Are LLMs to Python as C was to assembly?

Last updated on March 4, 2025

Topics: ai

When assembly came it made knowledge of transistors useless. When C came along it made knowledge of assembly useless. When Python came along it made knowledge of C useless.

Now as LLMs are coming along all coding will become useless.

But wait… did Python make C knowledge useless? During the transition from more lower level memory type languages like C to higher level dynamic languages like Python knowledge of how memory is structured was invaluable. The C knowledge provided a scaffold on which to hang the new Python knowledge.

But wait… was the knowledge of the lower level of abstraction just useful in the transition period. I mean surely now no one needs to know anything about how the memory is laid out. The answer is no! Right now the people building databases know a crazy amount about memory and low level structures! Everyone uses their stuff every day.

Perhaps every higher level abstraction that comes along provides the opportunity to ignore the lower level preceding it. With Python you can just ignore memory and allocation. But the option of getting into the details of how the memory works or how assembly works or how transistors work or how atoms work is always there.

What am I trying to say here? There is a lot of talk (citation not needed) about LLMs replacing coding and coding becoming obsolete. A lot of people take joy in writing their code and don’t want this to go away.

I believe that this will not happen. I believe it will be the same way as Transistors -> Assembly -> C -> Python. I believe there will always be a need for getting into the lower levels of the abstraction onion.

Perhaps with LLMs people can easily build simple webpages with 1% of the effort it would have taken before. But to do hard things like building databases we need people who understand things deeply.

Perhaps LLMs are a unique abstraction and they will not follow the trends of its predecessors. Perhaps this is “cope”. Time will tell.