<quote>
I just discovered, very belately, that these new-fangled AI engines or large-language-models or whatever it is that they are, or wish to be called, are actually capable of doing some pretty amazing things. I decided to test out Google's gemini after someone less than one third of my age told me about it, and I became amazed by it’s ability to translate code from one language to another, or write code snippets based on instructions that I gave it. It even seems to understand, more or less, the “borrow checker” in the Rust language. This was yet another humbling experience, but it confirmed what I have long thought: that writing code is not a particularly high mental function.
But what I found even more disconcerting, is that Gemini can actually translate the nom translators in the /tr/ folder into other languages. Which is very close to meaning that I didn't really have to bother spending all that time writing translators for nom. I could have just written one good one, and then used Gemini to tranlate it into other languages. There are actually a few caveats to the sentence above, but it is true that an AI engine can save vast amounts of hack-coding time.
And I intend to use it to translate, for example, the rust translator into other languages so that nom scripts can be translated into, and used from, those languages. It doesn't really matter if the translation is not perfect, because the structure of the pep virtual machine is pretty simple so it should not be difficult to debug the generated code if it has errors in it. In some ways it seems fitting to use an automatic translator (Gemini) to translate the automatic translator-translator which is the ℙ𝕖𝕡 🙵 ℕ𝕠𝕞 system.