Applause [LWN.net]

Applause

Posted Jul 23, 2026 16:26 UTC (Thu) by pizza (subscriber, #46)
In reply to: Applause by Poliorcetics
Parent article: Codeberg: Protecting our FLOSS commons from LLMs

> Just today I had to close 5 MRs on a GitHub project where the descriptions were about 50 times the size of the changes

Last week I closed two "security issues" (which to their credit, included "fixes") that were premised on an extremely verbose "analysis" that fundamentally misconstrued the operating environment of the code in question [1]. This reduced the scope from OMGCRITICAL "security vulnerability" to mere bugs, except it also got the specifics completely wrong.

Then, after closing those issues, I re-disabled MRs entirely as the GH repo has always just been a secondary mirror.

[1] Completely offline device. One would need direct physical access [2] to exfiltrate data after a successful attack. But if one already has direct physical access... why not just exfiltrate the data directly in the first place?
[2] ie connect to it via USB or eject an SD card and plug it into a card reader

to post comments

Applause

Posted Jul 25, 2026 19:37 UTC (Sat) by grmnsftphr (subscriber, #178591) [Link] (1 responses)

Same here too, total through-the-rough CVE where a service client had to shoot himself, with no RCE or PE possible. The reporter even copy and pasted the CVE text containing hebrew text. He had no clue what be was doing, this was very clear and he was the idiot controlled by the LLM bot and a hidden company behind it.

Also seen self-styled security experts sending patch slop but refuse to explain and fill in the COA or similar. Their users on github are full of nonsense fake slop repos and fake bios, but no publications could be found anywhere.

Send these AI freaks to all levels of Dante's inferno simultaneously, a plague on them and their supporters.

Not applauding anymore

Posted Jul 25, 2026 19:45 UTC (Sat) by corbet (editor, #1) [Link]

Please, we can discuss the technology without name-calling and such.