health
Healthy people use elective implants that materially enhance cognition by 2126.
Last revised Sep 13, 2026.
Current estimate
probability this has happened by each date
The full test below is met by 31 December 2035. New this cycle; this subject had no horizon grid before. Near zero, and the reason is a count rather than a judgment: the number of healthy people with no medical indication who have ever received an intracranial implant is zero, so the claim needs ten thousand of them in nine years starting from none. No regulatory channel exists in the EU, the United States or China through which such a device could be placed on a market, and the nearest cranial analogue with a severe indication and full reimbursement reached only eight thousand patients in twelve years.
The full test below is met by 31 December 2050. New this cycle. Twenty-four years requires the whole chain in sequence: a first healthy recipient, a regulatory channel built somewhere that admits an enhancement indication, then diffusion to ten thousand. The fastest implant diffusion on record, cochlear implants, took between four and eleven years from approval to ten thousand, which means approval would have to land around 2040 and therefore a submission in the early 2030s, in a field where every one of eight leading developers is currently filed for restoration only. The mass that remains sits almost entirely in a permissive jurisdiction shortcutting the sequence rather than in the sequence completing on time.
The full test below is met by 31 December 2100. New this cycle, and this is where nearly all of the probability accrues. Seventy-four years is long enough for the capability term to resolve, for one jurisdiction among many to build a channel, and for diffusion to a bar of ten thousand that every successful implant analogue has cleared by two orders of magnitude. The 2126 figure sits only three points above this one because the residual twenty-six years adds little to a process that will already have decided by then.
At least ten thousand people with no qualifying medical indication, in at least one jurisdiction, carry an implant that raises memory, reasoning, or learning above each recipient's own pre-implant baseline, sustained for at least twelve months and measured on a validated instrument.
Resolution criteria
- 2035
- The full test below is met by 31 December 2035. New this cycle; this subject had no horizon grid before. Near zero, and the reason is a count rather than a judgment: the number of healthy people with no medical indication who have ever received an intracranial implant is zero, so the claim needs ten thousand of them in nine years starting from none. No regulatory channel exists in the EU, the United States or China through which such a device could be placed on a market, and the nearest cranial analogue with a severe indication and full reimbursement reached only eight thousand patients in twelve years.
- 2050
- The full test below is met by 31 December 2050. New this cycle. Twenty-four years requires the whole chain in sequence: a first healthy recipient, a regulatory channel built somewhere that admits an enhancement indication, then diffusion to ten thousand. The fastest implant diffusion on record, cochlear implants, took between four and eleven years from approval to ten thousand, which means approval would have to land around 2040 and therefore a submission in the early 2030s, in a field where every one of eight leading developers is currently filed for restoration only. The mass that remains sits almost entirely in a permissive jurisdiction shortcutting the sequence rather than in the sequence completing on time.
- 2100
- The full test below is met by 31 December 2100. New this cycle, and this is where nearly all of the probability accrues. Seventy-four years is long enough for the capability term to resolve, for one jurisdiction among many to build a channel, and for diffusion to a bar of ten thousand that every successful implant analogue has cleared by two orders of magnitude. The 2126 figure sits only three points above this one because the residual twenty-six years adds little to a process that will already have decided by then.
- 2126
- At least ten thousand people with no qualifying medical indication, in at least one jurisdiction, carry an implant that raises memory, reasoning, or learning above each recipient's own pre-implant baseline, sustained for at least twelve months and measured on a validated instrument.
Probability history
| Revised | State | p(2035) | p(2050) | p(2100) |
|---|---|---|---|---|
hf-forecast-0210 | researched | 0.5% no prior revision | 5% no prior revision | 19% no prior revision |
hf-forecast-0021 | researched | — | — | — |
hf-forecast-0004 | initial prior | — | — | — |
Strongest evidence for
- The premise that an implant is frozen hardware is factually wrong. Neuralink recovered an implant that had lost approximately 85 percent of its recording threads by changing software alone, to performance that 'superseded Noland's initial performance', with no further surgery (hf-evidence-0195).
- The key uncertainty cycle 5 named for itself is partly resolved in the claim's favour: stimulation can write information into hippocampal encoding for a roughly 37 percent recall gain, and can produce form vision in people with no functioning optic nerve. No external tool can do either, so a categorical capability gap exists (hf-evidence-0195).
- The non-invasive neuromodulation rival has collapsed rather than improved: pooled across 143 effects in 1,386 healthy adults, transcranial alternating current stimulation's effect on working memory is Hedges g = 0.076 and falls to zero once publication bias is corrected, with a second meta-analysis agreeing and a preregistered direct replication of the leading direct-current result failing (hf-evidence-0196).
- The ten-thousand bar is low against the century. Cochlear implants passed it within about a decade of approval and reached roughly 736,900 devices by 2019, so once the preceding terms resolve the threshold is not the binding constraint (hf-evidence-0194).
- A no-indication elective implant category already exists and is not empty: subdermal identification chips are estimated at between 50,000 and 100,000 people worldwide, which is above this claim's threshold, though the estimate has no registry behind it and the procedure is not neurosurgery (hf-evidence-0194).
- China's newest instrument declares enhancement products in healthy individuals not to be medical devices, which removes them from device law rather than prohibiting them — an ambiguity that could cut toward permissiveness in at least one jurisdiction (hf-evidence-0191).
Strongest evidence against
- The regulatory picture is not an absence but a written exclusion. EU MDR Annex XVI created a category for brain-stimulation products with no medical purpose, and the common specifications in force since 22 June 2023 state that 'Invasive devices intended for brain stimulation, such as electrodes or sensors that are partially or totally introduced into the human body, should not be covered by this Regulation' (hf-evidence-0190).
- Two further regulators drew the same boundary independently and recently: the FDA's only non-medical-device framework excludes anything invasive or implanted by definition, and China's NMPA guidance of 30 June 2026 places cognitive enhancement in healthy individuals outside medical device regulation entirely (hf-evidence-0191).
- Cycle 5's second discriminating observation returned a clean null. No 2024 to 2026 study raises cognition above a healthy participant's own pre-intervention baseline with an implanted device, and every apparent counter-example dissolves on population — the most-cited one studied 12 depression and 2 OCD patients with therapeutic implants already in place (hf-evidence-0192).
- DARPA, the institution with both the mandate and the money to enhance healthy human performance, redirected its healthy-population memory programme to explicitly non-invasive methods and confines its invasive arm to people undergoing neurosurgery for other conditions (hf-evidence-0192).
- Stated willingness fell rather than rose as the technology became real: 32 percent of US adults would want a cognitive-enhancement brain chip in March 2016 (n=4,726) against 20 percent in November 2021 (n=10,260), with 63 percent in the later wave calling it meddling with nature (hf-evidence-0193).
- The nearest cranial analogue has not cleared the bar. NeuroPace's own 10-K states 'As of December 31, 2025, over 8,000 patients have received our RNS System' — twelve years of commercial availability, a severe indication, and reimbursement (hf-evidence-0194).
- Electrode durability has not improved. Utah-array recordings average 622 days of useful life with roughly 40 to 50 percent yield, reproducing rather than overturning the figures cycle 5 used (hf-evidence-0195).
- Zero healthy people are documented to have received an intracranial implant anywhere, including outside formal medicine; the closest case is an extracranial recording-only sleep device in 25 volunteers (hf-evidence-0192).
Present uncertainties
- Whether the regulatory exclusions mean prohibited or merely unregulated, since declaring a product not a medical device removes it from device law rather than banning it — though an implant still requires a surgeon and a place to be implanted, so the permissive reading may buy less than it appears to; and whether a capability demonstrated only in people who have lost a function transfers to people who have not, which is the gap every piece of pro-claim evidence in this cycle has to cross.
Next discriminating observation
First and most important, and chosen because it is the one observation that could arrive through a channel that actually exists: the first person with no qualifying medical indication to receive any intracranial device, for any reason, anywhere — including in a jurisdiction with no enhancement pathway and including outside formal medicine. The count is currently zero and the step from zero to one is the whole of this forecast's near-term mass. Second: whether any regulator extends a non-medical-purpose framework to invasive devices, the sharpest available form being an amendment to EU Implementing Regulation 2022/2346 that removes or narrows the Recital 6 exclusion, which would reverse the largest term movement of this cycle. Third: the result of NCT06829368, the Nottingham focused-ultrasound cognition trial in healthy participants, listed for completion in March 2026 and unreported — because a non-invasive technique that DOES work in healthy people would restore the risk-free competitor that hf-evidence-0196 has just removed, and would cut this forecast rather than raise it.
Sources
- The EU built a legal category for brain stimulation with no medical purpose, then excluded implanted devices from it by name — so the prior record's claim that no category exists is wrong, and what replaced it is worse for this forecast
Regulation (EU) 2017/745 Article 1(2) states that the Regulation 'shall also apply, as from the date of application of common specifications adopted pursuant to Article 9, to the groups of products without an intended medical purpose that are listed in Annex XVI'. Annex XVI point 6 lists 'Equipment intended for brain stimulation that apply electrical currents or magnetic or electromagnetic fields that penetrate the cranium to modify neuronal activity in the brain.' The common specifications that switch this on are Commission Implementing Regulation (EU) 2022/2346, which by its Article 3 'shall apply from 22 June 2023'. Recital 6 of that Regulation reads, verbatim: 'The group of products listed in point 6 of Annex XVI to Regulation (EU) 2017/745 is intended for brain stimulation where only electrical currents or magnetic or electromagnetic fields penetrate the cranium. Invasive devices intended for brain stimulation, such as electrodes or sensors that are partially or totally introduced into the human body, should not be covered by this Regulation.' Its Annex VII, Section 1 repeats the exclusion operatively: 'This Annex applies to equipment intended for brain stimulation that apply electrical currents or magnetic or electromagnetic fields that penetrate the cranium to modify neuronal activity in the brain as listed in Section 6 of Annex XVI ... Such equipment includes devices for transcranial alternating current stimulation, transcranial direct current stimulation, transcranial magnetic stimulation and transcranial random noise stimulation. This Annex does not apply to invasive devices.' Commission Implementing Regulation (EU) 2023/1194 amends only the transitional provisions, extending a deadline for devices under clinical investigation from 22 June 2028 to 31 December 2029, and leaves the scope untouched. So the framework is in force today and the exclusion of implants is unamended.
Limit: Annex XVI point 6 itself, as printed in the parent Regulation, is written in device-agnostic language; the exclusion of implants lives in the implementing Regulation rather than in the Annex, so a future amendment could in principle extend it without touching the MDR. No product has been confirmed CE-marked under this common specification, which was searched for and not found — the category exists in law and its use is unverified. This is one jurisdiction, and the claim being forecast needs only one permissive jurisdiction anywhere in a century.
The master fetched the consolidated EUR-Lex HTML of 32022R2346 directly and quoted Recital 6, the Annex VII scope sentence and the Article 3 application date from that fetch, because this is the single most load-bearing fact in the cycle. Article 1(2) of the parent Regulation was fetched separately from the consolidated 02017R0745-20230320 text. The lane that first surfaced this had independently retrieved matching text from a different EUR-Lex URL, so the quote is corroborated by two retrievals of two documents.
- Two more regulators drew the same boundary independently, and China's is the most explicit of the three: enhancement in healthy individuals is not a medical device at all
UNITED STATES. The General Wellness policy is the only FDA framework for a product without a disease-treating indication, and it is bounded to low-risk products: 'A product is not low risk if it is (1) invasive, (2) implanted, or (3) involves an intervention or technology that may pose risk to the safety of users and other persons if specific regulatory controls are not applied.' An implant therefore cannot qualify by construction. The FDA's one BCI-specific guidance is titled, verbatim in the Federal Register, 'Implanted Brain-Computer Interface (BCI) Devices for Patients With Paralysis or Amputation—Non-Clinical Testing and Clinical Considerations', issued 20 May 2021 — restorative by its own scope. CHINA. NMPA issued its first BCI-specific classification guidance on 30 June 2026. English reporting of it states that 'Products intended solely for performance enhancement (e.g., augmenting perception/cognition in healthy individuals), entertainment (e.g., game control, VR/AR interaction), or daily assistance (e.g., smart home control) are not considered medical devices', and that where NMPA does regulate an invasive BCI it is 'Always Class III (highest risk)', with no lower-risk enhancement track. COMPANY FILINGS. Every indication located across eight developers is restorative: Neuralink's PRIME study enrols 'severe quadriplegia (tetraplegia) due to spinal cord injury or amyotrophic lateral sclerosis (ALS)', with Blindsight designated for blindness; Synchron for severe chronic bilateral upper-limb paralysis and ALS; Precision Neuroscience's Layer 7 cleared under K242618 for temporary intraoperative cortical recording and mapping of under thirty days; Paradromics' Connect-One for speech restoration; Blackrock for paralysis; Motif Neurotech and Inner Cosmos for treatment-resistant depression; Science Corp's PRIMA for geographic atrophy. Not one is filed, designated, or publicly stated as targeting cognition in a person with no medical condition.
Limit: Two of the three load-bearing quotes here are secondary. The NMPA finding in particular is a translation of a document nobody in this cycle read, and it is the newest and most consequential of the three, so it carries the most risk of a translation artefact. The United Kingdom is a genuine open question that this cycle failed to close: a 2022 Regulatory Horizons Council report recommended regulating all neuromodulation devices 'irrespective of level of invasiveness or purpose', reportedly accepted by government in 2024, but every relevant primary PDF was unreadable and the recommendation, even if in force, creates safety oversight rather than an enhancement indication. Japan and South Korea were not researched to conclusion and no absence should be inferred there.
The Federal Register titles and dates were confirmed against federalregister.gov records. The FDA General Wellness sentence could not be read from the FDA PDF, which returned unparseable binary, and is reproduced here from two independent law-firm summaries of the same guidance that agree with each other. The NMPA quotes are English translations appearing on two independent Chinese regulatory-commentary sites that agree; the primary Chinese document on nmpa.gov.cn was not fetched by anyone in this cycle. The Precision clearance number was read from the FDA CDRH clearance PDF URL.
- Cycle 5's second discriminator returned a clean null across 2024 to 2026, and the strongest version of the null comes from the institution most incentivised to break it
DARPA's Restoring Active Memory programme page states that RAM Replay 'is developing new closed-loop, non-invasive systems that leverage the role of neural replay in the formation and recall of memory' and 'aims to support military training and effectiveness by using non-invasive interventions to accelerate and improve the performance of complex, military-relevant skills by healthy individuals.' The invasive arm of the same programme is confined to 'volunteers undergoing neurosurgery for other neurological conditions'. The apparent counter-examples each fail on population. Widge et al., whose title reads 'Deep brain stimulation of the internal capsule enhances human cognitive control and prefrontal cortex function' and which reports subjects 'on average 34 ms faster with DBS ON (t = -8.6, p < 1.33e-17)', studied 'Fourteen subjects (12 MDD and 2 OCD ... with VCVS DBS implants)' who 'had received VCVS DBS implants for a prior clinical study' — a within-subject stimulation-on versus stimulation-off comparison in treated psychiatric patients, not a pre-implant healthy baseline. The 2024 hippocampal-prosthesis paper draws subjects who were 'selected from patients with medically-refractory focal epilepsy who were undergoing seizure monitoring and localization through the use of implanted intracranial depth electrodes.' Nia Therapeutics' reported gains of approximately 19 percent in participants with a history of moderate to severe traumatic brain injury and 28 percent in epilepsy patients were produced on a research prototype running on electrodes placed for epilepsy monitoring; the company's own dedicated implant has not been placed in any human and its first-in-human study had not started. The closest documented case of an implant in healthy people with no medical indication is a subcutaneous EEG device placed 'deep to the galea' in 25 healthy volunteers for ultra-long-term sleep research — extracranial, recording-only, no stimulation capability and no cognitive claim.
Limit: This is a capped search of English-language indexed literature and web sources. Closed communities, non-English publication and private clinic marketing were not reached, so the absence of any healthy intracranial implant is an unfound case rather than a proven absence. The DARPA sentence most often quoted for the 'decades' judgment — that it will take decades before invasive brain implants can be used in healthy individuals — comes from a secondary defence-media source, not from the directly fetched DARPA page, and is reported as such.
The DARPA programme page, the Widge paper and the subcutaneous EEG paper were fetched directly and quoted verbatim. The 2024 hippocampal-prosthesis population sentence and the Nia figures are search-result synthesis of the underlying sources rather than direct fetches of the papers, and are reported at lower confidence; the Nia press release returned HTTP 403 and PubMed returned a cookie wall.
- Stated willingness to accept a cognitive-enhancement brain implant FELL by twelve points between 2016 and 2022, while the technology became visibly more real
In the 2016 wave, respondents were given 'a brief hypothetical scenario involving the surgical implantation of a computer chip in the brain for the purpose of giving people a much improved ability to concentrate and process information'; 32 percent said they would want this for themselves and 66 percent would not. The sample was n=4,726, fielded 2 to 28 March 2016 on the American Trends Panel, 4,243 by web and 483 by mail. In the wave fielded 1 to 7 November 2021 and published 17 March 2022, n=10,260 US adults, 78 percent said they would not want a chip implant to better process information and 20 percent said they would. In the same wave, 56 percent said widespread use of brain chips to enhance cognitive function would be a bad idea for society against 13 percent saying a good idea, 63 percent said it would be 'meddling with nature and crosses a line we should not cross', and 5 percent said they would like to get such a chip within the next year. The same respondents were far more favourable to restorative uses: 77 percent favoured brain chips to allow increased movement for people who are paralyzed and 64 percent favoured treating age-related mental decline.
Limit: The two waves are not a clean panel comparison. Wording differs — 'a much improved ability to concentrate and process information' in 2016 against 'better process information' in 2022 — and the 2016 wave had a mail component. A twelve-point fall across six years is larger than plausible wording noise in the same direction but this is a judgment, not a tested difference, and no significance test was performed or located. Stated willingness also massively overstates revealed willingness in every elective-procedure analogue, so 20 percent is an upper bound on something, not a forecast of uptake.
Both Pew pages returned HTTP 403 to direct fetch attempts by two independent lanes. The percentages, sample sizes and field dates above are reproduced from search-result text quoting those pages, and the 2016 methodology figures come from Pew's own separately indexed methodology page. No figure here was read from a directly fetched primary document, which is the weakest provenance of any load-bearing evidence in this cycle.
- The nearest cranial analogue has not reached ten thousand patients in twelve years, with a severe indication and full reimbursement — but every analogue that got past the bar went on to pass it by two orders of magnitude
NeuroPace's Form 10-K for the fiscal year ended 31 December 2025 states: 'As of December 31, 2025, over 8,000 patients have received our RNS System.' The same filing records that 191 patients were enrolled in the pivotal study between December 2005 and November 2008, with FDA approval in late 2013 and 'commercial rollout beginning in early 2014'. That is twelve years of commercial availability, for drug-resistant focal epilepsy, without clearing ten thousand. Cochlear implants are the counterweight: the NIH consensus conference of 1988 recorded 'approximately 3,000 patients had received implants' and the 1995 conference 'approximately 12,000 patients had received implants', so the ten-thousand line fell between four and eleven years after the 1984 adult approval; by December 2019 NIDCD reports 'approximately 736,900 registered devices have been implanted worldwide'. Penetration of the eligible population nonetheless stays low and flat: 'an estimated 5 to 7% of eligible US adults receive cochlear implants', and for deep brain stimulation 'as few as 10-15% of eligible PD patients are referred to DBS treatment centers'. The closest analogue to an implant with no medical indication at all is the subdermal RFID or NFC chip, estimated at 'between 50,000 and 100,000 people already have been chipped' worldwide after roughly eighteen years — an estimate with no registry behind it.
Limit: Every device in this table has a medical indication, a referring physician and an insurer, and the claim under forecast has none of those. The direction of that difference is not obvious: it removes the reimbursement engine that drove these curves, and it removes the gatekeeping that throttled them. The RFID figure, which is the only genuinely indication-free analogue, is the least reliable number here.
The NeuroPace sentence was read from the 10-K as filed on SEC EDGAR and is the strongest-provenance number in this cycle. The cochlear consensus-conference figures and the NIDCD total were read from the cited PMC and NIH pages. The subdermal-chip range is a community and industry self-estimate with no independent count and is reported as such. A widely repeated claim that pacemakers reached ten thousand units by the early 1960s was chased to its citation by the lane and dissolved — no primary source was found — so it is excluded rather than used. Two sources for cumulative worldwide DBS implants disagree by roughly fifty percent, 244,000 against 160,000, and were left unadjudicated.
- Two of the three premises behind this seed's own structural argument are wrong: an implant is not frozen, and there is a capability no external tool has
FROZEN HARDWARE. Neuralink disclosed in May 2024 that 'approximately 85 percent' of its first participant's implant threads had retracted, causing a sharp reduction in usable electrodes and in bits per second. The company modified the recording algorithm and the signal-translation techniques and pushed them without further surgery; the reported result is that these fixes 'produced a rapid and sustained improvement in BPS, that has now superseded Noland's initial performance.' An implant therefore does improve on a software cadence, on unchanged hardware, and can recover past its own pre-degradation baseline. WRITE CAPABILITY. The USC and Wake Forest hippocampal prosthesis extracted memory-encoding signals from participants and injected them back during recall, boosting performance by 'roughly 37 percent', described by its investigator as the first time researchers could 'identify a patient's own brain cell code or pattern for memory and, in essence, write in that code to make existing memory work better.' Separately, dynamic stimulation of 60 subdural electrodes over visual cortex produced recognisable form vision in blind subjects with no functioning optic nerve, with the six-subject Orion study running January 2018 to March 2025 and 'a loss of functionality on fewer than 4% of electrodes'. No external device can inject information into hippocampal encoding or produce vision without an optic nerve. DEGRADATION, WHICH SURVIVES. Utah-array longevity data give an 'average lifespan of available recordings' of 622 days, with 'nearly 50% of implants' showing year-long recordings above 40 percent electrode yield, and rare examples reaching nine years — reproducing rather than overturning the figures cycle 5 relied on.
Limit: Both write-capability demonstrations are in impaired populations — epilepsy patients for the memory prosthesis, blind subjects for the visual cortex work — so a capability that has no external substitute for a person who has lost a function is not yet shown to be a capability worth surgery for a person who has not. The software recovery is bounded by a ceiling the argument correctly identifies: firmware can extract more from the electrodes that exist and cannot add electrodes or move them, whereas a phone is replaced outright.
The Neuralink figures and the quoted outcome sentence are the company's own disclosure as reproduced by two independent news organisations; the primary Neuralink blog post was not fetched. The Cell paper, the USC release and the Utah-array longevity paper are cited at their primary URLs and were located by the lane rather than re-fetched by the master. The 37 percent memory figure is a 2018 result restated in a 2024 follow-up, not a new result.
- The risk-free rival that the structural argument assumed was improving has instead collapsed: transcranial stimulation's effect on working memory goes to zero once publication bias is corrected
The 2024 meta-analysis 'Transcranial alternating current stimulation barely enhances working memory in healthy adults' pooled '143 effects of tACS on working memory performance found in 42 reports published between 2014 and 2023, encompassing 1386 healthy adults', reporting 'an overall effect size of Hedges g = 0.076 [0.039, 0.113]. However, after correcting for significant publication bias this effect size dropped to zero.' A separate 2025 theta-tACS meta-analysis of 21 studies and 67 effect sizes found an initial Hedges g = 0.405 (95% CI 0.212 to 0.598) that 'became nonsignificant after correcting for publication bias (trim-and-fill adjusted Hedges g = 0.082, 95%CI: -0.052 to 0.217)'. For direct current stimulation, a preregistered replication in Communications Psychology reported that 'stimulation of either the posterior parietal cortex or the DLPFC failed to improve visual working memory capacity', testing 48 participants against the original study's 20. The one registered trial designed specifically to test transcranial focused ultrasound for cognitive enhancement in healthy participants, NCT06829368 at the University of Nottingham with completion listed March 2026, had not reported results.
Limit: This measures one family of non-invasive techniques on one cognitive domain. It says nothing about focused ultrasound, whose flagship cognition trial has not reported and is therefore open rather than negative, and nothing about the phone, which is the external competitor cycle 5 actually named and which was not tested here at all. Neither meta-analysis reported persistence of effect, so nothing here bears on duration.
None of these three sources was fetched directly. PubMed returned a cookie wall and the other two were not re-fetched; every figure here is search-result synthesis of the cited records, which is weak provenance for load-bearing evidence and is the reason this record is used to narrow an argument rather than to move a probability on its own.