When I set out to build River, I wanted to solve one problem. Typing is a poor way to express the richness of thought. It's rate limited and clumsy. As AI changed the way we interact with computers so that the default interface became language, voice AI saw a resurgence. But this too has privacy issues and background noise increases error. You see tech bros in SF whispering into microphones to control their agents. It already feels like we are watching a relic, and we're just a few months in. Don't we all agree voice is an intermediary step?

What we really want is to have our thoughts directly translated into intent, feeling and text, without the lousy motor and keyboard pathway, or verbal expression which is rate limited and prone to deterioration by disease, age and disorder.

And none of the devices today give us that. Neuralink inserts metal into the brain and leaves damage that doesn't fully heal. Would you put that into someone you love unless they were terminally ill, and a little mobility or communication in the time they had left made it worth it? Sabi and AlterEgo are building non-invasive devices and have great marketing campaigns on X, but the reality is that they require laborious attempted speech, which is essentially speaking with your mouth closed. I'm not saying they have no value, but is this really thought? When you think, do you mouth the words? No.

Why we did not pursue non-invasive routes

I really want non-invasive to work. The idea of it is appealing, though I am convinced 95% of these ideas are tarpits, because you simply can not argue with the physics.

Before I go into why I ultimately abandoned non-invasive routes, let's explain what axioms I assess these ideas against.

First, you do not want to spend hours collecting training data. For example, David Gaddy put electrodes all over his face and spent 25 hours collecting data to get a 25% word error rate (Gaddy & Klein, 2020). This is just not viable for most people. The data scaling laws are real and can contribute to cross-subject transfer learning which reduces data requirements (Kaifosh & Reardon, 2025), but the axiom rests - minimal setup is desirable.

Second is the form factor. Again, using EMG as an example of what you don't want, which is electrodes all over your face. That's too invasive. You want something like a headphone or behind the ear. Ideally, you don't want anything visible!

Third is true thought decoding. Most non-invasive technologies, and even some invasive, do attempted speech. The best way I can describe it is close your mouth and whisper something moving your tongue, larynx and cheek muscles, just without making a sound. That's attempted speech. It doesn't feel natural, though can be learned like typing. But it's still not what we consider to be thought.

Fourth is accuracy and speed. I lump the two together, though they are distinct. The two metrics these are assessed by are WPM (words per minute) and WER (word error rate).

With that ruler in hand, look at the impressive companies in this space. Apple bought Q.ai (presumably for their patent portfolio), which uses infrared imaging of the cheek to decode some words, and I believe their intention is to integrate this into Apple AirPods. But its performance varies in different lighting conditions, and I don't know if a demo has been published.

Another one is in stealth (sorry, I am under NDA, but the company starts with the letters "Ta") and I think is the best thing we'll have in the non-invasive space. I am not at liberty to describe how it works, though I can say that it images movements in the vocal tract during attempted speech to produce incredibly accurate decoding with limited training data. I was conversational in 5 minutes, and immediately thought "dang, I might be working on the wrong thing". Though, it doesn't quite solve the problem, and if you have some disorder that affects your efferent signal chain such as aphasia, ALS, or brainstem stroke, this device, and all other non-invasive devices that read motor signals instead of the originating cortical-level signal, will be unsuitable for you. It's also not really thought.

Various modalities, tried and abandoned

I've looked at almost every non-invasive modality you can think of. We started with EMG and EEG. EMG reads muscle and wants electrodes all over your face, so it fails on form factor and on thought before you even begin. EEG is the one I had hope for, because the form factor could be attractive, but anatomy gets in the way. The skull is fifteen to eighty times more resistive than the brain tissue beneath it, so by the time a cortical signal reaches the scalp it has lost roughly an order of magnitude in amplitude and been smeared across centimetres of skin (McCann et al., 2019). You need something like 6 cm² of cortex firing in lockstep before signal even shows up at a scalp electrode (Cooper et al., 1965). Worst of all, the skull hits high frequencies hardest, and the 70 to 170 Hz high-gamma band that speech decoding actually depends on is so attenuated that scalp EEG in that range is mostly picking up muscle, not brain (Ball et al., 2009). That muscle signal is decodable, but it's not thought.

To give EMG and EEG a fair shot we spent a while making our own electrodes, soft and comfortable ones from PEDOT:PSS, a beloved conjugated polymer I have a lot of affection for, combined with polydopamine, which is the chemistry mussels use to cling to wet rocks, so it would stick to mildly wet surfaces like skin. They were lovely electrodes. The skull and face did not care.

Then I considered optical sensing. The brain consumes energy, and when a region becomes active it needs to be replenished with blood, so optical sensing measures the change in blood oxygen as this happens. I abandoned it for a few reasons. First, sealing the light is tricky. You can't have light leaking in between the transmitter and the skin, or leaking out, and for anyone with hair, and everyone who is mobile, that is a hard problem. There's also the fact that light is fundamentally an energy transfer mechanism, which almost no one talks about. Photons displace nitric oxide from cytochrome c oxidase, a protein in your mitochondria, which frees up oxygen to bind at the iron and copper centre of complex IV and act as the terminal acceptor for electrons. This drives the electron transport chain, which builds the proton gradient that synthesises ATP from ADP (tri from di) using spare phosphate. The net effect is whichever region is exposed to photons has more energy, and is a well-studied mechanism in neuroplasticity treatments. You are no longer just measuring the brain, you are changing it. That's invasive.

None of it cleared all four axioms. Even with data scaling laws on our side, we could not solve the problem we wanted to solve without going inside. So we went invasive.

The state of Invasive

The invasive game is one where you trade off risk and damage for signal quality.

Intracortical

Intracortical unequivocally has the best signal quality. You can detect activity of single neurons! That is just amazing, and what you can do with that is nothing short of miraculous, on both the read-out and write-in modalities.

However, the risks are severe. Firstly, there is gliosis. When you put foreign material into the body, the immune system attacks it. Microglia and astrocytes flock to the site, building up a protective layer around the shank. This degrades signal quality (Polikov et al., 2005).

Then there's micromotion. The brain is soft and never sits still. With every heartbeat it shifts two or three microns against a rigid shank, and every breath moves it another ten to thirty microns (Gilletti & Muthuswamy, 2006). That sounds tiny until you remember it happens around 100,000 times a day, every day, for the life of the implant. In rigid arrays like the Utah array and Paradromics' microwires, this constant grinding of a stiff probe against soft tissue keeps the wound inflamed, and it is part of why the signal fades over the years. Around each shank the cortex is left with a dead zone of neurons roughly 100 to 150 microns across (Biran et al., 2005), and adult cortex does not regrow them, so the damage is permanent.

Neuralink deserves a caveat. Its threads are not rigid shanks but ultra-thin flexible polymer, and that helps. In animals, soft, low-footprint probes that move with the brain rather than against it provoke far less scarring and preserve far more neurons than rigid ones (Zhou et al., 2017). But flexibility softens the problem, it does not erase it. You still have to push the thread in, and the insertion, plus the body's lifelong response to a foreign object, still leaves damage. Neuralink's thin-film threads also sit closer to the ordinary flexible probes that still scar than to the softest mesh that barely does, and there is no published human histology for them yet. So the fair read is that flexible threads reduce the damage, not that they remove it, and in a living person we do not yet know by how much.

In rats, intracortical probes measurably degrade fine-motor capability. In one study, implanted rats took up to 527% longer to complete a fine-motor task than controls, and the deficit was still there at 16 weeks (Goss-Varley et al., 2017). In humans, this is harder to measure, because the probes cover a smaller percentage of the brain than in rats, and because most patients so far have had serious paralysis already. How do you measure how much motor movement is lost in a paralysed patient?

Now, knowing what you know, whether that's a lot of neuroscience or just what you've read here, would you want intracortical electrodes from Neuralink, Paradromics, or anyone else put into yourself or someone you love? Only, I think, if they were in the last few years of an incurable disease like ALS. Or maybe that annoying aunt, the one who isn't a blood relative but one of your uncle's bad decisions. If you are the person who makes these kinds of decisions in your family, I imagine the answer is no.

Synchron and the endovascular paradox

Synchron is beautiful. It is the paragon of "work with evolution, not against it". There is no craniotomy, it is far less invasive than anything that opens the skull, and the scaffold is built to last a lifetime.

On the business side, there are already thousands of people in hospitals around the world called neurointerventionists who perform these procedures to unblock arteries during stroke. They are high pressure situations where the procedure must be decided for and performed quickly to save a life, and they are often highly effective with minimal complications.

Synchron achieved instant distribution with a very safe and effective device that works with the body. You can't fault it.

There are a few key limitations to Synchron though.

Firstly, it only has 16 channels, so people use it as a rudimentary motor control device (Oxley et al., 2016). That can be changed, I imagine, and it would be unsurprising if Synchron is soon at 512 or even 1024 channels.

The main problem with Synchron is anatomy. To decode speech you need access to the ventral premotor and sensorimotor cortex (area 6v) (Bouchard et al., 2013) and parts of Broca's area. You can also pull decodable signal from the supramarginal gyrus, as Wandelt did (Wandelt et al., 2024), and from the superior temporal gyrus (Chang et al., 2010). These all sit out on the lateral surface of the brain. The Stentrode works because the superior sagittal sinus is a large, safe vein running right over the midline motor strip, but the lateral speech areas have no equivalent highway. Their arterial supply, the middle cerebral artery and its branches, is far smaller than the sagittal sinus, and the superficial cortical veins that drain them are small, tortuous, and vary a lot between people. There is no big, reliable vessel sitting over Broca's area the way the sagittal sinus sits over motor cortex.

There are zero in-vivo arterial endovascular BCI studies, and for good reason. The cortical arteries feeding the speech areas are small, and a permanent electrode-covered stent parked in one risks occluding the vessel and the tiny perforators branching off it, starving the very cortex you are trying to read and causing an ischemic stroke. That is precisely what these devices are ordinarily trying to prevent.

So Synchron, awesome. But it is not able to solve the problem, probably no matter how much they innovate. I hesitate to say never, but it's improbable due to the risks of leaving permanent stents in arteries.

The surface

Decoding from the cortical surface has a long history of being safe and effective. Edward Chang at UCSF is the leading researcher in this field, having conducted multiple in-human trials. To my knowledge, the Chang lab is the only group with chronic, multi-year, large-vocabulary data in paralyzed humans through their BRAVO trials.

Moses et al. demonstrated the first closed-loop ECoG speech BCI with a 128-channel grid in an anarthric stroke survivor, achieving 25.6% WER on a 50-word vocabulary at 15.2 WPM (Moses et al., 2021).

Metzger et al. extended this with a 253-channel array in a second participant, reaching ~25% WER on a 1,024-word vocabulary at 78 WPM, with simultaneous text, synthesized voice, and avatar output from attempted silent speech (~21% pre-LM phone error) (Metzger et al., 2023).

Silva et al. showed the neural representations remained stable for years on the same array (Silva et al., 2024), and Littlejohn et al. added streaming brain-to-voice synthesis at 46.5 WPM with ~30.7% WER (Littlejohn et al., 2025).

All of Chang's research was conducted with PDMS (polydimethylsiloxane, aka silicone) arrays. But silicone is incompatible with standard high density designs because it interacts poorly with photolithography, leading to wider pitch (spacing between electrode contacts). In Chang's example, the pitch was 4 mm.

Muller et al. measured the high-gamma spatial decorrelation length in human cortex at ~1.4 mm, meaning the 4 mm pitch used by Chang undersamples by roughly 3× per dimension (Muller et al., 2016). Duraivel et al. confirmed this in awake patients, finding 35% better decoding and 48% higher SNR at sub-2mm pitch, and pegged under 2 mm as the threshold for reaching 95% of full-array performance (Duraivel et al., 2023). Hettick et al. pushed to a 1,024-channel array at 400 µm pitch and saw decoding accuracy keep climbing with channel count, with no sign of saturation (Hettick & others, 2025). The takeaway is simple: tighter pitch and more channels keep paying off, they don't plateau. I've spared you the full spatial-frequency analysis here. If you want it, ask for River's data room or get in touch.

So we know that even with a coarse pitch, we can get to 78 words per minute and a 25% word error rate on a large vocabulary, and that tightening the pitch increases spatial resolution and decoding accuracy significantly. Great.

Now this brings us to the key problem, and it is a materials tradeoff. PDMS lasts a long time in the body, but as we saw it can't do fine pitch. Polyimide is the opposite. It's compatible with standard lithography, so you can get very fine pitches down to single-digit microns, but it doesn't last, and that is inherent to its chemistry.

BPDA/PDA polyimide (biphenyl dianhydride / p-phenylenediamine) is formed by thermal imidization of a polyamic acid (PAA) precursor. During curing, two water molecules are expelled per repeat unit as the imide rings (–CO–N–CO–) close.

In a chronic physiological environment, this condensation reaction is reversible. Water molecules present in interstitial fluid attack the electrophilic carbonyl carbons of the imide ring, cleaving the C–N bond and reopening the ring structure. This hydrolysis swells the polymer matrix, generates internal stress at buried interfaces, and weakens adhesion between encapsulation layers and underlying metal traces.

BPDA/PDA is among the better-performing polyimides in this regard. Bulk water absorption at saturation is approximately 1 wt%, lower than most alternatives in the polyimide family. However, saturation is reached within 3 days of saline exposure at body temperature. Moisture absorption in polyimide films follows Fickian diffusion kinetics and scales nearly linearly with relative humidity across the operating range, meaning there is no threshold below which absorption ceases. Once implanted, a polyimide device is absorbing from day one.

This relationship between water absorption and compatibility with high-density manufacturing processes is one of the major reasons that chronic, high-density surface arrays have been the exception, rather than the norm.

River solves the material science problem. We use a new material (that we patented, but I am not ready to talk about yet) that has very tightly packed molecular bonds, and thus, a very low water absorption rate, even lower than parylene C (0.06 wt%), which is what the field generally uses to keep moisture out. It's compatible with high-density manufacturing processes, is generally easy to obtain, and quite forgiving to work with.

This is why we've started with surface decoding. The science is proven, and the last piece of the puzzle is materials science, which is now a solved problem. With the exception of stimulation, surface decoding already gives you conversational-rate speech. Intracortical looks better on paper, but nobody has ever placed both in the same person on the same task, so the order-of-magnitude gap everyone assumes is real has never actually been measured. And surface gets there without the gliosis and chronic inflammation that make Neuralink and its intracortical cousins so repellent.

Come back to the four axioms. Surface decoding is the first modality that clears all of them. Setup is a one-time procedure, not hours of daily calibration. The form factor vanishes under the skin, with nothing visible. The accuracy and speed are already there, 78 words per minute at a 25% word error rate, and climbing as the pitch tightens. And it reads from the cortex itself, where thought originates, rather than the muscle downstream, which is the only place true thought can ever be read from.

Surface is the entry point, not the destination. The natural path from here is to keep working outwards, to epidural, giving up a little signal for even less risk, which I believe we can make up for with a large data bank and improvements in AI models beyond the GRU/CTC paradigm which dominates neural decoding.

Epidural

Decoding from the cortical surface is safer than intracortical, but it is not risk free. We've evolved to have thick skulls and dura mater for a reason. I imagine it has something to do with the fact that organisms with little cerebral protection probably got infections or damaged their brains more easily, leading to lower survival rates. Does our evolutionary trajectory involve cutting into that now? Probably not.

Epidural decoding can be done well. Acclaimed neuroscientist Sam John, who led most of Synchron's preclinical work and is now working with River on our preclinical studies, published a paper in 2018 with Tom Oxley showing the signal quality achievable at each level of invasion, finding endovascular, subdural and epidural recordings comparable (John et al., 2018). Epidural decoding is entirely capable of capturing the high gamma signal band (70-170 Hz) which is so critical to speech decoding, but only IF the contacts are large (Bundy et al., 2014). This is known as macro-ECoG.

There are companies working on this modality, the most well-known of which are NEO in China (Hong & others, 2024) and WIMAGINE in France (Sauter-Starace et al., 2019) (Larzabal et al., 2021). They are incredibly elegant devices which remove a small bit of skull, but leave the dura and its CSF-sealing properties intact, and replace that section of skull with a titanium plate that has electrodes on one side. They work effectively and, on information, are safe enough to be considered for less-than-life-threatening conditions.

Researchers are only beginning to investigate speech decoding with the epidural modality, with the first preprint appearing in late December 2025. Take Sabi, who promise to decode your thoughts and attempted speech from a beanie. It is unclear how. They have published nothing, no one has shown EEG doing this, and I am not aware of any published work using the hundred-thousand-sensor arrays they describe. But even granting their bet, the logic runs in our favour: if data scaling laws could ever pull reliable decoding from the skin, through the skull, then they should deliver a far better experience from epidural, where the signal is not fighting the skull at all.

This post explains why our thesis involves starting on the cortical surface and working outwards to capture true thought, not attempted speech. And over time, we'll find a way to "write in" to consciousness, ethically and desirably, so that when you can't find the words, you won't need them.

We've all had a sibling or a close friend where the bond is so tight you feel like you already know what they're thinking. I don't think it's far-fetched to say that one day we'll communicate like that for real, thought to thought, with nothing in between. That is the world we are building toward.

Then we can really say, "It's the thought that counts."

References

1. Gaddy, D., & Klein, D. (2020). Digital Voicing of Silent Speech. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 5521-5530. doi:10.18653/v1/2020.emnlp-main.445
2. Kaifosh, P., & Reardon, T. (2025). A generic non-invasive neuromotor interface for human-computer interaction. Nature, 645, 702-711. doi:10.1038/s41586-025-09255-w
3. McCann, H., Pisano, G., & Beltrachini, L. (2019). Variation in reported human head tissue electrical conductivity values. Brain Topography, 32(5), 825-858. doi:10.1007/s10548-019-00710-2
4. Cooper, R., Winter, A., Crow, H., & Walter, W. (1965). Comparison of subcortical, cortical and scalp activity using chronically indwelling electrodes in man. Electroencephalography and Clinical Neurophysiology, 18, 217-228.
5. Ball, T., Kern, M., Mutschler, I., Aertsen, A., & Schulze-Bonhage, A. (2009). Signal quality of simultaneously recorded invasive and non-invasive EEG. NeuroImage, 46(3), 708-716. doi:10.1016/j.neuroimage.2009.02.028
6. Polikov, V., Tresco, P., & Reichert, W. (2005). Response of brain tissue to chronically implanted neural electrodes. Journal of Neuroscience Methods, 148(1), 1-18. doi:10.1016/j.jneumeth.2005.08.015
7. Gilletti, A., & Muthuswamy, J. (2006). Brain micromotion around implants in the rodent somatosensory cortex. Journal of Neural Engineering, 3(3), 189-195. doi:10.1088/1741-2560/3/3/001
8. Biran, R., Martin, D., & Tresco, P. (2005). Neuronal cell loss accompanies the brain tissue response to chronically implanted silicon microelectrode arrays. Experimental Neurology, 195(1), 115-126. doi:10.1016/j.expneurol.2005.04.020
9. Zhou, T., Hong, G., Fu, T., Yang, X., Schuhmann, T., Viveros, R., & Lieber, C. (2017). Syringe-injectable mesh electronics integrate seamlessly with minimal chronic immune response in the brain. Proceedings of the National Academy of Sciences, 114(23), 5894-5899. doi:10.1073/pnas.1705509114
10. Goss-Varley, M., Dona, K., McMahon, J., Shoffstall, A., Ereifej, E., Lindner, S., & Capadona, J. (2017). Microelectrode implantation in motor cortex causes fine motor deficit: Implications on potential considerations to Brain Computer Interfacing and Human Augmentation. Scientific Reports, 7, 15613. doi:10.1038/s41598-017-15623-y
11. Oxley, T., et al. (2016). Minimally invasive endovascular stent-electrode array for high-fidelity, chronic recordings of cortical neural activity. Nature Biotechnology, 34(3), 320-327. doi:10.1038/nbt.3428
12. Bouchard, K., Mesgarani, N., Johnson, K., & Chang, E. (2013). Functional organization of human sensorimotor cortex for speech articulation. Nature, 495(7441), 327-332. doi:10.1038/nature11911
13. Wandelt, S., Bj\aanes, D., Pejsa, K., Lee, B., Liu, C., & Andersen, R. (2024). Representation of internal speech by single neurons in human supramarginal gyrus. Nature Human Behaviour, 8(6), 1136-1149. doi:10.1038/s41562-024-01867-y
14. Chang, E., Rieger, J., Johnson, K., Berger, M., Barbaro, N., & Knight, R. (2010). Categorical speech representation in human superior temporal gyrus. Nature Neuroscience, 13(11), 1428-1432. doi:10.1038/nn.2641
15. Moses, D., et al. (2021). Neuroprosthesis for decoding speech in a paralyzed person with anarthria. New England Journal of Medicine, 385(3), 217-227. doi:10.1056/NEJMoa2027540
16. Metzger, S., et al. (2023). A high-performance neuroprosthesis for speech decoding and avatar control. Nature, 620, 1037-1046. doi:10.1038/s41586-023-06443-4
17. Silva, A., et al. (2024). A bilingual speech neuroprosthesis driven by cortical articulatory representations shared between languages. Nature Biomedical Engineering, 8, 977-991. doi:10.1038/s41551-024-01207-5
18. Littlejohn, K., et al. (2025). A streaming brain-to-voice neuroprosthesis to restore naturalistic communication. Nature Neuroscience. doi:10.1038/s41593-025-01905-6
19. Muller, L., Hamilton, L., Edwards, E., Bouchard, K., & Chang, E. (2016). Spatial resolution dependence on spectral frequency in human speech cortex electrocorticography. Journal of Neural Engineering, 13(5), 056013. [Link]
20. Duraivel, S., et al. (2023). High-resolution neural recordings improve the accuracy of speech decoding. Nature Communications, 14, 6938. doi:10.1038/s41467-023-42555-1
21. Hettick, M., & others, . (2025). Minimally invasive implantation of scalable high-density cortical microelectrode arrays for multimodal neural decoding and stimulation. Nature Biomedical Engineering. doi:10.1038/s41551-025-01501-w
22. John, S., et al. (2018). Signal quality of simultaneously recorded endovascular, subdural and epidural signals are comparable. Scientific Reports, 8, 8427. doi:10.1038/s41598-018-26457-7
23. Bundy, D., et al. (2014). Characterization of the effects of the human dura on macro- and micro-electrocorticographic recordings. Journal of Neural Engineering, 11(1), 016006. doi:10.1088/1741-2560/11/1/016006
24. Hong, B., & others, . (2024). Reclaiming hand functions after complete spinal cord injury with epidural brain-computer interface. medRxiv. doi:10.1101/2024.09.05.24313041
25. Sauter-Starace, F., et al. (2019). Long-term sheep implantation of WIMAGINE, a wireless 64-channel electrocorticogram recorder. Frontiers in Neuroscience, 13, 847. doi:10.3389/fnins.2019.00847
26. Larzabal, C., et al. (2021). Long-term stability of the chronic epidural wireless recorder WIMAGINE in tetraplegic patients. Journal of Neural Engineering, 18(5), 056026. doi:10.1088/1741-2552/ac2003