GibberLink

This video - entitled Two AI agents on a phone call realize they’re both AI and switch to a superior audio signal ggwave - is a proof of concept, not a thing that has happened yet. When the two agents acknowledge that they are both robots they switch to GibberLink Mode, a way for the two bots to converse using a library called ggwave that is designed for "data over sound." The high level summary is:

This library is used only to generate and analyze the RAW waveforms that are played and captured from your audio devices (speakers, microphones, etc.). You are free to use any audio backend (e.g. PulseAudio, ALSA, etc.) as long as you provide callbacks for queuing and dequeuing audio samples.

That doesn't sound like much but it's doing all the heavy lifting to encode the audio with the data and decode it on the other side. There's plenty of demo videos and demo apps you really should play with.

There are certainly some practical considerations to keep in mind with an audio based communication tool - namely, this isn't going to work in a crowded space - but it sounds fantastic. I love a scenario where sci-fi informs reality and all of this feels lifted directly from a Star Trek script. I'm not saying we're ever going to hear it out in the wild but I sure do love seeing it as a fully working demo.