Meta releases version two of its brain-computer interface that can turn thoughts into keypresses — non-invasive magnetoencephalography scanner can measure changes in brain activity

At a glance:\n- Meta launches second‑generation Brain2Qwerty, a non‑invasive BCI that turns thoughts into keypresses using a magnetoencephalography scanner.\n- Accuracy jumps from 40% average (previous version) to 61%, with a top performer reaching 78% word accuracy after training on ten times more data.\n- The system remains clinically unready; MEG hardware is bulky and accuracy still feels hit‑and‑miss for everyday use.\n\n## Meta’s second‑generation Brain2Qwerty\nMeta released the second version of its Brain2Qwerty BCI, building on last year’s proof‑of‑concept. The system remains non‑invasive, using a magnetoencephalography (MEG) scanner to capture magnetic field fluctuations generated by neuronal activity. The captured signals are then mapped to keypresses on a virtual keyboard, allowing users to type purely by thought.\n\nThe new release incorporates a training regimen that leverages ten times more data per test subject than the first iteration. This expansion in dataset size has been a key driver behind the marked improvement in decoding performance, as Meta reports.\n\n## How the MEG‑based BCI works\nMEG sensors detect the faint magnetic fields produced by synchronous firing of cortical neurons. Unlike electroencephalography (EEG), which measures electrical potentials on the scalp, MEG captures these fields with high temporal resolution and is less susceptible to skull‑induced attenuation. The scanner in Brain2Qwerty is positioned around the participant’s head, allowing the device to read activity patterns without penetrating the skull.\n\nOnce the magnetic signatures are recorded, machine‑learning algorithms correlate specific patterns with intended keypresses. The virtual keyboard interface translates the decoded commands into keystrokes, enabling the user to compose text or control computer functions. Because the system is non‑invasive, it sidesteps the surgical risks associated with implanted electrodes.\n\n## Performance gains and remaining hurdles\nMeta’s latest metrics show an average accuracy of 61% across subjects, a significant rise from the 40% average achieved by the first version. The best participant in the new study reached 78% word‑level accuracy, compared to a 48% peak in the earlier release. Despite these gains, the accuracy still exhibits a hit‑and‑miss quality that would make real‑world conversation or complex typing challenging.\n\nAnother obstacle is the physical bulk of current MEG hardware. Existing sensors are larger than the user and the chair they occupy, limiting portability and clinical practicality. Meta acknowledges that ongoing sensor‑design research could yield smaller, more manageable devices, but such improvements remain in the future.\n\n## Comparison with invasive BCIs\nInvasive systems like Elon Musk’s Neuralink require surgical implantation of electrodes to record neural signals directly from the cortex. While these devices can achieve higher fidelity, they carry surgical risks and long‑term biocompatibility concerns. Meta’s Brain2Qwerty offers a risk‑free alternative, albeit with lower accuracy and larger equipment.\n\nOther non‑invasive efforts include a Georgia Tech team’s miniature BCI that can be slid under the scalp and a startup founded by Valve co‑founder Gabe Newell that aims for battery‑free operation. Meta’s approach sits alongside these innovations, contributing to a broader push toward safer, more accessible brain‑computer interfaces.\n\n## Future prospects and industry context\nMeta plans to continue expanding the training dataset and refining its decoding algorithms, hoping to push accuracy into a clinically viable range. The company also monitors advances in MEG sensor miniaturization, which could transform the system from a laboratory prototype into a bedside or home device.\n\nWhile the current performance falls short of enabling everyday tasks like webcam control or gaming, the technology represents a meaningful step in neuroprosthetics. If future iterations achieve higher accuracy and reduced hardware size, Brain2Qwerty could help patients with mobility impairments regain communication and autonomy.\n\n## Conclusion\nMeta’s second‑generation Brain2Qwerty demonstrates that non‑invasive brain‑computer interfaces can reach respectable decoding accuracy without surgery. The jump from 40% to 61% average accuracy marks progress, yet the remaining challenges—sensor bulk, accuracy variability, and clinical readiness—highlight the distance still ahead.\n\nWhile the neurotechnology ecosystem matures, competition among non‑invasive solutions will likely accelerate. Meta’s commitment to data‑driven improvements and hardware research positions it as a key player in shaping the next wave of brain‑computer interaction.