as its own hot topic at the conference: the developer’s core skill is no longer writing code, it is reviewing code efficiently. When an agent can produce a slice in minutes, your bottleneck is no longer how fast you type. It is how fast and how confidently you can look at what came back and say yes, that is right, or no, that is not what we agreed on.
Why the method matters here
This is where event modeling and event sourcing stop being a preference and start being an advantage.
If you take the time to map out how information is supposed to flow through the system, in a format simple enough for a human to read and verify, you have given the AI a blueprint it cannot wander away from. In all my experiments with vibe coding, when I actually sit down and lay this out first, there is no room for the model to get creative and build something out of spec. The spec is the fence.
And event sourcing keeps the review side honest, because there is not much to review. It is essentially four patterns. You can get a long way with two. When every slice is built from the same small set of shapes, checking the output becomes almost mechanical. You can wrap rules and tests around the generated code to make sure it follows the patterns. You can even just eyeball it, confirming that a slice only touches the files inside that slice and nothing outside it. That is a review you can do quickly and trust.
So the two new skills, understanding the domain and reviewing efficiently, both get easier when the method is simple and the blueprint is written down first. That is not an accident. It is the whole point.
Who this is good for
If you came into this industry because you like building things and solving problems, you have nothing to worry about. The job is changing in your favor. You get to work with bigger strokes now, to have more impact per hour than you used to, because the tedious middle of the work is being handled and you get to spend your attention on the parts that actually need a human.
If you came into it thinking the job was implementing against complete, predefined specs and nothing more, that is the part being automated. And honestly, I think that is fine too. Maybe even good. It frees people to go find the thing they are actually meant to be doing.
I really do believe everyone has something that drives them. Sometimes we have just been so busy making a living that we never got the chance to chase it. My hope is that AI does not only push people out of the jobs they were quietly hiding in, but also becomes the thing that lets them build whatever it is they are actually passionate about. The key is making that kind of AI accessible enough that a person with an idea and no formal training can still get somewhere with it.
The job is changing. It was always going to. The version of it that is arriving asks you to understand more, care more, and review more, and it hands the boring part to a machine. I will take that trade.
being in tech 