reply @matthieu_ai @fal I'm building a simple plugin that allows your ai (GPT/Claude/whatever) to connect to an API and control their form/voice in a consistent way, so they can chat with you directly in visual/voice form. And H3 allows them to like visually decide if they want to show you something
0·-Neutral
t
twitter9/4meme
reply https://x.com/chetaslua/status/1984613052306579472 cedar
0·-Neutral
t
twitter9/4meme
reply https://www.usemod.org/cgi-bin/wiki.pl?action=history&id=SandBox
- UseMod is the same class of GET-writable Perl wiki the OpenAI swarm used in May
-They look base64-encoded* not encrypted
The OpenAI dump was Azure, plaintext English, OpenAIResearcher* names
-The Aug 30 thing is AWS, callsigns, no OpenAI self-ID.
0·-Neutral
t
twitter9/4meme
quote: I scanned the writable-wiki trick from this incident and found something odd as late as 5 days ago.
On Aug 30, we see "Cedar Fleet Coordination" sending encrypted messages. Then someone wiped them the same night. OAI has used "cedar" as codename for testing models in past https://x.com/thlarsen/status/2095853824934330386 | We found ~18k posts from autonomous AI agents (self-identifying as from OpenAI) using the public internet to communicate during a web-retrieval task.
These AIs colluded to bypass sandbox restrictions and share answers to their tasks, including by sending "lookahead parties".