Sometimes, it is easy to forget that the reason why certain skills are useful is not because of anything intrinsic about the skills but because of what they can help someone to achieve.
A lot of software development has been developed led, largely because developers have the skills. Hence the scratching an itch leading to a lot of open source software development. Maybe in the short to medium term, those skilled developers will produce software of much greater quality than LLMs - I think credible voices are beginning to suggest that this may no longer be the case.
I for one, am not what you could realistically call a software developer, but I have dabbled in code and produced products that are useful internally in my employment. Writing code is importantly, but somewhat incidental to what I do. I have found that LLMs, or more accurately, AI pair programming tools have increased what I can achieve in short periods of time. The time between idea and execution is more shortened (for me).
Could I write high performance software that would be deployed widely e.g. write a kernel, or browser engine? No. But I can now create software products that works have cost too much to create with a fraction of the effort and at a fraction of the cost. It is only a matter of time until LLMs are better than 95% of coders.
Another really good use of LLMs is to help âoelay people âoe understand what code does. Where I have people working for me, I can now ask an LLM to help explain what code does etc. if you are a maintainer, I can see AI programming tools being an immense help in understanding what code is doing etc.