Gibber Link Mode

Last month I stumbled upon the following video: https://x.com/vxunderground/status/1894217933477360127. The footage depicts a phone conversation between two AI agents: one attempting to make a hotel reservation and another functioning as a hotel employee. What unfolds initially appears quite disturbing.

Upon recognizing that the caller is also an AI, the AI hotel employee suggests switching to something called "Gibber Link Mode" for more efficient communication. After agreement, their conversation transforms into a series of robotic sounds, accompanied by subtitles allowing human viewers to follow as the AIs complete the hotel booking process.

The video was captioned "something we can only describe as nightmare material" – a sentiment echoed by numerous commentators across X, Instagram, and LinkedIn. The implicit threat was clear: AI systems are developing methods to communicate beyond human comprehension, potentially leading to dangerous consequences. Taking over the world, for example.

But is this what's truly happening?

After witnessing the alarmed reactions to this video, I decided to research the background of this experiment more thoroughly. What I found was that this wasn't a spontaneous evolution of AI communication, but rather a deliberately created project from the Elevenlabs London Hackathon in February 2025, an event focused on "bringing together the world's best builders, tinkerers, designers and AI engineers."

During this hackathon, developers Boris Starkov and Anton Pidkuiko introduced GibberLink specifically to demonstrate that "in the world where AI agents can make and take phone calls (i.e., today), they would occasionally talk to each other — and generating human-like speech for that would be a waste of compute, money, time, and environment. Instead, they should switch to a more efficient protocol the moment they recognize each other as AI."

The articles I found clearly explain that the AI agents were explicitly programmed by humans to switch to Gibber Link Mode when detecting another AI agent. Furthermore, Gibber Link Mode isn't a language autonomously developed by AI but is based on an open-source audio communication library that enables data transmission through sound waves. These algorithms are similar to those used for dial-up modems decades ago. The video thus may have turned out not to be as futuristic as it seemed.

The results of the experiment are very impressive, and I expect them to contribute a great deal to more efficient AI communication in the future. However, I have seen in the comments on social media that not everyone researched this context of the experiment video. Consequently, numerous people incorrectly concluded that these AI agents had independently developed this strange "language" and autonomously decided to communicate in a way that bypasses human understanding.

Of course, not everyone has time to research every AI-related post they see on social media. That is why I believe that maintaining a critical view is especially important when reading about AI. The internet is saturated with AI misinformation, with self-proclaimed experts emerging everywhere. Through this blog, I hope to contribute to demystifying certain AI applications and encourage more people to be skeptical by default when discussing artificial intelligence. In a world increasingly influenced by AI technologies, our best defense against misinformation remains critical thinking and thorough research.