The lab-to-life gap
Almost everything in Volumes I and II was measured under laboratory conditions: a seated, still, cooperative participant; a quiet electromagnetic environment; gelled electrodes; a fresh calibration; a short session. Those conditions are not cheating — they isolate the phenomenon — but they are also nothing like a person using a device at home for a year. Ecological validity asks how well a result obtained in the lab predicts behaviour in the messy world, and out-of-the-lab robustness is the engineering property that closes the gap.
Three arenas of real-world BCI
Real-world BCI is not one thing. It spans an invasiveness spectrum. At one end are implanted systems used at home: BrainGate participants have operated intracortical systems in extended, less-supervised sessions, and the endovascular Stentrode is delivered through a blood vessel with no craniotomy, aiming squarely at everyday independent use. In the middle sit wearable non-invasive systems — dry-electrode EEG, ear-EEG, wearable fNIRS, optically-pumped MEG. At the far end is the consumer market of consumer neurotechnology sold directly to the public.
These arenas trade the same currency: signal quality against invasiveness and everyday practicality. This is the invasiveness-versus-signal tradeoff seen from the user's side. An intracortical array gives the richest signal but needs surgery and upkeep; a minimally-invasive endovascular device gives less but no craniotomy; a dry EEG headset gives least but you can put it on yourself. No single point on this curve is 'the' real-world BCI — the right choice depends on who the user is and what they need.
The three enemies
Leaving the shielded room releases three enemies that a lab keeps caged. First, everyday non-stationarity: signals drift with electrode settling, arousal, fatigue, caffeine, time of day and posture, so a decoder trained this morning is subtly wrong tonight. Second, artifact-heavy signals: real users blink, chew, walk and sweat, and motion and physiological artefacts can be larger than the brain signal itself. Third, adherence: a device only helps if people keep wearing and using it, which makes long-term adherence and human factors first-class engineering problems, not afterthoughts.
How to read this track
The rest of the track follows the pipeline of a device that must survive the world. Guide 2 surveys the wearable and minimally-invasive hardware — dry EEG, ear-EEG, fNIRS, wearable MEG, and where mobile brain-body imaging fits. Guide 3 is about fighting the noise: artifacts and non-stationarity, and the alignment methods that tame them. Guide 4 is honest evaluation: evidence standards, chance levels, information rate, and adherence. Guide 5 confronts the field's open problems: overclaiming, reproducibility, and neural-data privacy in a consumer world.