Cognitive Dependence

Very quick post, forgive the brevity. We know that using AI as a cognitive scaffold can result in skill atrophy. The science is still a little indeterminate, but there's enough evidence to show that phenomena like cognitive surrender and dependence can set in with extended offloading (delegating intellectual labour to AI). So my question is the exent to which this is happenign to SWEs and how this will affect model development. If we accept the science and we adopt a long singularity time horizon, then relying on models for development could plausibly lead to a scenario where engineers simply don't have the skills - particularly the younger generation. The result would be a kind of plateau, where models aren't advanced enough for RSI (recursive self improvement) and engineers are unable to move the needle appreciably. There are huge holes in the argumentation - I see them. Blame it on the brevity, but I do think that there's a kernel of a point here. The vibecoding stories I'm hearing from SWEs are astonishing, some report that they haven't written a line of code for months. Is this sustainable? I also think that this points to a broader concern about how reliant society may become on AI and how that will affect knowledge production. There's a dystopian scenario where timelines are extended and humanity settles into a kind of stagnation, reliant on AI machines that have yet to fully individuate and unable to bootstrap to the singularity.