← Blog

Nov 7, 2026 · ryangtanaka

AI "Expertise" is a Literacy Problem, not a Labor Problem

I think one misconception that people have about AI right now is that its current version runs largely on LLMs - which is a translation service, not a service that actually does any "work". It's made explicit right in the definition itself: Large LANGUAGE Models!

The debate over whether or not AI can actually replace human software engineers is actually better understood when you look at it from a literacy angle, rather than a labor angle.

I've been using AI to translate English into Japanese, and visa versa, and for the most part, it actually does it very well. I get a bit nervous when I use AI for communication (it can come across as disrespectful if you're not careful) but since I'm fluent in both, I can at least double-check the "result" to see if what I'm saying is OK.

You can apply the same idea of "fluency" to basically everything else: coding, art, writing...with most creative activities, using AI to assist you with the work you're already doing is generally seen as OK, even among most skeptics out there. Using AI to spit out "work" you don't understand, less so. If you're living in the modern world, the use of machines is an inevitability - but at the end of the day, you need to understand what you are actually looking at.

In a lot of ways, that's where the "productivity" of AI is actually coming from - it's value largely comes from the act of "translating" the entangled mess of programming languages that makes computer systems frustrating to work with - it translates my plan English instructions into python, React, Javascript, GraphQL, Michelson, and a host of languages that I never had the time to learn and turns it into a "result".

For a lot of folks living in those worlds, this resolution feels like some "work" has been done. But it's actually closer to the act of filing paperwork or having navigated a bureaucratic process that...well, maybe shouldn't have existed to begin with. And therein lies the problem - if people start confusing procedures with real work, the idea of the robots "replacing" human beings starts feeling like it could be a real possibility. But real work comes from solving problems - and solving problems don't come from translations. They are both important in their own ways, but they are not the same thing.

The confusion between the two is real. Clouded by the immense amount of money being dumped into research right now, it's quite easy to get caught up in the irrational sides of the hype - and I was no exception to it, either. But after experimenting with these tools for a little while, I think this is what the industry will settle into, eventually. It's just a matter of time.