Now I believe that the current hype around AI and LLMs may cause an increase in average salary in the software industry.

Being part of the startup ecosystem from the very first day, I always carried the doubt of not having a job for a while, but not for very long. After I reached a point where I proved myself to myself, I didn’t have much of a doubt about job security, excluding catastrophic situations. The spark of LLMs took me to the very first days of my career.

The burning question in the wake of LLM-powered world was, and probably still is, will I, a software engineer with 10 years experience, lose my job and financial stability? Maybe not today, hopefully not tomorrow, what about a few years from now?

I have been working with LLMs integrated into my daily work for a while. Since I am working with code and configuration, using LLMs is a productivity boost without a question. Ability to understand existing codebase, auto-generate well-defined functions and snippets, auto-complete using Copilot and alike, etc. are all fantastic tools we wouldn’t anticipate 10 years back. Although I hate that it doesn’t push back enough like a normal person does, they are quite useful when somebody knows what they are doing.

When I think about what I do every day, the level of detail I think and operate, LLM-powered AI is not likely to replace me, as an operator, a software craftsman, or a manager, in its current state.

Having said that, it shows hard-to-miss signs that it is changing the toolset I’m accustomed to, the very definition of my day-to-day job, the mental models I have built so far. I don’t think the goal at its core does not differ much, but day-to-day activities are changing a lot.

When distilled, explosion of LLM-powered practices in software shops comes down to two simple outcomes:

• The speed and amount of code generated per time unit has increased (e.g. vibe coding, AI-powered coding, agents etc.)
• Business owners tend to see this as an opportunity to reduce the cost of building software

In even simpler terms, we have even smaller teams working on larger contexts, with higher cognitive load. We work on multiple stacks, making a dream come true: the real full-stack engineer, working end-to-end, from refining requirements and UX optimizations to database administration and deciding load balancer strategies.

My current mental models tell me that the last sentence above is an exaggeration, no one can do that. I want to believe that as well, but I’m open to breaking this wall too. In a way, I’m preparing myself.

New responsibilities

Starting very recently, the definition of a software craftsman started to include more and more things:

• Wide understanding of the overall system
• Ability to articulate what you want to do
• Ability to analyze & sign off on what AI suggests

I believe these are necessary to keep a good operational stability for our products. If your service does not require acceptable quality, I assume you can let some coding agent take the helm and steer the ship, someone will fix it when it’s necessary.

I still want to believe these are good practices to follow, to keep our daily interaction with software technologies sane. Keeping standards above acceptable levels is an every-day effort. Using LLM-powered GenAI to develop these systems requires a special kind of developer. It means that the average developer quality has to increase.

With higher capacity, higher level of understanding, and much better tooling, one-person teams deliver more output; therefore I suppose their compensation will increase proportionally.

Fewer people will earn more. The total expenditure of companies on maintaining software teams will decrease. Hiring decisions will be made over one simple constraint: cognitive load per person. AI-supported developers will have more capacity and handle more work.

What about competition among developers?

There is already an intrinsic barrier of entry to the industry from a technical perspective. I believe this barrier is now higher than before and it will get even higher. With a higher barrier, a more capable person will have the advantage, so competition should follow similar pattern as before.

In my home country, there was a recent spike in the last decade because many people thought software industry is easy to enter, salaries are high, and quality of life is rather better compared to the alternatives. Nowadays, I read post after post that people are desperate and looking into alternative career paths. I suppose this is a common issue generally, since hirings have stagnated globally starting 2023.

I think we will start to see this pattern once the initial fever is over. We already see it in crazy salaries paid to AI engineers in frontier companies, such as Meta paying football player transfer fees to engineers, but it looks like it is isolated to a very small group of people right now. For the rest of us, there are many things to learn.