Human-written. I was trying to be short and straight to the point.

LLM-powered rewrites and huge refactors are better done using 1 additional step "convert the code to <something> that represents it best".

The simplest example is, for a CRUD app it can be swagger description. The more complex behaviour exhibit the app, the more raw information should be provided.

Like ontologies, "A is a child of B" model can derive and enforce that "B is a parent of A", and so on.

On top of that, I write that Fable is reasonably cheap if one uses it solely for agent orchestration.

▲ UltraSane 4 days ago | parent | next [-]

I found this part to be interesting/clever:

1. Extract the data representation

Ask the LLM to represent your code as any combination of:

    graphs
    ontologies
    hierarchical state machines
    UML process charts
    constraints
    math formulae

2 Operate on the representations

3 Convert representations back to code

▲

richstokes 4 days ago | parent | next [-]

Do we think that was necessary? What would have happened if OP had just asked it to rewrite and test/validate each piece as it went until everything is verified and complete?

My gut feeling is this is doing way too much, and it would've figured it out.

▲

aka-rider 4 days ago | parent | next [-]

But we know, Bun was 535496 lines for $165000.

The whole point of this experiment was to try and make the rewrite as cheap as possible.

▲

never_inline 3 days ago | parent [-]

Oh no, bun had quite a bit of machinery: https://bun.com/blog/bun-in-rust#loops-that-write-review-cod...

	▲	aka-rider 3 days ago \| parent [-]
		True, the main problem is how to split the work. They decided to go file by file. To be fair, I haven't try exactly "rewrite in Rust" prompt, but I cannot imagine for this to work.

▲

UltraSane 4 days ago | parent | prev [-]

converting code to more abstract and denser representations and then manipulating them makes sense to me. Finding better representations is like half of mathematics.

	▲	richstokes 2 days ago \| parent [-]
		Is code being converted to / represented in maths that much more helpful for an LLM that deals in language primarily, by definition? I have no idea, interesting though!

▲

aka-rider 4 days ago | parent | prev [-]

this is the meat, yes.

▲ brazukadev 4 days ago | parent | prev [-]

honestly it is hard to believe that seeing your replies and this heading: "The secret sauce".

Giving the benefit of doubt, we all might be writing a bit like claude nowadays.

If that is the case, I'd recommend reviewing the content before publishing to see if it sounds like a LLM.

Or if you are trying to create "better" AI slop and think that is enough to say the text is human-written, don't do that, just say it was AI-generated or assisted.

▲

aka-rider 4 days ago | parent [-]

I wrote this elsewhere. Reading so much LLM output may have affected how I write.

Probably I need some fresh air and a good fiction book.

	▲	abrookewood 4 days ago \| parent [-]
		We'll all be writing this way before long ...