Connexus

In June, neurosurgeons at the University of Michigan drove 421 platinum-iridium microwires into the speech-motor cortex of a Michigan woman who has lost the ability to speak clearly to motor neuron disease. The wires, each about a millimeter and a half long and thinner than a hair, went in all at once in under a second, fired by a spring-loaded inserter — fast on purpose, because a quick clean entry does less damage to the tissue the array has to live inside for years.

The wires read individual neurons — the resolution everyone agrees you want; whether more channels actually buys more fluent speech is the open bet, not a settled result.

This is Paradromics' first human implant, and the pitch is density. Where existing implants read crowds of neurons, Connexus listens to individual cells, 421 channels of them — the most ever placed in a person — with the whole system buried under the skin: a module smaller than a dime in the skull, a transceiver in the chest beaming the signal out wirelessly, no wires through the scalp. The competing approaches split on how invasive to be; Paradromics chose maximum resolution and is betting that raw bandwidth is what eventually turns neural static into fluent sentences.

The thing not to lose: nothing has been decoded yet. This is an early-feasibility study cleared by the FDA for safety — the device is in, and the woman will be followed for six years. Restoring her speech is the goal the trial is built to test, not a result it has shown. The brain-computer field is now a commercialization race among a half-dozen companies, each implanting its first patients and each, for now, mostly proving that the hardware can sit in a human skull without harm.