John Carmack on AI in game dev

"My first games involved hand assembling machine code and turning graph paper characters into hex digits. Software progress has made that work as irrelevant as chariot wheel maintenance."

John's talking about tooling (computers are probably the ultimate tools to make tools,) but let's not discount the increases in compute capability. Let's put his "first games" phase at 1990 for the sake of making a nice even number. Wow, computers were slow then vs 2025.

Component

Speed Increase Since 1990

CPU (per core)

~1,000x

CPU (multi-core)

10,000x+

Storage

~5,000x

RAM

~1,000x

Combine that with the rise of custom tooling, open source, and so many abstraction layers, and one could argue that there've been some massive improvements in programmer productivity. With that said, I think there are some complicated demand curves that make it hard to draw a line to more/less jobs in the field.

What I do think is that increased awareness of what's possible has led to increased demand for people who can make those things possible, and I think there's grounds to believe that that cycle will continue. What's up for debate, perhaps, is how those things will be made possible in the coming years, which is going to directly affect the kinds of jobs that are available.