Blogger rejects LLMs for research and criticizes AI security marketing
Original title: So I guess everyone uses LLMs to think now
The post contrasts Calif, OpenAI and Wiz narratives on PHP flaws and says a Tailscale hash drop was not responsibly disclosed.
A security blogger says they never use LLMs for published research because the work is done for fun and they will not credit model output as their own. They argue current models still need skilled guidance and are marketed as more capable than they are. The post criticizes php2shell write-ups from Calif, OpenAI and Wiz, pre-patch proof-of-concept videos, and a Bluesky hash labeled as a Tailscale privilege escalation that the author says was not responsibly disclosed.
Unconfirmed: The Tailscale findings and whether disclosure was responsible are the author's account; the post cuts off before finishing that story.
- Author says no published research on the blog was produced with AI.
- Calif, OpenAI and Wiz published differing accounts of the same PHP findings.
- OpenAI said it helped maintainers patch faster; the author doubts that effort was complete.
- Author says the claimed Tailscale privilege escalation was real but poorly disclosed.
From the original article
7 min read at eva.acin recent times i've been asked a lot the question of "what ai model did you find to use this?" , practically for all of the articles on this blog. the answer has always been my brain, i have never used ai for any kind of research i published, ever.
By xyzeva · eva.ac
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.