https://sourceware.org/bugzilla/show_bug.cgi?id=34498
Serhei Makarov <serhei at serhei dot io> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |ASSIGNED
Assignee|unassigned at sourceware dot org |serhei at serhei dot io
Ever confirmed|0 |1
Last reconfirmed| |2026-08-12
--- Comment #4 from Serhei Makarov <serhei at serhei dot io> ---
This issue also applies to the eu-stackprof code, and requires an adjustment to
the libdwfl_stacktrace arch constant patches currently under review. Timely
catch.
my thoughts re: LLM review,
- a human writeup of the LLM finding tends to be more concise (saves the reader
time)
and signals willingness to triage the results and engage
- an LLM writeup is not limited by human fatigue and tends to be
verbose/redundant;
the work of compressing/evaluating the results is shifted to the recipient.
- submitting multiple LLM writeups in a row crosses a certain line:
the quality of the work is unknown/unverified until the wall of text is waded
through,
so these kinds of contributions are at greater risk of being deprioritized
- it's hard to formulate defined policies, because this is a politeness issue
rather than a violation of some rule a project could plausibly enforce
- There's an empathy gap because people who work with LLMs extensively are
desensitized to the fatiguing qualities of LLM prose and trained to skim,
hence have difficulty understanding how disadvantageous it looks next
to more concise reports, from the PoV of an audience used to reading every
word.
--
You are receiving this mail because:
You are on the CC list for the bug.
