Focusing on the mechanics of execution rather than the output of the model is the only way to move past the current hype cycle. When teams prioritize the process of validation over the initial generation, they build a defensive posture that actually holds up under pressure. This shift from simple automation to structured verification mirrors the trap of choosing ignorance for the sake of compliance. Relying on checkboxes to satisfy audits creates a false sense of security that crumbles the moment a real threat appears. True operational maturity comes from choosing to see the gaps that others prefer to ignore.
Good challenge, Antonio, and it's the question I get most often on this topic.
I'd push back slightly on the premise though. The junior work that's actually at risk is the repetitive execution: cleanup, meshing to spec, running the load matrix, formatting the report. That was never where juniors learned judgment anyway, it was just the toll they paid to get access to the model. If automation removes the toll but not the access, a junior can get to the judgment-building work (why did we idealize it this way, why doesn't this result look right) much faster than they used to.
The real risk isn't "no juniors." It's a generation of engineers who never did the repetitive work and so never built the pattern recognition that comes from doing it badly a few hundred times before doing it well. That's a training design problem, not an automation problem, and it's on us (the people building courses and onboarding programs) to solve it deliberately rather than assume it'll happen by osmosis like it did for our generation.
I get you and I like the point you say about the "training design problem", I agree that we are talking about a generation that will not learn anymore through a long process of trial and error.
It is just difficult to imagine for me this reality, since I find it hard to memorize a concept or a solution to a problem without the "learn by doing" or "testing" approach.
Focusing on the mechanics of execution rather than the output of the model is the only way to move past the current hype cycle. When teams prioritize the process of validation over the initial generation, they build a defensive posture that actually holds up under pressure. This shift from simple automation to structured verification mirrors the trap of choosing ignorance for the sake of compliance. Relying on checkboxes to satisfy audits creates a false sense of security that crumbles the moment a real threat appears. True operational maturity comes from choosing to see the gaps that others prefer to ignore.
https://cyrilsimonnet.substack.com/p/the-cheapest-way-to-comply-is-to?utm_source=substor&utm_medium=substack&utm_campaign=comment
That’s an interesting topic to talk about.
I have frequently read/heard opinions stating that in companies the positions more at risk to be substituted by AI are those of Juniors.
The reason is that usually they are the ones assigned of repetitive or systematic tasks that do not require much judgment.
But can we really imagine a future without juniors?
Good challenge, Antonio, and it's the question I get most often on this topic.
I'd push back slightly on the premise though. The junior work that's actually at risk is the repetitive execution: cleanup, meshing to spec, running the load matrix, formatting the report. That was never where juniors learned judgment anyway, it was just the toll they paid to get access to the model. If automation removes the toll but not the access, a junior can get to the judgment-building work (why did we idealize it this way, why doesn't this result look right) much faster than they used to.
The real risk isn't "no juniors." It's a generation of engineers who never did the repetitive work and so never built the pattern recognition that comes from doing it badly a few hundred times before doing it well. That's a training design problem, not an automation problem, and it's on us (the people building courses and onboarding programs) to solve it deliberately rather than assume it'll happen by osmosis like it did for our generation.
I get you and I like the point you say about the "training design problem", I agree that we are talking about a generation that will not learn anymore through a long process of trial and error.
It is just difficult to imagine for me this reality, since I find it hard to memorize a concept or a solution to a problem without the "learn by doing" or "testing" approach.