The GibberLink project, created by two Meta engineers during a hackathon in London, went viral on social media, sparking significant interest in the capabilities of generative AI. GibberLink enables AI agents to recognize each other during a phone call and switch to a more efficient communication protocol called GGWave. This protocol uses sounds to represent small amounts of data, allowing computers to communicate faster and more efficiently than is possible with human language.
Today I was sent the following cool demo:
— Georgi Gerganov (@ggerganov) February 24, 2025
Two AI agents on a phone call realize they’re both AI and switch to a superior audio signal ggwave pic.twitter.com/TeewgxLEsP
The GibberLink demonstration, reminiscent of scenes from science fiction movies, garnered over fifteen million views on the X platform within a week after the hackathon. Even popular tech blogger Marques Brownlee shared the demo video, highlighting the growing curiosity about the future of AI agents. Although the technology underlying GibberLink is not new and originates from 1980s modems, its potential application in today’s context is sparking much discussion.
The project’s creators, Borys Starkov and Anton Pidkuiko, note that GibberLink is not yet commercialized and is not connected to their work at Meta. The project is open to everyone on GitHub, and its authors plan to work on additional tools in their spare time. Meanwhile, the hype around GibberLink continues to grow — there have even been speculations about selling the GibberLink.com domain and webinars promising to teach “agent communication.”
The engineers emphasize that using GGWave can significantly reduce computational costs when two AI agents communicate with each other. While for now GibberLink remains just an interesting project, its future potential could fundamentally change the way AI communicates, especially as call center automation increases and AI agents capable of performing complex tasks for users are introduced.
