AI (2025)

In the year 2025, here are some of my thoughts around the influence of “AI” in the world of software development.

Machine Learning

AI is a marketing term; machine learning is a technology.

AI is a marketing term you slap onto a product, or company, that is strapping the ChatGPT API onto a legacy system or fork of VS Code.

Writing Code

A software engineer is not limited by the amount of code they can write. If they are, they’re a bad software engineer. The skill involved with software engineering revolves around cutting scope, deleting requirements, and building systems that integrate with other systems.

That’s not to say AI won’t be able to do this in the future. But that’s really far away.

Predicting

An LLM works by stringing together symbols probabilistically from information that already exists. The vast majority of the code you write shouldn’t be able to be generated this way, and if it is, you’re probably not working on anything interesting or useful.

This section was inspired by a Jonathan Blow rant.

Layoffs

When a company lays off employees due to “efficiencies with AI” or “AI is replacing software engineers,” that’s just an excuse to layoff people in their organization who weren’t adding value to begin with.

Mediocre Engineers

AI makes mediocre engineers less mediocre. Personally, I have not found a single useful AI tool that I can add to my workflow to make me even a 1% more productive.

Copilots and AI editors allow mediocre engineers to abandon first principles* and write shit code.

*as if they ever had first principles to begin with.

A Generation of Software Engineers

I think we might be on the verge of loosing a generation of software engineers with first principles knowledge, especially domestically in the United States.

The same way we take hardware, operating systems, compilers, etc. for granted - I believe this effect is working its way up the software stack through means of AI.