The depth problem
Start with the wall from guide 4. Because scattering caps non-invasive optical access to roughly the cortical surface, deep optical read/write in a large brain requires implanted optics: GRIN-lens relays, implantable microLED / optrode arrays, or fibres that carry light and collect fluorescence past the scattering. That reintroduces exactly the invasiveness optics promised to avoid, and adds a new power/heat budget on the implant.
Higher nonlinear order n (three-photon) and longer wavelength lengthen the scattering length, buying depth — but in millimetres, not centimetres. Deep and whole-brain optical access remains genuinely open.
The genetics problem
The deepest barrier is not optics at all — it is genetics. Every optical read or write requires a transgene, delivered (typically by AAV), expressed at the right level in the right cells, and kept stable and safe for years. Microbial opsins are foreign proteins and can be immunogenic; expression can drift or silence; dosing, reversibility, and off-target expression are all open. This — not the physics of light — is why electrical interfaces remain the clinical default.
What has actually reached humans
Keep an honest ledger. The clearest first-in-human optogenetic result is in the eye: retinal optogenetic therapy has restored partial, task-level vision in a blind patient with advanced retinitis pigmentosa, using a red-shifted opsin plus light-amplifying goggles. That is a genuine milestone — but partial, single-report-scale, and in the most optically accessible tissue in the body. The optical cochlear implant promises finer frequency resolution than electrical stimulation, because light focuses better than current (electrical vs optogenetic restoration), yet it remains preclinical. And near-infrared and upconversion optogenetics, which aim to sidestep the depth wall, are early-stage animal work.
The pattern is telling: optical neuroprostheses reach humans first where the target tissue is thin, superficial, and already optically privileged — the retina, the cochlea. The general case, deep cortical read/write in a human, is not close.
Where it is going
Read this honestly and the near-term value of optical BCI is scientific before it is prosthetic. All-optical, cell-type-specific, closed-loop control is an unmatched instrument for dissecting the neural code — reading and writing defined circuits to learn how they compute. Two walls still stand: depth (scattering) and the biomimetic-write problem (what to write). The plausible medium-term futures are hybrid — optics combined with electrical and other modalities, bidirectional by design — and human therapy advancing selectively where geometry cooperates. Separating this demonstrated trajectory from the aspirational one is exactly the discipline clinical translation demands.