I think the core insight of this article, and it's something I haven't thought about before in this way, is that any successful programming language needs to "market itself" to LLMs and LLM developers.
We had the "first wave" of languages that were marketed to decision makers (Java), the second wave that was marketed to developers (Ruby on Rails, Rust), I think a third wave is coming, where the language will be designed to be LLM friendly from the ground up.
I was looking for a programming language that's being written with LLMs in mind, and used ChatGPT to give me some pointers so I can start some research https://chatgpt.com/share/6841d3b7-13a8-800f-9bcc-0f7859114c...
In my mind (I am a non-academic that never really designed a new programming language), I can imagine strong typing, terseness, and no optional syntax (you can do this multiple ways) would be really cool for LLMs.
I did come across a few languages / people trying to design programming languages to be used with LLMs / AI Agents, and I think the most promising thing I found is https://www.moonbitlang.com/blog/moonbit-ai - which takes the whole idea of working with LLMs to another level, with actual features designed to work with agents (not only syntax).
Moonbit lang is really interesting. Looks like it's been submitted but ignored by HN.
If you wanted to go all in, you'd want a language that is denser so you get more efficiency out of tokens. Since it's being used by LLMs, readability by humans can take a back seat. Obviously the answer is: Perl.
But seriously though, a high-level language designed for LLMs would be interesting, maybe with a translation layer to something more developer friendly on top.
High information density is good for LLMs. https://ash-hq.org benefits from this, but hurts due to being fairly new. https://hexdocs.pm/usage_rules is helping.
> If you wanted to go all in, you'd want a language that is denser so you get more efficiency out of tokens.
That's Lisp because nothing compresses the number of tokens you have to write more than Lisp Macros.
APL is the densest language I have come across.
There may not be enough code out there for LLMs to pick it up though.