The read side (the mature half)
The sender's half is ordinary decoding, which you already know from Volumes I and II. Non-invasively, a BBI typically taps one of a few reliable EEG paradigms: motor imagery read as mu-rhythm desynchronization, SSVEP frequency-tagging, or P300 / speller selection. Each maps brain A onto a small alphabet of discrete symbols reliably enough to act on.
Invasively, spikes carry far more, but that requires an implant in the sender — which for a communication system, as opposed to a clinical one, is a very high bar. So most BBIs are read-limited only in the trivial sense that they choose coarse, robust EEG features. The real limit is on the other side.
The write side is the bottleneck
Writing into a brain non-invasively is crude. TMS over visual cortex evokes a phosphene — a formless flash — worth roughly one bit per pulse; over motor cortex it evokes an involuntary twitch. The TMS phosphene write is the workhorse of human BBIs precisely because it is the only non-invasive way to reliably put a detectable, self-attributable event into another person's experience.
The emerging write modalities push on this. Transcranial focused ultrasound and temporal interference stimulation can reach deeper with a tighter focus than TMS, but their information payload is still tiny and their perceptual effects poorly characterized. At the invasive extreme, intracortical microstimulation — the same channel used for somatosensory touch restoration — is higher-resolution, but it needs an implant in the receiver.
It's a computer-mediated cascade
Put the halves together: brain A goes to a decoder, which emits bits over a digital link, which drive a stimulator, which writes into brain B. The encoder-decoder brain pair never touch — a classical channel carries the bits, which is exactly why the honest label is brain-to-computer-to-brain. Each half is a noisy channel, and we can model the simplest case as a binary symmetric channel.
Capacity of one binary symmetric half-channel with per-symbol error p. With today's coarse read and write, p is far from zero, so a single pulse carries only a fraction of a bit.
Who is acting? Agency at the receiver
A BBI raises a question no ordinary BCI does: at the receiver, who acts? In the reflexive setups, a TMS pulse over motor cortex produced the receiver's hand movement directly, bypassing conscious choice. In the phosphene setups, the receiver consciously perceived a flash and then chose to act. That gap between an evoked response and an interpreted cue is the seed of the ethics in guide 5 — it is the difference between a message and a command. See receiver agency and the sense of agency.