A $500 Lamp Cut TB Transmission 90 Percent. Almost Nobody Buys One

Nathan Labenz, host of *The Cognitive Revolution*, cross-posts a Complex Systems episode from Patrick McKenzie (better known online as Patio11) with Aerolamp CEO Misha Gurevich and chief scientist Vivian Belenky, a Columbia University researcher. The primer is 222-nanometer far-UVC: a wavelength that can inactivate airborne pathogens in shared rooms. Lab science is already well understood, Labenz says, even as field evidence is still arriving. The headline result is a South African hospital trial with preliminary numbers: about 90 percent suppression of tuberculosis transmission in wards.

Philanthropist Jeff Kaufman, now working full-time on catastrophic-pandemic prevention at Secure Bio — an organization Labenz has supported with a modest donation — specifically recommends Aerolamp. At $500 the fixture can feel expensive, Labenz argues, until you price a single cold in lost productivity. For the immunocompromised the value is higher still. He and his wife installed a unit directly over their son's bed during cancer treatment last winter. The child is back at school; Labenz plans to buy more for the classrooms. If anything like a 90 percent drop in transmission shows up, the social return would be outstanding.

He also puts a larger chip on the table. AI systems are becoming extremely capable; the first autonomous cyberattacks have already appeared, and similar abilities, by many accounts, are coming to biology. Buying a 222-nanometer lamp now, from Aerolamp or otherwise, is a way to invest in one's own health and to expand a nascent market for what might soon be critical infrastructure. McKenzie, for his part, calls far-UVC one of the sleeper picks among important technologies being developed today.

Belenky's chemistry is blunt. Far-UVC is absorbed by pathogen DNA and RNA and by essentially all proteins. That protein absorption is what makes it much safer than other germicidal UV. A microorganism without the protections humans and animals have is inactivated: still in the air, no longer able to replicate, extremely fast. The effect is like an extremely strong air purifier, except instead of one or two extra air changes you can get the equivalent of 30 to 50. McKenzie assumes the wavelength is already in sunlight. It is not. Sunlight is mostly UVA and UVB; UVB is the skin-cancer and cataract worry. UVC is blocked entirely by the ozone layer, which makes evolutionary sense: if it reached the surface, microbial life — and the planet — would look different. Sunlight does kill germs, Belenky says, only because there is so much of it that mildly germicidal photons still add up. The lawyer in McKenzie notes that Oliver Wendell Holmes has been empirically disproven: sunlight is not, per photon, the best disinfectant.

Safety, she continues, is more mechanical than chemical. UVC of any wavelength is not good for living cells. Humans have a roughly 20-micron layer of dead skin packed with protein that absorbs essentially all far-UVC — unique to the 200-to-235-nanometer band, commercially 222 nanometers. Longer germicidal wavelengths used in water disinfection, 254 to 265 nanometers, lack that protein wall. They are thought less carcinogenic than UVB, but they are not pleasant, and they demand careful upper-room installs. What far-UVC does not absorb in the stratum corneum is stopped in the upper living layers, which slough off in a couple of days. Belenky has never seen a study showing significant biological activity at the basal layer, where DNA damage would raise cancer worries.

Eyes are harder: there is no convenient dead-cell layer. Eyelids, lashes, brows, and the brow ridge still cut dose the way they do for sunlight, and the tear film's lipids and proteins absorb only about 15 percent. The rest hits the epithelium. A relatively lower dose can produce pain or discomfort — which Belenky treats as good news, because pain makes you flinch, unlike an infrared laser that can blind you with no sensation. Ambient sitting in a room should not suddenly hurt; the accident case is climbing up and staring into the emitter for minutes. One-year and three-year eye-safety studies are running in Japan. Industry practice stays conservative, with a lower safe eye dose than skin dose. Power is modest: emitters around 100 milliwatts, falling toward zero across a room, enough for rapid germicidal effect without a blinding flood.

Current units are corner-mounted because the beam is narrow: point from one corner to the opposite to maximize path length and average dose. The long-term form factor, Gurevich and Belenky say, is a boring ceiling fixture — like a smoke alarm in a drop ceiling. Today's lamps are an existence proof that the technology is ready: no insurmountable technical or logistical wall, not a $10,000 government-lab toy. McKenzie's world-theory is that anything that fits in a shipping container from China craters in price as production scales. Medical interventions often stay expensive under Baumol's cost disease; this one should eventually cost not much more than lighting. Almost no institution builds a hospital and then cannot afford the lights.

Rule of thumb: about 250 square feet per lamp. A standard classroom takes two to three, maybe four if large; a small building four to five figures to outfit; a university hundreds of lamps and six figures before installation. If you hire electricians, installation roughly equals lamp cost; DIY is basically free — ten minutes, drywall anchor, plug. Electricians already know light fixtures; 222 is innately safer than 254, so a botched upper-room mercury install is no longer the failure mode. Manufacturer claim is at least 10,000 hours at 70 percent output, with data suggesting 13,000 to 14,000. Eight-hour workdays imply five or six years; 24/7 in an airport or station is about a year and a half per bulb.

The best emitters, krypton chloride excimer lamps, currently come from one Japanese firm in a high-end, high-margin business. Even stripping their margin leaves something like $15 to $20 per meter, plus hydrogen fluoride in the process — feasible to scale, not LED-easy. Belenky pushes back that LED chips were *more* capital-intensive, tens of billions into fabs, after which scale is ferocious. Krypton chloride sits on a gentler cost curve and will never match white-LED prices, but the floor is nowhere near. Solid-state far-UVC chips are far out. She would be joyful to have the problem of shaving cost on tens of billions of lamps. Competitors sell at $2,000, someone recently at $3,500: high margins for tiny companies, not physics. Aerolamp's $500 bet is an inflection: Gurevich sees no reason the price could not hit on the order of $100 in a couple of years with zero technological innovation. You can order from aerolamp.net, open a box, and stick it on a tripod, a bookshelf, or a wall.

McKenzie asks the obvious question: why is this not already in every hospital? Belenky says he hit the nail on the head. It is will, plus people not knowing the option exists. Far-UVC currently sits with "things only weird people think about," like wearing a respirator everywhere. It needs to become hand sanitizer in the lobby, or ordinary ventilation. She came into the field hunting bottlenecks — cost, research, regulation — and found none fatal. More research on best practice would help. There is no good reason it cannot be everywhere.

McKenzie ranks public-health history. Water disinfection is centralized: convince one utility. Surgeons washing hands is an unfinished negotiation because every doctor must do it every time. Lamps are a one-time, room-by-room intervention. You do not need every room, but you need many rooms where people congregate. A building owner can decide unilaterally. If infection prevention is written into codes, it rides the roughly ten-year commercial renovation cycle. The only current standard is ASHRAE 241, assembled quickly, still under construction, not yet broadly adopted by authorities having jurisdiction. If the standard names a clean-air target and UVC is the cheapest way to hit it, people will do that.

A heckler's veto is the research-shaped trap. Formal trials need IRBs; one uncomfortable person can kill the protocol. A building owner who simply installs is much less exposed, and employees, the guests find, are usually more eager than the manager who spends the money. Once the hardware is in because someone wants the benefit, nothing stops you from studying it later. Compared with vaccines — time, a jab, visceral unpleasantness — a dark box in the corner that emits nothing visible is an easier political object, even in a post-pandemic climate of public-health skepticism.

Asked where to prioritize, Belenky puts elementary and maybe middle schools first: often extremely poorly ventilated, children immunologically naive. Healthy adult volunteers are a lousy way to study airborne flu; they often simply do not catch it. Children are another story. Parents, though, will be conservative about a relatively new technology, so she expects occupational settings, long-term care, and hospital waiting rooms to move earlier even if they are not the largest benefit. Waiting rooms have a useful dose profile: no one spends plural days there, so photobiological worry is weaker, while aggregate dwell time across hundreds of patients is large. Private schools may lead public ones because, on McKenzie's envelope math, a single parent who gets the school on board could fund it. Some of his tech friends are already discussing that.

Randomized trials of environmental interventions are hard to design; early adopters should expect a sigmoid. Install in the office and nobody shares flu there, then the kid's school has none and you catch it at home. Coverage has to hit a community threshold before suppression takes off. Gurevich is more optimistic about specialized, low-mixing settings and about pathogens that are either especially susceptible or, usefully, especially resistant. Tuberculosis is not an American dinner-table topic, but it clusters in institutions and countries with less social mixing. The South African guinea-pig study — animals sharing only air with patients — already has that 90 percent figure. TB is about ten times more resistant to far-UVC than typical respiratory viruses, so the number is a conservative lower bound, not a best case. Flu and coronavirus do not transmit exactly like TB, Belenky cautions, but there is reason to be optimistic. Long-term care and boarding schools should show results faster than a mixed community school.

When infections drop on the order of 90 percent, McKenzie says, anecdotal evidence will pile up in certain communities without waiting for confidence intervals. A long-term care chain, PE-owned or otherwise, prefers residents not to die — humanitarian mission plus census-based revenue. Install in one site, check the month: 34 infections in the median location, three in the treated one, then make the obvious decision. Knock on wood, that kind of viral coefficient inside a community of practice could run on a scale of months to a few years, then a broader rollout. Google-scale employers who already have someone accountable for employee health only need arithmetic; caregiver absenteeism, not just worker sick days, is a large slice of colds' economic cost. Finance firms bought ventilation and filtration early in COVID without a state mandate. A corner box is easier to sell than needles.

Gurevich's most exciting case is pandemic prevention, which he avoids in ordinary marketing because people are collectively traumatized and do not want to hear it. He is more uncertain about stopping the average cold — often a long close interaction — than about kneecapping a future COVID-like respiratory pandemic. Paradoxically, the more contagious a pathogen, the more surface area there is to keep it from getting as bad as it could. Traditional 254-nanometer UVC was used to control measles, with a reproduction number around 20, probably the ceiling of contagiousness. COVID-19 at its worst was something like one-point-something. Relatively low coverage in transport hubs and gathering places where people are not already a community could do pandemic suppression at relatively low cost.

McKenzie maps R0 onto the K-factor video-game marketers use. 1.08 goes exponential; 0.99 fails, and you have to keep dumping marketing budget — which viruses do not have. A small absolute impairment that takes a COVID-like bug from slightly above one to slightly below one is the difference between a pandemic and not. Because the intervention lives in the building, far fewer people have to opt in than for vaccination or masking. One owner, or a board, can cut infections in a space. HVAC upgrades that actually move more air are often impossible or ruinously expensive in existing buildings; you plop a light in the corner instead. Portable MERV-13 cleaners are additive — pathogens are not the only pollutant, and UVC's effect on allergens via protein absorbance is modest next to filtration. In auditoriums, lecture halls, and gyms, meeting a clean-air standard without UVC is essentially cost-prohibitive. Vaccines do not get worse because less virus is circulating. The combination is probably multiplicative. McKenzie once bet on broad U.S. deployment by 2030 and now expects to lose. Worldwide sales may be a couple of hundred lamps a year, not even clearly rising. LED-style takeoff still implies at least a decade after the idea goes viral, riding that ten-year renovation cycle. Belenky lies awake because people assume there must be a regulatory, cost, or critical-safety gap. There isn't. It is social diffusion. She has spent a career in research, not in global awareness campaigns, and wishes she had.

Should you get an Aerolamp at home? Gurevich usually says no. Most private houses do not have twenty occupants; transmission inside a small family is limited. Immunocompromise, unusual wealth, or an unusual taste for not getting sick can flip the math. At $500, for most people, it does not price out; at $100, maybe. Belenky pushes back. Societal benefit is not concentrated in individual homes, but individual benefit can be. She had a baby two months ago. Infants essentially lack immune systems, she wanted postpartum visitors, and she has two lamps in the living room that she switches on for gatherings — like a kitchen ventilator, a home-health appliance. A fever in a tiny baby means an emergency trip, hospitalization, possibly a spinal tap. You should be extremely willing to pay to cut that probability. Labenz's hospital-bed install is the same logic at the extreme. Consensus: homes are a personal-risk call; the social-suppression payoff is in schools and transport hubs.

McKenzie asks the investor's clarifying question: if this fails to deliver societal benefits, why? Gurevich's answer is transmission dynamics. If a huge chunk of ordinary cold and flu is strongly short-range, with little long-range airborne spread, the lamps will not do much for "normal" disease. Highly contagious pandemic pathogens probably *need* more long-range spread, so you could land in an uncomfortable world: crucial anti-pandemic insurance, weak everyday business case, because people hate paying for prevention they will not see this year. Belenky's numbers: a typical far-UVC install is about one equivalent air change every two minutes, 90 percent coronavirus or influenza reduction in about eight minutes, 99 percent in about fifteen. If infection is an hour of shared air, far-UVC cuts that path. If it is a two-foot conversation and a massive dose, no environmental intervention does much, though mixing the airstream can help a little, and a smaller inhaled dose can still mean a milder illness — as COVID showed. No benefit at all, she thinks, is quite unlikely.

McKenzie floats the precautionary objection: children are immunonaive, and maybe decreasing pathogen contact is bad. Belenky pushes back hard. The modern hygiene hypothesis does not say you need clinical illness to train immunity. Training comes from environmental and commensal bacteria. There does not seem to be a benefit to an episode of sickness. RSV in a child looked neutral to negative for future illness; measles wrecks immunological memory and is purely bad. The emerging view is that all viruses are like this: you would rather not catch them, and if you must, later is better than earlier. End viral infections and, in her view, that is overwhelmingly good. McKenzie notes that if some optimal viral load ever appeared, we already have a technology for controlled exposure: vaccines. Asthma and allergy are up; childhood mortality is way down, and those allergy trends have nothing to do with viral infections. Indoor microbiomes have not been explicitly studied under far-UVC; surface bacteria are hardy, UVC is much less effective on surfaces than in air, and outside there is no shortage of dirt and bugs. If there is a cost, she bets it is minor.

McKenzie is harsher on the precautionary principle: in a million-dimensional world you can always find one axis a technology makes worse, which is a rule for never doing anything. Infectious-disease death numbers are horrific. A history-book line between pre-ending and post-ending of flu would be a world-historical achievement — not something they 100 percent expect from far-UVC, but inside the outcome set. Belenky suspects long COVID is not special; many viruses have long-term sequelae, measles among them. It is possible, McKenzie adds, that urban humanity has been living under unlabeled "long flu." Some viruses cause cancers; he still coughs for two months after a bad winter cold. Unknown-unknown downsides exist; so do unknown-unknown upsides. Relative to vaccines, this intervention is easy to desist: turn the lamp off and the dose stops. You are mostly irradiating dead outer skin. At some point reasoning about exotic protein-diffusion edge cases, he says, just gets nuts.

Asked for a closing thought, Gurevich speaks as a capitalist who would love people to buy Aerolamps, then as someone in the industry: awareness matters more. A global information campaign could benefit the world more than almost anything else you can do. Most people have never heard of UV for disinfection; among those who have, almost nobody has heard of 222. The pitch is easy science fiction — special wavelengths, less sickness — and everyone is eager once they know. Belenky adds that far-UVC is only one piece of clean air, with filtration and ventilation. Parents should lean on schools, daycares, workplaces, and community centers: how clean is this air, can we make it cleaner, can I help? Aerolamp is cheap by comparison and open source; CAD files are on GitHub, so in principle anyone can build their own. Still-COVID-cautious groups are fighting the good fight and it is not enough. Clean air needs a social movement and a broad coalition to upgrade buildings that make people sick. McKenzie asks whether a well-resourced capitalist's first million should go to trial deployments and papers or to getting Taylor Swift to adopt. Belenky leans toward the latter as the limit, then Gurevich prefers trials as a more robust way to spread the word — too much celebrity attention this early could backfire. If he had to pick a celebrity it would be Paris Hilton: high-end hotels associated with clean air. Meanwhile, lamps already sit in San Francisco and Berkeley venues and in D.C. event spaces, becoming ordinary infrastructure by being around.

Belenky's practical ask is not a celebrity campaign. If you have a child, lean on the school, the daycare, the workplace, the gym, the community center: what is the air like, is there filtration, ventilation, or UVC, and can you help. Parent pressure is itself an effective move. A classroom is two to four lamps at $500 each; electrician cost roughly matches hardware; a single parent can fund it without waiting for district budget. Households with a pregnancy, a newborn, an immunocompromised member, or frequent hospitalizations should do the arithmetic against one ER visit. Offices and shops can forward the link to whoever owns absenteeism: one cold's lost work can exceed $500. Google-shaped firms with a person accountable for average employee health only need to run the numbers, as finance firms did with ventilation during COVID.

Labenz's closer is the same chip he put down at the start. AI is pushing biological risk closer. The first autonomous cyberattacks are here; similar capability will spread into biology. Before that diffusion, $500 for a lamp is health insurance and an early vote that 222 nanometers becomes a life-saving piece of ordinary infrastructure. The fixture is for sale. You plug it in. Aerolamp.net.