Kernel (neurotechnology company)

Kernel is an American neurotechnology company, founded in 2016 by the payments entrepreneur Bryan Johnson, that builds non-invasive headsets for measuring brain activity. It began as a plan for an implanted Memory prosthesis and dropped that goal within about six months. Since 2020 its main product has been Kernel Flow, a helmet that uses time-domain functional near-infrared spectroscopy (TD-fNIRS) to track blood-oxygen changes in the outer cortex. As of October 2026 Kernel sells Flow to researchers and offers "brain age" scans through a few longevity clinics; no Kernel product is cleared by the US Food and Drug Administration to diagnose or treat disease.

§Overview

Kernel is often grouped with Neuralink and Synchron as a Brain–computer interface company, and Johnson calls Flow a "brain interface". The label hides a basic difference: Kernel's current devices are worn, not implanted, and only record. They do not stimulate the brain, drive a cursor or decode the content of thought. Flow measures changes in oxygenated and deoxygenated haemoglobin, a signal that lags neural activity by seconds and comes only from the outer few centimetres of cortex.

The stated aim is to make brain function as routinely measurable as heart rate or blood chemistry. A 2025 company newsletter says the team held itself to "no gimmicks, no toys, no sci-fi fantasies",¹ a long way from the 2016 launch, when Johnson likened programming yeast to programming the brain's "neural code".² Since 2025 Kernel has sold wellness scans in the same consumer-longevity market as Consumer blood testing and Wearable health sensors.

Measurement, not mind-reading

Headlines have repeatedly called Kernel's helmets "mind-reading". A TD-fNIRS headset estimates how much oxygenated blood reaches patches of cortex. It can show that a region worked harder; it cannot say what a person was thinking, and it does not reach the hippocampus or other deep structures.

§History

Kernel launched in October 2016, when Johnson announced that he would put $100 million of his own money into it; TechCrunch reported that he had turned down outside investors. The first product was to be a neuroprosthesis for cognitive decline such as Alzheimer's disease, built on the hippocampal memory research of Theodore Berger of the University of Southern California, the company's chief science officer. Johnson gave a timeline of seven to ten years.²

Berger had left by early 2017 and memory implants were dropped from the near-term plans. "The basic reason is it was going to take too long," Berger told MIT Technology Review; Johnson called the work "really interesting, but not an entry point". In February 2017 Kernel acquired Kendall Research Systems, a Cambridge spin-out of Edward Boyden's MIT lab that made neural recording equipment, and re-aimed at a general platform for measuring and stimulating many neurons.³ By 2020 that had narrowed to non-invasive measurement.

1. 2016
   Launch
   Johnson announces a $100 million personal commitment to a company built around Theodore Berger's hippocampal memory-prosthesis research.
2. 2017
   Implants dropped
   Berger leaves and Kernel acquires Kendall Research Systems; memory implants are removed from its near-term plans.
3. 2020
   Flow and Flux
   Kernel unveils an optical headset and a magnetoencephalography headset, offers them as Neuroscience as a Service, and raises $53 million from outside investors.
4. 2022
   Flow described
   Kernel engineers publish a peer-reviewed description of the first-generation Flow system.
5. 2023
   New chief executive
   Johnson hands the CEO role to chief technology officer Ryan Field and stays on as chairman.
6. 2025
   Clinic scans
   A mild cognitive impairment study is published, and brain-age and cognition scans launch in wellness clinics.
7. 2026
   Retail scans
   Flow is offered at about half a dozen US longevity clinics, and a small depression study is terminated for low enrollment.

In 2023 Johnson handed the chief executive role to Ryan Field, until then chief technology officer, and stayed on as chairman, saying the company would focus next on depression and cognitive decline.⁴ Now known mainly for his anti-aging regimen, Johnson frames Flow in those terms: the brain was the organ he could not measure.⁵

§Technology

§Kernel Flow

Time-domain fNIRS fires very short laser pulses into the scalp and records when individual photons return. Later photons have, on average, travelled deeper, so the timing separates light that reached cortex from light scattered in skin and skull better than the continuous-wave systems most fNIRS research uses. The technique had been confined to bulky laboratory instruments. Kernel miniaturized the laser drivers, integrated circuits and detectors into modules that tile the head. The first-generation system had 52 modules and weighed 2.05 kg; its validation was a set of standard phantom protocols plus a finger-tapping task in two participants.⁶

The paper's headline 200 Hz is each detector's histogram rate; because modules fire in turn to avoid crosstalk, the whole head is sampled at about 7 Hz, still far faster than the haemodynamic response.⁶ The second-generation Flow2 has 40 optical modules plus a few dry EEG electrodes, weighs 5.5 pounds and was listed at $117,200 in 2026.⁵

§Kernel Flux

Flux, unveiled with Flow in 2020, was a magnetoencephalography helmet whose optically pumped magnetometers detect the faint magnetic fields of synchronized neural activity.⁷ A 2021 SPIE conference paper describes it as a whole-head system; as of October 2026 Kernel's product listings offer only Flow-based systems.

§Research programme

Nearly all published work with Flow comes from Kernel's own scientists. A ketamine study funded by Cybin, a developer of drugs for Psychedelic therapy, gave subanaesthetic ketamine and saline placebo to 15 healthy volunteers, single-blind and non-randomized, and reported reduced low-frequency signal fluctuations across the brain, lower prefrontal connectivity and "preliminary evidence" of links to mystical experience and mood.⁸ Kernel has also published an alcohol study and a reliability study.

The clinical programme has two arms. In the IMPACT study, 50 people with mild cognitive impairment and 51 age-matched controls did verbal-fluency and working-memory tasks while wearing Flow. A classifier given self-report and task behaviour alone reached an area under the curve of 0.79. Adding the neural measures raised it to 0.92. The authors, funded by and mostly employed by Kernel, call testing the model on a separate cohort "a critical next step".⁹ The work bears on screening for Alzheimer's disease but has not been tested in that role.

The second arm, the PREDICT study, scans depressed patients before and after treatments including transcranial magnetic stimulation, a form of Non-invasive neuromodulation, to learn whether a baseline scan predicts response. Kernel said in late 2025 that it expected a full readout by the end of that year.¹ As of October 2026 the registry lists the study as still enrolling, with no results posted. A companion study of accelerated TMS was terminated in 2026 for low enrollment after 11 participants.

The consumer products came last. In 2025 Kernel launched a Brain Age Tracker, a short passive scan summarized as a single age, and a 30-minute Cognition Tracker for longevity and wellness clinics. Five clinics were offering scans that autumn.¹ The score has spread into other companies' research: Minicircle's pilot of an injectable Klotho gene therapy lists Flow's "Estimated Brain Age", the output of proprietary models, as a trial outcome. None of the peer-reviewed papers Kernel lists validates the score, which leaves it among the Aging biomarkers and Biological age measures whose relation to health outcomes is unsettled.

§Funding

Johnson's 2016 figure of $100 million was a commitment, and it is the number most often repeated; the sums reported as actually invested are smaller. When Kernel took its first outside money, a $53 million Series C led by General Catalyst with Khosla Ventures and others in July 2020, Johnson had invested $54 million.⁷ A 2025 company newsletter says he founded Kernel with $50 million,¹ and Business Insider reported in 2026 that he had put in more than $60 million of over $150 million invested overall.⁵

The 2020 business model was "Neuroscience as a Service": paying clients used Kernel's headsets on demand, sometimes remotely, rather than buying them.⁷ It has since shifted toward hardware sales to laboratories, data subscriptions, funded research partnerships and clinic scans. Kernel said in 2026 that more than 30 commercial developers and academic labs had bought its systems.⁵

§Reception

One scientist quoted in 2017 called Johnson's original business plan "metaphysical". Konrad Kording, a Northwestern neuroscientist who had advised him, warned that "we are far away from meaningfully understanding the brain".³

Independent researchers have been kinder about the hardware than about the claims made for it. David Boas of Boston University, a long-time fNIRS researcher, called it "just beautiful" and bought modules for his own lab, but said "right now there's just not enough data" and estimated that a decade or more would pass before the measurements yield actionable brain-health guidance. Theodore Huppert of Stony Brook University noted that middle-aged brains look broadly alike on such scans, limiting what one brain-age reading can tell an individual.⁵ A University of Toronto group testing Flow on three people over 20 sessions found that it reproduced resting-state patterns known from fMRI, with less spatial detail.¹⁰

Company-generated evidence

Every clinical claim for Flow rests on studies funded by Kernel and written mainly by its employees. The headline cognitive-impairment result comes from one cross-validated cohort of 101 people with no external test set, and none of Kernel's published papers validates the brain-age score sold in clinics. The open question is not whether the helmet records a real haemodynamic signal but whether the classifiers hold up in independent cohorts.

Privacy concerns have followed the move into consumer scans. Huppert warned that companies could exploit brain signals commercially without fully understanding them, and a JLL executive in the same article stressed Mental privacy at work.⁵ Flow's resolution is far below what the Neurorights debate imagines, but large datasets of brain responses to media are that debate's raw material, and Kernel has been inviting volunteers to watch television in the headset to build training data.

§Outlook

Kernel's future depends on whether a measurement company can become a medical one. Its planned clinical "Markers", for cognitive impairment and depression-treatment response, would need prospective validation and regulatory clearance before a doctor could use them, and the PREDICT readout promised for the end of 2025 is still unpublished. Meanwhile the wellness scans bring revenue and data without the evidentiary bar a diagnostic must meet, the trade-off made across consumer longevity testing.

Johnson's 2016 talk of programming neural code belonged to the Human–AI merger and Intelligence amplification vision that still drives much invasive-BCI investment. Kernel now sells a careful optical instrument whose best-documented uses are pharmacology studies and group-level classification. The strongest objection to its consumer pitch is Boas's decade-long estimate, from a researcher who admires the hardware.

§See also

• Bryan Johnson
• Brain–computer interface
• Neural decoding
• Non-invasive neuromodulation
• Mental privacy
• Neurorights
• Memory prosthesis
• Quantified self

§References