Re:Show HN: Marsha – An LLM-Based Programming Language
by ModernMech on Tuesday July 25, 2023 @03:19PM
(#36867913)
Because a lot of the time, exactly how is very important. The devil is in the details.
Re:Show HN: Marsha – An LLM-Based Programming Language
by pombo on Tuesday July 25, 2023 @04:01PM
(#36868479)
I would say that it depends. Python and JS, two of the most used programming languages, abstract away memory management and threading details that most of the time you don't need, but you can "drop down" to write code that does pay attention to that within the language (building an object pool and reusing it, or memoization, etc) and if that's not enough, go to the "deeper" language like C/Rust to handle those cases with first-class primitives.
Re:Show HN: Marsha – An LLM-Based Programming Language
by jsight on Tuesday July 25, 2023 @05:02PM
(#36869490)
That's what all the C engineers told me about manual memory management.
Of course, there are cases where they were right.
Re:Show HN: Marsha – An LLM-Based Programming Language
by pombo on Tuesday July 25, 2023 @04:03PM
(#36868506)
I'm not sure I agree with the fact that an abstraction is always the answer, but your last sentence outlines precisely one of our motivations behind Marsha. You can specify an imperative set of steps in the description of a function/program, or you can (in the future) write a Python function that you use from Marsha
Why not abstract away the "how" layer and focus completely on the "what" layer?
For a lot of data processing and integration problems, this would both eliminate a lot of work and increase reliability.