The impact of LLM's insoftware development

Yesterday I wrote an article about the changes AI is bringing to software development.

In this article, I made a prediction of how AI would impact software development and software engineers. In my opinion, at least in the short-term, developers won't be completely replaced as AI will have a bigger impact on things that are typically done by junior professionals.

Coincidentally, I stumbled on an articleby Sergey Tselovalnikov today that corroborates my prediction and expands on the theme, adding some fascinating insight on the subject.

LLM Impact on Software Engineers

The core point of the author's theory is that LLM's impact on people's jobs will largely depend on their proficiency level. That notion can be illustrated by the following curve:

Source

Let's take a look at each level and the potential impacts that LLM's may have.

Junior engineer

Understandably, this is where LLM impact will have the highest impact. Junior engineers are typically in charge of working on simple features, upgrading libraries and peer-reviewing very simple pieces of code.

LLMs can be a huge help to Junior Engineers, but they must be careful because if they rely too much on LLMs to do their job for them, it may not be long before managers realize junior devs can be easily replaced by AI-powered assistants.

Junior engineers should strive to understand what they are doing and show they can do more than AI alone as soon as possible or their role is at serious risk.

Mid-Level engineer

Mid-Level engineers are more skilled in coding and have a much better understanding of algorithms and logic. They can use LLM's to speed up their coding by providing assistance with stuff like syntax which really shouldn't be a big deal.

But, again, even a mid-level engineer cannot "outcode" an LLM, so they must make sure they can do things that LLMs can't yet, such as translating customers' concerns and desires into actual understandable requests that can be coded into software.

I think AI agents will even be able to do that, but that may take a little longer.

Senior engineer

Senior engineers are usually in a very different position compared to their more junior counterparts because, as ironically as it may sound, they tend to do less and less coding as they progress "up the ladder".

They have a deep understanding of the codebase, the project as a whole, and the customers, both internal and external. Because of that, senior engineers have an almost strategic role where they are in charge of helping to make sure the code is aligned to the business goals of the project.

Again, it's possible that AI will be able to fill that role one day but I think it will take a while so senior engineers are much safer.

Staff+ Engineer

This is an interesting point which I had not considered in my previous article. Sergey makes a very valid point that Staff+ roles rely heavily on experimentation and because of that LLM's starts to shine again due to their capability of easily generating proof of concepts to explore and validate a multitude of scenarios.

I'm not sure I entirely agree with his point, mai

nly because in my teams, the Staff+ engineers are also very present in customer relationships and sales processes, but I understand that can be very particular to each company, so I will take his word for it.

Final thoughts

As I had mentioned in my previous article, LLMs - and AI in general - will disrupt the software engineering field, but the disruption will not be equal at all levels, at least in the short term.

Roles that rely purely or mostly on actual coding skills are much more likely to be replaced by AI because you just cannot compete with artificial intelligence when it comes to repertoire and syntax. AI "knows" pretty much every use case of code in existence and will be able to apply it better and faster than you.

What's left for humans is to fill the gaps where AI just can't perform as well, at least for now.

Posted Using INLEO