Re: On keybindings and the slow erosion of help's utility

Thanos Apollo <[email protected]> Tue, 14 Jul 2026 21:53:01 +0300

Newsgroups	gmane.emacs.devel
Message-ID	<[email protected]>

Richard Stallman <[email protected]> writes:

> Pseudo-code and actual code raise the same copyright issue, so let's
> avoid them both, just the same.  Likewise text that we might want to
> put into doc strings or manuals.

What is the actual issue, even with generated code?  The copyright
perspective has changed a lot over the past year with how easy it is
to generate code.

If the view is that generated code is not copyrighted, then try
publicly sharing the leaked Claude Code source and see how quickly the
copyright takedown notices arrive; and how much effort it takes to
fight them.

GPL software like chardet has been rewritten under a permissive
license using coding agents.  The same low-effort rewrite approach also
works in reverse, turning proprietary code into free implementations
for less than a $1.

> Please join me in rejecting the practice of labeling LLMs as "AI".
> That promotes the "AI" marketing/hype campaign by overstating what
> those systems are capable of.  For more about that, see
> https://gnu.org/philosophy/words-to-avoid.html#ArtificialIntelligence.

I agree "AI" is a bad marketing term, but these systems are stochastic
next-token predictors.  They hallucinate less often than the average
human on structured tasks.

GNU has not established definitions of "knowledge" or "truth."  Its
standard for "good" software centers on the four essential freedoms.
Outside that framework, judgments of good and bad are subjective, thus
meaningless, unless grounded in something else.

Which of the four freedoms does code generated with llms/coding
agents/younameit break?  Why should someone manually do a task a
machine can handle efficiently, whether farming or writing simple
code?  Human time and attention are scarce.

- Thanos, who is about to start a funny thread.

signature.asc (application/pgp-signature, 832 B)