#schemacs — Public Fediverse posts

The challenge of writing portable Scheme code is that every implementation has myriad little idiosyncrasies. Today I discovered the #MITScheme tries to read #Emacs -style file header comments (for example ;; -*- encoding: utf-8 -*-) and this can cause the reader to crash for some reason. I am not sure if it is a bug or a feature, but it did throw me off for a whole day while I tried to figure out why MIT Scheme wouldn’t compile a library file.

I also happen to be hung-up on a bug in #Guile 3.0.11. It seems to me like the optimizer is emitting bad code, because I get a SEGFAULT on an expression that #Gauche has no trouble executing. I am still working on a bug report, and trying to figure out a work-around.

#tech #software #Scheme #Schemacs #SchemeLang #R7RS #Lisp #GuileScheme #GaucheScheme

The challenge of writing portable Scheme code is that every implementation has myriad little idiosyncrasies. Today I discovered the #MITScheme tries to read #Emacs -style file header comments (for example ;; -*- encoding: utf-8 -*-) and this can cause the reader to crash for some reason. I am not sure if it is a bug or a feature, but it did throw me off for a whole day while I tried to figure out why MIT Scheme wouldn’t compile a library file.

I also happen to be hung-up on a bug in #Guile 3.0.11. It seems to me like the optimizer is emitting bad code, because I get a SEGFAULT on an expression that #Gauche has no trouble executing. I am still working on a bug report, and trying to figure out a work-around.

#tech #software #Scheme #Schemacs #SchemeLang #R7RS #Lisp #GuileScheme #GaucheScheme

The challenge of writing portable Scheme code is that every implementation has myriad little idiosyncrasies. Today I discovered the #MITScheme tries to read #Emacs -style file header comments (for example ;; -*- encoding: utf-8 -*-) and this can cause the reader to crash for some reason. I am not sure if it is a bug or a feature, but it did throw me off for a whole day while I tried to figure out why MIT Scheme wouldn’t compile a library file.

I also happen to be hung-up on a bug in #Guile 3.0.11. It seems to me like the optimizer is emitting bad code, because I get a SEGFAULT on an expression that #Gauche has no trouble executing. I am still working on a bug report, and trying to figure out a work-around.

#tech #software #Scheme #Schemacs #SchemeLang #R7RS #Lisp #GuileScheme #GaucheScheme

The challenge of writing portable Scheme code is that every implementation has myriad little idiosyncrasies. Today I discovered the #MITScheme tries to read #Emacs -style file header comments (for example ;; -*- encoding: utf-8 -*-) and this can cause the reader to crash for some reason. I am not sure if it is a bug or a feature, but it did throw me off for a whole day while I tried to figure out why MIT Scheme wouldn’t compile a library file.

I also happen to be hung-up on a bug in #Guile 3.0.11. It seems to me like the optimizer is emitting bad code, because I get a SEGFAULT on an expression that #Gauche has no trouble executing. I am still working on a bug report, and trying to figure out a work-around.

#tech #software #Scheme #Schemacs #SchemeLang #R7RS #Lisp #GuileScheme #GaucheScheme

I am borrowing the term, “The Prime Directive,” from the popular science fiction series Star Trek to describe my AI policy:

Every single line of code, every S-expression, every comment, every piece of documentation, every part of this software that serves a function must have that function be well-understood by at least one person (and of course, AI is not a person).

The term “well understood” is deliberately vague so that we (the Schemacs developer community) can decide on a case-by-case basis whether a person understands a function well enough.

Vibe coding is banned.

The definition of “Vibe Coding” is that you don’t bother to understand what the computer code means, if the software appears to behave in the way you expect you declare the code “correct.” By definition, vibe coding violates the Software Engineer’s Prime Directive.

Some LLM-generated output is OK.

If you suffer from some form of disability, and use an AI coding assistant to write code faster, this is perfectly acceptable so long you adhere to the Prime Directive — you must be able to defend the code you write by explaining the function of all of it’s component parts.

The term “disability” is deliberately vague because I don’t believe we as a community should define tests to decide whether someone qualifies as “having a disability.” If your disability is that you don’t like writing text, that is fine. Just obey the Prime Directive.

Unofficial AI Policy: we don’t need it.

This is a project for people to have fun. If you don’t like writing computer code, then don’t contribute computer code. There are plenty of other ways to contribute without needing AI coding.

There is no hurry, there is no deadline. Write code because you enjoy it, or because you want to learn how computers work. It is OK to take your time and enjoy yourself. There is really no reason to use AI for this project.

#tech #software #AI #LLMs #LLM #Scheme #R7RS #SchemeLang #Lisp #Schemacs #Emacs #EmacsLisp #Slop #VibeCoding #ClaudeCode #ChatGPT #SlopMachine

I am borrowing the term, “The Prime Directive,” from the popular science fiction series Star Trek to describe my AI policy:

Every single line of code, every S-expression, every comment, every piece of documentation, every part of this software that serves a function must have that function be well-understood by at least one person (and of course, AI is not a person).

The term “well understood” is deliberately vague so that we (the Schemacs developer community) can decide on a case-by-case basis whether a person understands a function well enough.

Vibe coding is banned.

The definition of “Vibe Coding” is that you don’t bother to understand what the computer code means, if the software appears to behave in the way you expect you declare the code “correct.” By definition, vibe coding violates the Software Engineer’s Prime Directive.

Some LLM-generated output is OK.

If you suffer from some form of disability, and use an AI coding assistant to write code faster, this is perfectly acceptable so long you adhere to the Prime Directive — you must be able to defend the code you write by explaining the function of all of it’s component parts.

The term “disability” is deliberately vague because I don’t believe we as a community should define tests to decide whether someone qualifies as “having a disability.” If your disability is that you don’t like writing text, that is fine. Just obey the Prime Directive.

Unofficial AI Policy: we don’t need it.

This is a project for people to have fun. If you don’t like writing computer code, then don’t contribute computer code. There are plenty of other ways to contribute without needing AI coding.

There is no hurry, there is no deadline. Write code because you enjoy it, or because you want to learn how computers work. It is OK to take your time and enjoy yourself. There is really no reason to use AI for this project.

#tech #software #AI #LLMs #LLM #Scheme #R7RS #SchemeLang #Lisp #Schemacs #Emacs #EmacsLisp #Slop #VibeCoding #ClaudeCode #ChatGPT #SlopMachine

I am borrowing the term, “The Prime Directive,” from the popular science fiction series Star Trek to describe my AI policy:

Every single line of code, every S-expression, every comment, every piece of documentation, every part of this software that serves a function must have that function be well-understood by at least one person (and of course, AI is not a person).

The term “well understood” is deliberately vague so that we (the Schemacs developer community) can decide on a case-by-case basis whether a person understands a function well enough.

Vibe coding is banned.

The definition of “Vibe Coding” is that you don’t bother to understand what the computer code means, if the software appears to behave in the way you expect you declare the code “correct.” By definition, vibe coding violates the Software Engineer’s Prime Directive.

Some LLM-generated output is OK.

If you suffer from some form of disability, and use an AI coding assistant to write code faster, this is perfectly acceptable so long you adhere to the Prime Directive — you must be able to defend the code you write by explaining the function of all of it’s component parts.

The term “disability” is deliberately vague because I don’t believe we as a community should define tests to decide whether someone qualifies as “having a disability.” If your disability is that you don’t like writing text, that is fine. Just obey the Prime Directive.

Unofficial AI Policy: we don’t need it.

This is a project for people to have fun. If you don’t like writing computer code, then don’t contribute computer code. There are plenty of other ways to contribute without needing AI coding.

There is no hurry, there is no deadline. Write code because you enjoy it, or because you want to learn how computers work. It is OK to take your time and enjoy yourself. There is really no reason to use AI for this project.

#tech #software #AI #LLMs #LLM #Scheme #R7RS #SchemeLang #Lisp #Schemacs #Emacs #EmacsLisp #Slop #VibeCoding #ClaudeCode #ChatGPT #SlopMachine

I am borrowing the term, “The Prime Directive,” from the popular science fiction series Star Trek to describe my AI policy:

Every single line of code, every S-expression, every comment, every piece of documentation, every part of this software that serves a function must have that function be well-understood by at least one person (and of course, AI is not a person).

The term “well understood” is deliberately vague so that we (the Schemacs developer community) can decide on a case-by-case basis whether a person understands a function well enough.

Vibe coding is banned.

The definition of “Vibe Coding” is that you don’t bother to understand what the computer code means, if the software appears to behave in the way you expect you declare the code “correct.” By definition, vibe coding violates the Software Engineer’s Prime Directive.

Some LLM-generated output is OK.

If you suffer from some form of disability, and use an AI coding assistant to write code faster, this is perfectly acceptable so long you adhere to the Prime Directive — you must be able to defend the code you write by explaining the function of all of it’s component parts.

The term “disability” is deliberately vague because I don’t believe we as a community should define tests to decide whether someone qualifies as “having a disability.” If your disability is that you don’t like writing text, that is fine. Just obey the Prime Directive.

Unofficial AI Policy: we don’t need it.

This is a project for people to have fun. If you don’t like writing computer code, then don’t contribute computer code. There are plenty of other ways to contribute without needing AI coding.

There is no hurry, there is no deadline. Write code because you enjoy it, or because you want to learn how computers work. It is OK to take your time and enjoy yourself. There is really no reason to use AI for this project.

#tech #software #AI #LLMs #LLM #Scheme #R7RS #SchemeLang #Lisp #Schemacs #Emacs #EmacsLisp #Slop #VibeCoding #ClaudeCode #ChatGPT #SlopMachine