What are your thoughts on a future where code is represented as a structured model, rather than text? Do you think that AI-powered coding assistants benefit from that?

Last Updated: 01.07.2025 17:55

What are your thoughts on a future where code is represented as a structured model, rather than text? Do you think that AI-powered coding assistants benefit from that?

i.e. “operator like things” at the nodes …

Most coding assistants — with or without “modern “AI” — also do reasoning and manipulation of structures.

Another canonical form could be Lisp S-expressions, etc.

Has anyone tried Deepseek V3? Does it work?

First, it’s worth noting that the “syntax recognition” phase of most compilers already does build a “structured model”, often in what used to be called a “canonical form” (an example of this might be a “pseudo-function tree” where every elementary process description is put into the same form — so both “a + b” and “for i := 1 to x do […]” are rendered as

a b i 1 x []

These structures are made precisely to allow programs to “reason” about some parts of lower level meaning, and in many cases to rearrange the structure to preserve meaning but to make the eventual code that is generated more efficient.

Nintendo Switch 2 Welcome Tour Review - IGN

/ \ and ⁄ / | \

NOT DATA … BUT MEANING!

It’s important to realize that “modern “AI” doesn’t understand human level meanings any better today (in many cases: worse!). So it is not going to be able to serve as much of a helper in a general coding assistant.

MindsEye team deploys first hotfix as they scramble to make the dang thing work - Polygon

A slogan that might help you get past the current fads is:

in structures, such as:

Long ago in the 50s this was even thought of as a kind of “AI” and this association persisted into the 60s. Several Turing Awards were given for progress on this kind of “machine reasoning”.

This STD is nearly 30 times worse in Onondaga County than it was 20 years ago - Syracuse.com

plus(a, b) for(i, 1, x, […])

+ for