I dont think On the foolishness of "natural language programming" by Dijkstra is a gatekeeper, it is just how solving problems is, it is always better to have a specific meaning condensed to a symbol, sure it is difficult to learn, but it pays off.
I still think that at better prompting even in loops is becoming a PM, never a programmer, because in fact you dont program anymore. Yeah, you can use prompts to write code I did for an elixir app wich worked, but I didnt learn anything really about elixir... and I have learn probably more than 10 or 20 or 30 programming languages in my life, but sure I didnt learn a thing when I used the LLM, as any PM doesnt become a programmer just because s/he can talk with you about a feature to develop.
Why people want to use LLMs to do freebsd or any other operative system you want more output like for example check https://github.com/LiberNT/LiberNT which in some months is at least 2x or 3x than the most prominent commiter in reactos in a lot of years of effort. Do people want free bsd or any bsd become libernt really? (also why not just do a fork and code all with claude or watever from now on like he and be X times more productive?).
What would be the purpose of allowing LLMs? just use the tokens you dont have any use of? finally being able to contribute code (I want to contribute too, but I will take the long path and ask humans and maybe some day send a PR by hand, instead of ask the LLM to let me do it in a few hours... maybe jsut read some summaries or ask about things I dont want to botter people here on the forum or on discord...).
If you thought the speed of writing code was your problem - you have bigger problems I link this with Braess's paradox when you think you are solving a hicup but instead you are not solving the actual issue at hand. As the article says... what value are you trying to add allowing llms?
* just using the new thing?
* burn tokens?
* contribute by sending various PRs?
* want to actually learn the internals? (zig doesnt allow llm code as they want the language also be used as a learning plataform for people that can handle that level of complexity).
I also find curious that in the original article it says that they want to add documentation with LLMs... why not all just join a hackathon a weekend and do some actual contribution as a community? why being lazy and not organized at all??? why use the tool when you have the oportunity to make community... or this community doesnt care about inviting people to do "trivial" task of write documentation?
I remember a documentathon on a library I used some years a go, it got almost all covered in some days, maybe some weeks.... not sure how much that documentation helped people later.
I still think that at better prompting even in loops is becoming a PM, never a programmer, because in fact you dont program anymore. Yeah, you can use prompts to write code I did for an elixir app wich worked, but I didnt learn anything really about elixir... and I have learn probably more than 10 or 20 or 30 programming languages in my life, but sure I didnt learn a thing when I used the LLM, as any PM doesnt become a programmer just because s/he can talk with you about a feature to develop.
Why people want to use LLMs to do freebsd or any other operative system you want more output like for example check https://github.com/LiberNT/LiberNT which in some months is at least 2x or 3x than the most prominent commiter in reactos in a lot of years of effort. Do people want free bsd or any bsd become libernt really? (also why not just do a fork and code all with claude or watever from now on like he and be X times more productive?).
What would be the purpose of allowing LLMs? just use the tokens you dont have any use of? finally being able to contribute code (I want to contribute too, but I will take the long path and ask humans and maybe some day send a PR by hand, instead of ask the LLM to let me do it in a few hours... maybe jsut read some summaries or ask about things I dont want to botter people here on the forum or on discord...).
If you thought the speed of writing code was your problem - you have bigger problems I link this with Braess's paradox when you think you are solving a hicup but instead you are not solving the actual issue at hand. As the article says... what value are you trying to add allowing llms?
* just using the new thing?
* burn tokens?
* contribute by sending various PRs?
* want to actually learn the internals? (zig doesnt allow llm code as they want the language also be used as a learning plataform for people that can handle that level of complexity).
I also find curious that in the original article it says that they want to add documentation with LLMs... why not all just join a hackathon a weekend and do some actual contribution as a community? why being lazy and not organized at all??? why use the tool when you have the oportunity to make community... or this community doesnt care about inviting people to do "trivial" task of write documentation?
I remember a documentathon on a library I used some years a go, it got almost all covered in some days, maybe some weeks.... not sure how much that documentation helped people later.