AI advantage?
One possible advantage of Nytril being a programming language is
that writers might not need to learn it in depth. They could describe
what they want and let an AI coding tool write the Nytril code. AI
tools are also strong at writing code, so documents defined in code
might give them more to work with than prose-based formats. Is that
a realistic workflow today, given that Nytril is new and LLMs haven't
seen much of it in training? What are you doing to make Nytril easy
for AI tools to work with (for example, reference docs, examples, or
an llms.txt file)?
You are right about the potential. Given that Nytril allows you to specify the entire document
in code, it allows an AI to mix queries, logic, translations and formatting together to create
code for the requested documents.
So far, we have been able to get LLMs to understand the content of our site. Since our site is
simply a big Nytril project, this means that users who publish to the site will benefit from
having their content published in a way that is digestible to Google and AI crawlers.
As far as AI generated Nytril code, we have had success there as well. The following diagram
was generated by ChatGPT after reading our
tutorial document
, and being asked to produce a
drawing of a machine in Nytril code. The generated source for this diagram is at the very end
of this document.
It is early days on the LLM front, but we foresee that AI systems will be able to deliver
documents in 'source form' rather than as DOCX or PDF files. This is vastly more useful for
content creators. It will allow writers to edit and correct the inevitable hallucinations, and use
that work as a permanent resource for future documents.