AI Optimism

This is a long-overdue follow-up to my 2023 essay AI Pessimism.

So much has changed in the last couple of years, and maybe even more has changed in the last nine months. December of 2025 was the inflection point in the capabolites of agentic coding tools. I was wrong about them.

We already knew LLMs were really good at writing code. It makes sense. It’s highly structured data and there’s mountains of free training data available. What I didn’t expect was how good LLMs would be at making tool calls, and how straightforward an “agent harness” actually is.

But the cambrian explosion of agentic software engineering happened. It’s no longer the bleeding edge. Developers around the world are expected to use these tools or move on.

There is perhaps no major job seeing as radical a change right now as my own (software engineer). I should be terrified…but I can’t be. I cannot think of a time I have been more excited to wake up every day and get working. The job has changed, but it’s only allowing us to do more, faster.

Jevons Paradox is the by far the leading economic theory people like to talk about with AI in software engineering. Early on, we were worried that making software engineering cheap would remove the need for software engineers. But as LLMs continue to get better and cheaper (I’m in the camp that they’re a commodity), the demand for software engineers will increase. The cost to make software goes down, and the amount of software we will make will explode. We are seeing this play out in real time. Software Engineering jobs are increasing - not falling.

There has never been a better time to be an IC (individual contributor).

Now this is just one example. Yes, there are places where AI will absolutely change or destroy jobs. But my own experiences over the last few years show me that there’s still room for optimism.