On 2026-09-05 05:38, Jose E. Marchesi wrote:
The GCC project has adopted the policy that no LLM generated code which
is legally significant shall be installed, and the policy explicitly
refers to the GNU Project's definition of "legally significant".

Gnulib's HACKING file also mentions that definition. That being said, its 
current wording about single LLM prompts is a bit worrisome, as it can be too 
easily misinterpreted to mean that it's OK to stitch together many prompts or 
to use an agent to stitch together prompts for you. (Plus, to some extent the 
idea of prompt engineering is soooo 2024.... :-)

How about the attached patch to Gnulib? It attempts to clarify HACKING to avoid 
this kind of misinterpretation. The idea is that a single prompt is just an 
example, and what matters is the larger principle.
From b7656b3847c7ae7a947a05ada36e241d4a37d448 Mon Sep 17 00:00:00 2001
From: Paul Eggert <[email protected]>
Date: Sat, 5 Sep 2026 10:14:52 -0700
Subject: [PATCH] Generalize LLM advice to non-prompt models
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* HACKING: Don’t assume traditional LLM prompt model.
---
 ChangeLog | 5 +++++
 HACKING   | 4 +++-
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/ChangeLog b/ChangeLog
index 7fb59f0be5..ddd79187e0 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,8 @@
+2026-09-05  Paul Eggert  <[email protected]>
+
+	Generalize LLM advice to non-prompt models
+	* HACKING: Don’t assume traditional LLM prompt model.
+
 2026-09-04  Oleg Tolmatcev  <[email protected]>
 
 	fts: don't use FTS_CWDFD where openat is emulated via fchdir
diff --git a/HACKING b/HACKING
index 9782291de1..ed32ebf448 100644
--- a/HACKING
+++ b/HACKING
@@ -261,7 +261,9 @@ We need to avoid two problems with that:
 
 To this end:
 
-  1) Code included in this package that comes from a single LLM prompt
+  1) Code derived from an LLM must respect copyrights and licenses
+     in the source material that the LLM is based on.
+     For a traditional, single LLM prompt this means a code response
      must be limited in size: it must be at most 5 lines long.
 
   2) As a submitter, you assert that you have reviewed such code that you
-- 
2.53.0

Reply via email to