Replying to
@ansuz@gts.cryptography.dog @david_chisnall@infosec.exchange While I agree, that companies hire for this, I am not convinced they should be looking for „problem recognition“ alone. I remember a few times, where clever engineers thought they recognized the problem to solve, who then were stuck in the „solution space“, and could not describe how their idea connects with the problem of a client.
AI amplifies this. You can get your description of a problem made into a program faster at a fraction of the cost. But does it solve your problem?
I have not yet seen an AI questioning the prompt (the problem to solve) and only few engineers, who are confident and brave enough to do that. I have neither seen a hiring manager or client look for this.
IMO having an automated library of problem-solution-pairs is great. It makes life easier, and custom software way more accessible. Yet one still needs to understand the problem to be able to solve it reliably.