←back to thread

Anthropic’s paper smells like bullshit

(djnn.sh)
1160 points vxvxvx | 2 comments | 16 Nov 25 11:32 UTC | HN request time: 0.432s | source

Earlier thread: Disrupting the first reported AI-orchestrated cyber espionage campaign - https://news.ycombinator.com/item?id=45918638 - Nov 2025 (281 comments)
Show context
gpi ◴[16 Nov 25 14:44 UTC] No.45945486[source]▶
The below amendment from the anthropic blog page is telling.

Edited November 14 2025:

Added an additional hyperlink to the full report in the initial section

Corrected an error about the speed of the attack: not "thousands of requests per second" but "thousands of requests, often multiple per second"

replies(2): >>45946057 #>>45947275 #
1. wging ◴[16 Nov 25 18:34 UTC] No.45947275[source]▶
> The operational tempo achieved proves the use of an autonomous model rather than interactive assistance. Peak activity included thousands of requests, representing sustained request rates of multiple operations per second.

The assumption that no human could ever (program a computer to) do multiple things per second, nor have their code do different things depending on the result of the previous request is... interesting.

(observation is not original to me, it was someone on Twitter who pointed it out)

replies(1): >>45947920 #
2. sublimefire ◴[16 Nov 25 19:58 UTC] No.45947920[source]▶
Great point, it might be just pure ignorance. Even OSS pentesting tooling such as metasploitable have great capabilities. I see how LLM could be leveraged to build custom modules on top of those tools or how can you add basic LLM “decision” making, but this is just another additive tool in the chain.