Strangely enough, I work with AI models as a programmer and do not use any AI tools to code. The creation of "cognitive debt" as AI researchers describe it means that crucial components of my code are offloaded into a black box understanding.

"Across groups, NERs, n-gram patterns, and topic ontology showed within-group homogeneity. EEG revealed significant differences in brain connectivity: Brain-only participants exhibited the strongest, most distributed networks; Search Engine users showed moderate engagement; and LLM users displayed the weakest connectivity."

In other words, working with LLMs reduces the level of engagement in our brains and offloads much of our cognitive activity.

When working with models that require a high level of abstraction and optimization, offloading these tasks to LLMs that essentially scrape and combine code from various sources can lead to disaster. For something like a basic CRUD app such factors are not nearly as important.

I can see the future convergence of two clear trends in software: No-code tools and AI tools. The development of AI agents which handle not only the generation of code but the full development cycle of creating, testing, and deploying projects with a limited level of programming knowledge required.

As an additional note:

Both of these software trends aim to automate and replace many responsibilities of the programmer in a commercial sense. I believe that future programmers will need to become familiar with these emerging technologies in order to remain competitive in the job market.

Namely, math skills. Linear regression, multivariable calculus, the sort of math applied by scientists and researchers. If you do not understand the mathematics at the core of neural networks, now is the time to learn.