Engineering video
1:19 · 4 August 2025
In the speaker’s view, coding models excel at low-complexity tasks in popular languages and frameworks like JavaScript and React due to abundant training data, making AI particularly efficient at generating frontend component code compared to backend languages like Haskell or enterprise Java.
Hear it at 0:00 →In the speaker’s view, because AI is most efficient in the frontend, an estimated twenty-five percent or one in four developer jobs are being automated, reducing the need for certain frontend developers.
Hear it at 1:09 →why is that AI works better in the front? So one of the best data we have in software engineering productivity comes from Stanford and what they found by studying private repositories in the enterprise. Look at the actual work getting done, they found out that the lower the task complexity and the more popular the language, the more abundant data in model tuning you have. So, coding models work better with the most popular languages and frameworks.
And if we look at the open source code that those models were trained on, a lot of it was JavaScript and specifically React JS, because this was uh the most popular framework for the last 10 years since 2015 when it came out. And so it's a lot easier to generate component code, for example, with AI than it is to generate Haskell code for some backend endpoint, or even things like Java do not have that much data, uh because again, those models were trained on public code and a lot of Java code that's written in enterprise is not public.
So all in all, it basically means AI, it's most efficient in the front end. So a lot of uh developer work getting automated. We're talking about 20 25%. That's one developer in four. That you probably don't need one front end developer because of AI.
