-
Space Elevators in 2026: Engineering Fantasy or Eventual Reality?
The concept is simple enough to explain on a napkin and difficult enough to build that it’s been 130 years since anyone first described it and we’re still nowhere close. You put a satellite in geostationary orbit—35,786 kilometers above the equator—and you lower a tether all the way down to the surface of the Earth. Anchor it at the bottom. Attach a counterweight above geostationary altitude to keep the whole thing taut. Then you send climbers up the tether, hauling cargo to orbit without a single gram of rocket fuel. The cost per kilogram to geostationary orbit drops from roughly $20,000 on a conventional launch vehicle to an estimated $500. The environmental impact drops to essentially zero. You could, in theory, send 170,000 metric tons to orbit per year on a mature system.
The Russian scientist Konstantin Tsiolkovsky described the basic idea in 1895 after visiting the Eiffel Tower. Arthur C. Clarke popularized it in his 1979 novel The Fountains of Paradise. A 2003 NASA Innovative Advanced Concepts study concluded that a space elevator “could be built in the near future with acceptable risk and less funding than some current space programs.” The key word in that sentence turned out to be “could,” because twenty-three years later we still cannot manufacture the tether material, and the tether material is the entire problem.
One material, and everything depends on it
A space elevator tether needs to be 100,000 kilometers long, roughly one meter wide, and about as thick as plastic wrap. It needs to support its own weight—which, at that length, is enormous—plus the weight of multiple climbers carrying payloads. The required specific strength (tensile strength divided by density) is approximately 50 to 60 GPa·cm³/g. For reference, the specific strength of steel is about 0.25 GPa·cm³/g. Kevlar is about 2.5. The best carbon fiber composites reach maybe 4. You need a material that is roughly 15 to 25 times stronger per unit weight than the best structural material in common industrial use.
In 2026, there are exactly three known materials with the theoretical tensile strength to serve as a space elevator tether: carbon nanotubes, single-crystal graphene, and hexagonal boron nitride. Carbon nanotubes have been the poster child for space elevator materials since the 1990s. Their theoretical tensile strength is approximately 150 GPa—more than adequate. The problem is manufacturing them. The longest single carbon nanotube ever publicly reported is 0.5 meters. Nanotube “forests”—bundles grown on a substrate—have reached 14 centimeters at Waseda University, at a growth rate of one meter every 186 hours. The tether requires 100,000 kilometers of continuous, defect-free material. The gap between 0.5 meters and 100,000 kilometers is not a gap that incremental manufacturing improvements are going to close on any human timescale.
This is why the International Space Elevator Consortium—yes, there is one, and they publish monthly newsletters on tether materials research—has increasingly shifted its focus to graphene. Graphene was isolated for the first time in 2004 and won the Nobel Prize in 2010. Its theoretical tensile strength is approximately 130 GPa, comparable to carbon nanotubes. But here’s the critical difference: polycrystalline graphene can already be manufactured at lengths of one kilometer and speeds of two meters per minute. Multiple industrial companies are producing it commercially. The material isn’t at tether quality yet—you need single-crystal graphene with no grain boundaries or defects, manufactured as a continuous sheet at industrial scale—but the trajectory from “lab curiosity” to “industrial product” is incomparably more advanced than the trajectory for nanotubes.
ISEC’s current leading candidate is what they call “graphene super laminate”—multiple layers of single-crystal graphene bonded together through a process they describe as “spot welding” using covalent carbon-carbon bonds. In theory, this creates a material where each layer retains graphene’s extraordinary in-plane strength while the interlayer bonds prevent the shearing weakness that plagues regular multilayer graphene. In September 2025, ISEC reported that spot-welding layers of graphene had been demonstrated in the lab and produced a material with diamond-like properties. In February 2026, they published research on atomic oxygen corrosion resistance of graphene super laminate—addressing one of the critical environmental hazards a tether would face in low Earth orbit.
Whether graphene super laminate can actually be manufactured at 100,000-kilometer continuous lengths, at tether-quality purity, at a speed that doesn’t require decades of production time, and at a cost that makes the project economically viable rather than merely physically possible—that remains entirely undemonstrated. The gap has narrowed. The gap is still enormous.
Everything else that’s also impossible
The tether material gets all the attention because it’s the most obvious bottleneck, but it’s worth cataloging the other engineering challenges that would need to be solved even if someone handed you a perfect tether tomorrow.
The climber system needs to ascend 35,786 kilometers to geostationary orbit. At reasonable speeds, that’s a multi-day journey—Obayashi Corporation’s 2012 design estimated eight days. The climber needs to be powered the entire way, and it can’t carry all its fuel because that would make it too heavy. Proposed solutions include ground-based lasers beaming power to photovoltaic cells on the climber, which introduces its own set of engineering problems including atmospheric attenuation, beam tracking accuracy across thousands of kilometers, and what happens to anything that accidentally flies through the beam path.
Space debris. The tether passes through low Earth orbit, where thousands of tracked objects and millions of untracked fragments are traveling at orbital velocity—roughly 7.8 kilometers per second. A collision between a piece of debris the size of a marble and a tether the thickness of plastic wrap would be catastrophic. ISEC published a June 2025 analysis titled “The Space Elevator Tether and Space Debris: Irresistible Force Meets Impenetrable Object?” The paper-thin ribbon design proposed by researcher Bradley Edwards would help—a ribbon can survive small punctures because the stress redistributes across its width—but routine avoidance maneuvers for tracked debris would still be necessary, and the tether can’t exactly dodge.
Atmospheric hazards. The bottom portion of the tether passes through the troposphere, where it encounters wind loads, lightning strikes, and weather of every variety. The portion passing through low Earth orbit encounters atomic oxygen, which corrodes most materials. The Van Allen radiation belts degrade molecular bonds over time—though studies suggest carbon nanotubes could survive radiation for over 1,000 years. Gravitational perturbations from the Moon and Sun create oscillations in the tether that need damping systems.
The anchor station. The tether needs to be anchored at the equator, ideally on an ocean platform to allow positional adjustments and avoid geopolitical complications. Building and maintaining a floating platform capable of anchoring a structure under millions of newtons of tension, in equatorial waters, indefinitely, is itself a major engineering project.
Where the money actually is in 2026
A market research report published in March 2026 values the “space elevator market” at $720 million, projected to reach $1.16 billion by 2030. These numbers require some decoding, because there is no space elevator to buy or sell. What the market consists of is materials research (primarily carbon nanotubes, graphene, and boron nitride), climber system design, tether dynamics modeling, and related R&D. The major corporate names are Obayashi Corporation (which still maintains its 2050 target date for a completed space elevator), Shimizu Corporation, Tethers Unlimited, and the LiftPort Group—whose CEO admitted in 2019 that “little progress had been made” on their original space elevator ambitions despite years of effort.
Google X investigated the concept around 2014 as part of its Rapid Evaluation R&D team. They concluded that nobody had manufactured a perfectly formed carbon nanotube strand longer than a meter and put the project in “deep freeze,” where it has remained. They reportedly keep tabs on materials science advances, which is Silicon Valley’s polite way of saying “we’ll wait.”
The honest assessment
The International Academy of Astronautics published feasibility assessments in 2013 and 2019, both concluding that Earth-based space elevators are feasible in principle and that the critical bottleneck is the tether material, which they projected could achieve the necessary specific strength “within 20 years.” That projection, made in 2013, would put the material breakthrough at approximately 2033. The graphene super laminate research is consistent with that timeline in the sense that the trajectory is visible, even if the destination hasn’t been reached.
Obayashi’s 2050 target—an operational space elevator with an eight-day trip to geostationary orbit—is the most specific commitment from a credible engineering firm. Whether it’s achievable depends almost entirely on whether single-crystal graphene or an equivalent material can be manufactured at scale within the next decade, which is a materials science question that nobody can answer with confidence.
The comparison that clarifies the situation: in 1903, the Wright brothers flew at Kitty Hawk. In 1969, Apollo 11 landed on the Moon. Sixty-six years from first powered flight to lunar landing. The space elevator concept has existed for 130 years. The materials science necessary to build it has been actively researched for roughly 30. The longest carbon nanotube is half a meter. The required tether is 100,000 kilometers. The concept is not fantasy—the physics is sound, the engineering challenges are understood, and the materials are making measurable progress. But “eventual reality” is doing a lot of work in that phrase, and “eventual” might mean 2050, or 2080, or later, depending on breakthroughs that cannot be scheduled.
We cover space elevators alongside 23 other civilization-scale engineering challenges—from fusion reactors to ocean thermal energy to asteroid mining—across our Moonshot 2169 course. If the gap between “the physics works” and “we can actually build it” is where your brain lives, the course is 24 lectures of exactly that tension.
-
Are There Any Successful Utopian Communities Still Operating in 2026?
The short answer is yes—depending entirely on how you define “successful” and how generous you’re willing to be with the word “utopian.” The Foundation for Intentional Community maintains a directory of over 1,000 intentional communities worldwide, housing an estimated 100,000 people. These range from income-sharing communes in rural Missouri to cohousing developments in suburban Denmark to ecovillages in Scotland. Some have been operating continuously for over a century. Others launched last year and may not survive to next year. The survival rate for utopian communities has always been brutal—most experiments in communal living fail within a decade—but the ones that endure tend to share a set of characteristics that are worth studying, because they tell you something about what human beings can actually sustain versus what sounds good on a manifesto.
The ones that are still here
The most straightforward examples of long-running communal experiments that are still operating in 2026:
The Hutterites are probably the most successful communal living experiment in Western history, if longevity and scale are your metrics. Founded in the 16th century during the Radical Reformation, Hutterite colonies practice complete communal ownership of property—no private possessions, no individual income, shared meals, shared labor, shared child-rearing. There are approximately 475 colonies across the northern United States and western Canada, with a total population around 50,000. They’ve been doing this for nearly 500 years. The reason nobody writes breathless magazine features about them is that they’re deeply religious, socially conservative, and not particularly interested in being studied or admired. They’re also extremely good at agriculture, which turns out to be a more durable economic base for communal living than artisanal crafts or newsletter subscriptions.
The kibbutzim in Israel represent the largest-scale secular communal experiment of the 20th century. At their peak in the 1980s, roughly 270 kibbutzim housed about 130,000 people under a model of collective ownership, shared labor, and communal child-rearing. The vast majority have since privatized—shifting to differential salaries, private property, and market-based economics—such that the classic kibbutz model now exists mainly as a historical reference point. A handful of traditional kibbutzim still practice full income-sharing, but they represent a tiny fraction of the movement. The privatization wave is itself one of the most instructive case studies in the entire history of utopian experiments: the model worked, for decades, at significant scale, and then the children and grandchildren of the founders decided they’d rather have their own stuff.
Twin Oaks in Louisa, Virginia, is probably the most frequently cited operating commune in the United States. Founded in 1967—inspired, improbably, by B.F. Skinner’s utopian novel Walden Two—Twin Oaks has about 100 members living on 450 acres, sharing income, labor, and resources. Members work a quota of roughly 42 hours per week across the community’s businesses (hammock manufacturing being the most famous) and domestic labor, and receive no individual salary. The community allocates a small personal allowance, provides housing, food, and healthcare, and makes decisions through a combination of planners and community-wide input.
Twin Oaks in 2026 is dealing with a genuinely dramatic period. In March 2024, a wildfire consumed 227 acres of the property and destroyed the building that housed its hammock business—one of the community’s primary revenue generators. The fire also brought an unexpected development: the deeply conservative surrounding community of Louisa County, which had maintained an arm’s-length relationship with Twin Oaks for decades, rallied to support the commune during the crisis. Neighbors showed up with supplies and equipment. The metaphorical wall between the commune and the county cracked. As of late 2025, Twin Oaks was rebuilding and reassessing its economic model—a process that will determine whether a community founded on mid-century behavioral psychology can adapt its revenue base after losing its signature industry to fire.
East Wind Community in the Ozark Mountains of southern Missouri was founded in 1974 and operates on roughly 1,145 acres with about 72 members. Like Twin Oaks, it’s an income-sharing egalitarian commune—members share farming, domestic work, housing, and self-governance. East Wind manufactures nut butters as its primary commercial operation. The community has attracted attention in recent years as younger people—priced out of housing markets, exhausted by gig-economy precarity, and skeptical that conventional employment will ever deliver financial stability—have started seeking out intentional communities for reasons that are more pragmatic than ideological. The New York Times described it as part of a “new generation of self-created utopias” embraced by millennials who want fewer moving parts in their lives.
Dancing Rabbit Ecovillage, also in Missouri, represents the environmentalist wing of the intentional community movement. Founded in 1997, it operates as a land trust with covenants requiring ecological sustainability—no personal vehicles, organic agriculture, renewable energy. The Foundation for Intentional Community, the movement’s main coordinating organization, is headquartered there.
Christiania in Copenhagen occupies a unique position: an 84-acre self-proclaimed autonomous neighborhood in the middle of a European capital, established in 1971 when squatters occupied an abandoned military barracks. Christiania has its own informal governance, prohibits private property ownership on its land, and has been in a continuous legal and political negotiation with the Danish government for over fifty years. In 2012, residents purchased the land from the Danish state through a collective foundation, partially resolving the ownership question while maintaining Christiania’s distinctive character as a car-free, collectively managed neighborhood that exists in a kind of negotiated autonomy with the surrounding city. It’s probably the only utopian community where you can walk to a Michelin-starred restaurant.
Auroville: The cautionary tale of 2025
Any 2026 survey of operating utopian communities has to reckon with Auroville, because Auroville is simultaneously one of the most ambitious experiments in communal living ever attempted and one of the most dramatic institutional crises in the history of intentional communities, and both of those things are happening right now.
Founded in 1968 in Tamil Nadu, India, Auroville was conceived by Mirra Alfassa—known as “the Mother,” the spiritual partner of Indian philosopher Sri Aurobindo—as a “universal township” where people of all nationalities could live in peace, devoted to the evolution of human consciousness. At its founding ceremony, 5,000 people from 124 nations gathered around a banyan tree while All India Radio broadcast Alfassa’s charter, which declared that Auroville “belongs to nobody in particular” and “belongs to humanity as a whole.” UNESCO endorsed the project. The Dalai Lama blessed it.
Over fifty years, roughly 3,000 residents from over 50 countries transformed a barren plateau into a functioning township with forests, organic farms, water systems, schools, and cultural facilities. The reforestation alone—turning eroded wasteland into thriving forest—became an internationally recognized achievement. The community was governed under the Auroville Foundation Act of 1988, which established a three-body structure: a Governing Board appointed by the Indian government, an International Advisory Council, and a Residents’ Assembly with authority over admissions and community affairs.
In late 2021, the Indian government appointed a new Governing Board and Secretary, Dr. Jayanti Ravi, who began implementing a rapid urban development plan—the “Master Plan: Perspective 2025″—by force. Bulldozers entered at night. Approximately 20,000 trees were cut. Residents were evicted with days of notice. The Residents’ Assembly’s authority over admissions was stripped. Agricultural land—Annapurna Farm, which supplied over 30 percent of Auroville’s food—was leased to IIT Madras for a truck test track over the objections of 16,000 petition signers.
In March 2025, India’s Supreme Court reversed a Madras High Court ruling that had provided some protection to residents, effectively affirming the Governing Board’s authority as Auroville’s sole administrative body. The Governing Board approved the stationing of 15 Central Reserve Police Force members—India’s largest paramilitary force—in Auroville for five years. A parliamentary committee of 30 members adopted a unanimous report in December 2025 identifying “deep flaws” in the Governing Board’s functioning, but its recommendations have not been implemented. Residents who opposed the administration faced intimidation, threatened termination from the Register of Residents, and potential expulsion.
Auroville still exists. People still live there. The banyan tree still stands. But the experiment as originally conceived—a self-governing community of international residents collaboratively building a new model of human society—is in the most severe crisis of its 58-year history. Whether it survives the current administration in any recognizable form is genuinely uncertain.
What the survivors have in common
The communities that last tend to share a few structural features that have nothing to do with the idealism of their founding documents:
A durable economic base. Hutterite agriculture, Twin Oaks hammocks (until the fire), East Wind nut butters, kibbutz farming and later light industry. Communities that depend on member donations, external grants, or ideological enthusiasm for their operating budget tend to collapse when the enthusiasm fades and the grants dry up.
Clear membership boundaries. Who’s in, who’s out, and what the process is for each. Communities with fuzzy membership—where anyone can show up and stay indefinitely—tend to attract free riders who consume resources without contributing labor, which generates resentment that kills the experiment faster than any external threat.
Governance that actually functions. Not governance that sounds beautiful in a charter, but governance that can resolve conflicts, allocate resources, and make unpopular decisions without tearing the community apart. Auroville’s crisis is fundamentally a governance failure—the three-body structure that was supposed to balance resident autonomy with institutional oversight collapsed when the institution decided to override the residents.
Willingness to evolve. The kibbutzim that survived privatized. Twin Oaks is rebuilding after fire. East Wind attracts members who are there for economic pragmatism as much as ideological conviction. The communities that insist on ideological purity tend to select for members who agree with everything and can’t adapt to anything, which is a recipe for a very pleasant five years followed by dissolution.
The honest assessment is that successful utopian communities in 2026 are small, rare, and modest in their claims. The ones that work have traded grand visions for functional systems, replaced manifestos with operational procedures, and discovered that the hardest part of building a better society isn’t imagining one—it’s doing the dishes when it’s not your turn and not resenting the person who didn’t do them yesterday.
We cover the full arc of utopian experiments—from Robert Owen’s New Harmony in 1825 through Auroville’s 2025 crisis—across our Utopian Societies course. If the question of why these experiments keep failing in the same ways is more interesting to you than the question of whether the next one will succeed, the course is built for that.
-
Landmine-Detecting Rats: How Giant Pouched Rats Are Saving Lives in Cambodia and Mozambique
A single African giant pouched rat can search an area the size of a tennis court in 30 minutes. A human deminer with a metal detector takes up to four days to cover the same ground. The rat weighs about 1.5 kilograms—too light to trigger the pressure plates on anti-personnel mines, which are typically calibrated to detonate under the weight of a human footstep. The rat doesn’t care about the rusty nails, shell casings, bottle caps, and miscellaneous scrap metal buried in every former conflict zone on earth, because it’s not detecting metal. It’s detecting the scent of TNT. When it smells explosives, it scratches at the ground, its handler marks the location, and a demolition team moves in. The rat gets a piece of banana. The mine gets destroyed. The land gets returned to the people who have been afraid to walk on it for thirty years.
This is not a thought experiment. This is a program that has been running for over two decades, has located more than 155,000 landmines and unexploded ordnances, has released nearly 86 million square meters of land back to civilian use, and has directly improved the safety of nearly six million people across seven countries. The organization behind it—APOPO, a Belgian-registered NGO whose Dutch acronym translates to “Anti-Personnel Landmines Detection Product Development”—was founded because a product design student in Antwerp watched a documentary about landmines and thought about his pet rats.
The origin story
Bart Weetjens was a graduate student at the University of Antwerp in the 1990s when the idea occurred to him. He’d kept rodents as pets since childhood and had recently read about gerbils being used as scent detectors. The connection was immediate: rats have an extraordinarily acute sense of smell—comparable to dogs in sensitivity—combined with a trainability that, while different from canine obedience, is robust enough for operant conditioning. They’re cheap to breed, cheap to feed, native to the tropics where most landmine-affected countries are located, and resistant to many endemic diseases. They can be trained in about nine months. They have a working lifespan of six to eight years.
When Weetjens proposed using trained rats as landmine detectors, the response from the demining community was roughly what you’d expect: he was laughed at for several years. The Belgian government gave him a research grant in 1997 anyway. He recruited his friend Christophe Cox—now APOPO’s CEO—and established a training and research center in Morogoro, Tanzania. The first 11 rats received accreditation under International Mine Action Standards in 2004. By 2006, APOPO’s rats had become what the organization describes as Africa’s preferred landmine countermeasure technology. By 2008, APOPO was the sole operator tasked with clearing Gaza Province in Mozambique.
The species they use—Cricetomys ansorgei, the southern giant pouched rat—is worth a moment of description, because the name “giant pouched rat” undersells both the animal and the weirdness of the whole enterprise. These are not sewer rats. They’re cat-sized, with large dark eyes, prominent whiskers, and cheek pouches they use to store food. They can weigh up to 1.4 kilograms. They’re nocturnal, social, and—according to everyone who works with them—genuinely affectionate. They climb onto shoulders. They lick their handlers. They have individual personalities and are given names: Magawa, Poppy, Peter Parker, Ronan. The APOPO visitor center in Siem Reap, Cambodia, lets tourists hold them, which reportedly divides visitors cleanly into people who love rats and people who discover they do not.
The training
Rats begin socialization at about four weeks old, handled daily so they’re comfortable around humans. Formal scent detection training starts at around five weeks. The rats are taught through clicker training—a click signals a correct response, followed by a food reward—to associate the smell of TNT with positive reinforcement. They learn to indicate a detection by pausing at the scent source and scratching at it. Training takes approximately nine months and includes progressively more complex scenarios: buried samples, outdoor environments, distracting scents, variable weather conditions.
Once certified under International Mine Action Standards, each rat works as part of an integrated team that typically includes manual deminers with metal detectors and sometimes mechanical ground preparation equipment. The rats don’t replace conventional methods. They accelerate them. Because a metal detector alerts on every piece of metal in the ground—and less than three percent of landmine-suspected land actually contains landmines—deminers spend the vast majority of their time investigating false positives. A rat that ignores scrap metal and responds only to explosive compounds eliminates most of that wasted time. APOPO’s integrated mine detection teams can triple the efficiency of a land release process compared to manual clearance alone.
The practical limitations are real. Rats can’t search reliably in thick vegetation. They work in short bursts because they overheat in tropical climates—typically 20 to 30 minutes per session. They search more erratically than human deminers, which means they offer a lower level of assurance that every square meter has been covered. They work best as a complement to other methods, not a standalone solution. APOPO is currently the only organization in the world that uses giant rats for mine detection, which tells you something about both the novelty and the niche nature of the approach.
Mozambique: The proof of concept
Mozambique was the program’s defining success. The country’s civil war, which ended in 1992, left an estimated two million landmines across the country—buried in roads, farmland, river crossings, and the areas around schools and hospitals. APOPO began operations in Mozambique in 2006, working with the government’s national demining authority. Tasked as the sole operator to clear Gaza Province, APOPO completed the work in 2012, one year ahead of schedule. The government then expanded APOPO’s mandate to Maputo, Manica, Sofala, and Tete provinces.
On September 17, 2015, Mozambique was officially declared free of all known landmines. APOPO had assisted with clearing five of the country’s most affected provinces, releasing over 13 million square meters of land. The declaration didn’t mean every mine had been found—residual clearance continued, with 16 rats maintained in-country for mop-up operations—but it represented a milestone that many in the demining community had not expected to reach on that timeline.
Cambodia: The ongoing operation
Cambodia presents a different and in some ways more challenging context. An estimated four to six million landmines were laid during the country’s decades of conflict, predominantly in the northern regions along the Thai border. Cambodia has among the highest rates of amputees per capita in the world—more than 40,000 people have lost limbs to explosive remnants of war. Agricultural land remains unusable. Communities remain displaced. The scale of the problem dwarfs what was faced in Mozambique.
APOPO began operations in Cambodia in 2015, partnering with the Cambodian Mine Action Centre in the Siem Reap area. In January 2018, the APOPO Visitor Centre opened to the public, offering guided tours, live demonstrations of mine detection by the rats, and exhibits on the science of scent detection. Tourists can watch a rat named Jordan traverse a simulated minefield, locate a buried TNT sample, and receive his banana reward. Proceeds from the $10 admission go directly back into Cambodia’s clearance program. The operation has also expanded to include HeroDOGs—trained detection dogs that complement the rats in areas where vegetation or terrain makes rat deployment less effective.
APOPO now operates across seven countries for mine action—including Angola, Zimbabwe, Colombia, and, most recently, efforts in Ukraine in response to the massive contamination from the ongoing conflict—and runs tuberculosis detection programs in Tanzania, Mozambique, and Ethiopia. The TB program uses the same scent detection methodology: rats sniff sputum samples through a glass chamber and indicate positive results by pausing and scratching. Since 2007, the TB program has evaluated hundreds of thousands of samples, identified over 13,000 tuberculosis patients who were missed by conventional microscopy at their clinics, and prevented an estimated 32,000 additional infections. In Maputo, the rats increased the TB detection rate by 40 percent.
The Magawa legacy
The most famous HeroRAT was Magawa, an African giant pouched rat who worked in Cambodia from 2016 until his retirement in June 2021. Over a five-year career, Magawa detected 71 landmines and 38 items of unexploded ordnance, clearing more than 225,000 square feet of land. In 2020, the British charity People’s Dispensary for Sick Animals awarded Magawa its gold medal for animal bravery—the first rat to receive the honor in the organization’s 77-year history. He was described as “physically strong” and exceptionally driven, and his handlers utilized him more frequently than other rats because of his consistent performance.
Magawa died in January 2022, shortly after his retirement. His legacy, as PDSA noted, “will live on for decades to come in the lives he has helped to save.” The statement is more literal than it sounds—every mine Magawa found is a mine that didn’t kill someone walking to school or working a field. The cumulative effect of 155,000 detected explosives and 86 million square meters of released land is measured not just in mines destroyed but in the ordinary, unremarkable activities—farming, walking, playing—that became possible again because a rat the size of a small cat scratched at the dirt and got a piece of banana.
The whole thing started because a guy in Antwerp liked his pet rats. Sometimes the most important innovations in the world look absolutely ridiculous from the outside, and the people who propose them get laughed at for years before the results make the laughter stop.
We cover APOPO’s HeroRATs alongside military dolphins, carrier pigeons, and a dozen other cases of animals deployed in service of human conflicts and crises across our Animal Heroes course—including why the best detection technology in a Cambodian minefield weighs less than a Chipotle burrito.
-
Drone Delivery in 2026: Why It’s Taking So Long and What Actually Has to Happen
Zipline has completed over two million commercial deliveries across 125 million autonomous miles with zero serious injuries. Wing, Alphabet’s drone delivery subsidiary, has passed 450,000 deliveries. Walmart has completed over 150,000 drone deliveries since launching the service in 2021. These are real numbers representing real packages arriving at real homes. And yet the odds that a drone will deliver your next Amazon order—or your prescription, or your burrito—remain essentially zero unless you happen to live in a handful of specific zip codes in Texas, Arizona, or a few other test markets.
The drone delivery industry in 2026 is a $1.47 billion market projected to reach somewhere between $6.7 billion and $27.5 billion by 2031, depending on which analyst you believe and how broadly they define the ecosystem. The technology works. The economics are getting closer. The regulatory framework is being built in real time. And the gap between “this demonstrably functions” and “this is available to you, specifically, right now” is still measured in years—not because any single problem is unsolvable, but because the problems stack.
The regulatory bottleneck that matters most
Every conversation about why drone delivery hasn’t scaled starts and ends with four letters: BVLOS. Beyond Visual Line of Sight. Under the FAA’s current framework, commercial drone operations generally require a human observer who can see the drone at all times. This is the regulatory equivalent of requiring a person to walk in front of every automobile with a red flag—a rule that made sense when the technology was new and makes progressively less sense as the safety record accumulates.
Without BVLOS authorization, drone delivery can’t scale. You can’t deliver packages across neighborhoods, let alone cities, if a human has to maintain eyeball contact with the aircraft for the entire flight. The companies that are actually delivering at volume—Zipline, Wing—have obtained individual BVLOS waivers from the FAA, each one negotiated separately through a cumbersome approval process. Zipline holds a waiver effective from August 2025 through August 2027 that allows operations without ground-based visual observers in the Dallas area, and following Trump’s June 2025 executive order titled “Unleashing American Drone Dominance,” the company has secured BVLOS authorization across all 50 states.
But waivers are not rules. Each one is a case-by-case approval that doesn’t automatically extend to new locations, new aircraft types, or new operators. What the industry needs—and what the FAA has been working toward—is Part 108, a proposed rulemaking that would create a permanent, standardized framework for routine BVLOS operations. Part 108 is being watched as the drone industry’s equivalent of Part 107, which in 2016 opened commercial drone operations to licensed pilots under a clear set of rules rather than individual exemptions. Part 108 would do the same for autonomous, beyond-line-of-sight flights.
The rule isn’t finalized. FAA staffing shortages have slowed the approval process. The proposed framework was announced in August 2025 and is still working through the rulemaking process. Until it’s codified, every operator expanding to a new market has to go back through the waiver system, which means the pace of expansion is gated by regulatory bandwidth rather than technological capability.
The three companies that actually matter
The competitive dynamics of drone delivery in 2026 have clarified considerably. Three operators have separated themselves from the field, and their approaches tell you almost everything about where the industry is going—and where it’s stuck.
Zipline is the company that the logistics industry points to when it needs to demonstrate that drone delivery is real. Founded in 2014, Zipline built its operation in Rwanda and Ghana delivering blood, vaccines, and medical supplies to facilities that couldn’t be reached quickly by road. The safety record—125 million autonomous miles, zero serious injuries—is the dataset that regulators and investors find compelling. Zipline’s P2 drone carries up to eight pounds, delivers within a ten-mile radius, and uses a tether system to lower packages with precision to a specific location—a porch, a table, a parking spot. The company raised over $600 million in January 2026, boosting its valuation to $7.6 billion from $5 billion in 2024. Walmart partnerships are expanding. The company was producing a new drone every hour at its manufacturing facility by end of 2025. It received a $150 million State Department contract to expand medical deliveries across five African countries.
Wing, the Alphabet subsidiary, has taken a different approach—focusing on frequent, small consumer deliveries in suburban markets. In the Dallas-Fort Worth area, customers order coffee, prescriptions, and household items through the app and receive delivery in as little as ten minutes. Wing’s drones use a hybrid design that hovers for delivery and flies like a fixed-wing aircraft for transit. The company announced a 150-store expansion with Walmart in early 2026, extending service to Los Angeles. Wing’s advantage is integration—Google’s AI infrastructure, seamless third-party app integration, and a delivery model designed around the kind of small, frequent purchases that are most expensive for traditional last-mile logistics.
Amazon Prime Air is the company that most people think of when they think “drone delivery” and is, by most operational metrics, the furthest behind. Amazon’s MK30 drone carries up to five pounds within a 7.5-mile radius and delivers within 60 minutes. The aircraft underwent 1,070 flight hours for FAA certification and was the first drone to receive BVLOS approval through the standard certification process. The ambition is enormous—Amazon has stated a target of 500 million annual deliveries by 2030.
The execution has been rough. Amazon paused drone operations in early 2025 due to altitude sensor failures caused by dusty conditions and resumed in April after software fixes. Since then, the MK30 has been involved in at least seven significant incidents: a controlled landing at an Arizona apartment complex in May, a package dropped into a swimming pool in July, two drones crashing into a construction crane in Tolleson in October—sparking a fire and hazmat response—a drone landing five feet from a resident checking his mailbox, a severed internet cable during ascent in Waco in November, and a crash into a Richardson, Texas apartment building in February 2026. The FAA and NTSB have opened multiple investigations. Amazon resumed flights within 48 hours of the crane incident and launched new markets days after the apartment building crash.
The weight differential explains a lot. Amazon’s MK30 has a maximum takeoff weight of 83 pounds. Zipline’s P2 and Wing’s drones weigh between 10 and 40 pounds. When a 15-pound drone has a problem, it’s an inconvenience. When an 83-pound drone hits an apartment building at speed, people smell smoke and watch propeller fragments fall to the sidewalk. Internal cost projections reported in late 2024 showed Amazon spending roughly $63 per delivery against customer pricing of $4.99 to $9.99. Amazon can absorb that because it’s Amazon. Whether the unit economics ever flip is an open question.
The problems that aren’t regulatory
Even if Part 108 were finalized tomorrow and every airspace question were resolved, drone delivery would still face constraints that don’t have regulatory solutions.
Noise is the first one. Drones are loud. Wing has emphasized that its aircraft are quieter than many leaf blowers, which is true and also a comparison that reveals how low the bar is. Amazon touts the MK30’s reduced noise profile. But “quieter than a leaf blower” is not “quiet,” and a neighborhood experiencing dozens or hundreds of drone flights per day is a neighborhood experiencing a new and persistent noise source. Community pushback in test markets has been real, and noise is consistently cited as the top concern.
Weather limits operations. The MK30 is unreliable in high winds, heavy rain, and snow—which describes a substantial percentage of days in most American cities. Zipline’s fixed-wing P1 is more weather-resilient but still has operational limits. No commercial delivery drone currently operates in severe weather conditions, which means the service has reliability gaps that ground-based delivery doesn’t.
Payload constraints limit the addressable market. Five to eight pounds covers a lot of consumer goods, prescriptions, and restaurant orders. It does not cover most grocery orders, large packages, or anything that weighs more than a medium-sized cat. The economics of drone delivery work best for small, high-urgency items—medications, missing ingredients, last-minute purchases—not for the bulk of e-commerce volume.
Airspace integration remains unsolved at scale. Individual operators can manage their own fleets with proprietary software, but as drone traffic grows, integration with traditional air traffic control and future urban air mobility services—air taxis, emergency medical flights—requires interoperable unmanned traffic management systems that don’t exist yet in standardized form. Zipline’s FAA-approved airspace management system is an early example, but the infrastructure for managing thousands of simultaneous autonomous flights over a metropolitan area hasn’t been built.
Where this actually goes
The honest trajectory for drone delivery is not the one that any company’s investor deck shows. It’s not 500 million deliveries by 2030. It’s not a replacement for ground-based logistics. It’s a specific tool for specific use cases—medical supplies in areas with poor road infrastructure, urgent small-package delivery in suburban markets, and high-frequency low-weight consumer goods in neighborhoods where the economics and regulatory approvals align.
The technology works. Zipline has proved that comprehensively. The regulatory framework is being built, slowly, by an FAA that is understaffed and cautious—appropriately so, given that these are autonomous aircraft operating over populated areas. The public acceptance question is real and varies enormously by community. And the economics are viable for niche applications today, with the potential to improve as production scales, battery technology advances, and operational density increases.
What drone delivery is not, in 2026, is imminent for most people. The companies doing this well are doing it carefully, in selected markets, with specific use cases. The company doing it fastest is also the one crashing into apartment buildings. There’s probably a lesson in that.
We cover drone delivery technology, autonomous navigation, and the regulatory landscape alongside the full humanoid robotics industry in our Humanoid Robots & Drones course—including why the companies that are actually scaling are the ones that started with blood deliveries in Rwanda, not same-day retail in suburban Texas.
-
The Uncanny Valley Problem: Why Humanoid Robots Are So Unsettling (And Whether It Matters)
In January 2026, a team at Columbia Engineering published a paper in Science Robotics announcing that they’d built a robot capable of learning realistic lip movements by watching its own reflection and studying human videos. The robot could speak in multiple languages and sing. Hod Lipson, the lab’s director, framed the significance plainly: “There is no future where all these humanoid robots don’t have a face. And when they finally have a face, they will need to move their eyes and lips properly, or they will forever remain uncanny.” His co-researcher Yuhang Hu added: “We are close to crossing the uncanny valley.”
That phrase—uncanny valley—has been floating around robotics and cognitive science for over fifty years, and it describes a problem that gets more commercially urgent every quarter. Tesla, Figure AI, Agility Robotics, Boston Dynamics, and a growing roster of companies are building humanoid robots designed to operate in spaces built for humans—warehouses, hospitals, factories, homes. The machines are getting better at walking, grasping, navigating, and following instructions. The question of whether people will actually want to be around them is a different engineering challenge entirely, and it’s one that can’t be solved with better actuators.
The graph that launched a thousand nightmares
Japanese roboticist Masahiro Mori proposed the concept in 1970 in an essay for the journal Energy. The idea is deceptively simple: as a robot becomes more human-like in appearance, people’s emotional response becomes more positive—up to a point. A cartoon robot is charming. A robot with a humanoid shape and some facial features is engaging. But as the resemblance approaches near-human levels without quite getting there, the response doesn’t just plateau. It collapses. The emotional curve drops into a trough of revulsion, unease, and what Mori called bukimi—a Japanese word that translates roughly to “eeriness.” That trough is the uncanny valley.
Mori’s original graph wasn’t based on experiments. It was based on his personal observations and intuitions, which is a detail that tends to get lost in the retelling. The essay was more philosophical provocation than empirical finding, and it sat in relative obscurity for decades before roboticists and CGI animators in the 2000s rediscovered it and realized they’d been stumbling into the valley independently. The 2004 Robert Zemeckis film The Polar Express—in which Tom Hanks was motion-captured into a CGI character that audiences found deeply unsettling despite technically impressive animation—became the canonical example of uncanny valley in popular culture. The technology was extraordinary. The result was a children’s movie that gave children nightmares.
What makes the uncanny valley genuinely interesting, rather than just a fun piece of trivia to reference when a new humanoid robot demo goes viral, is that nobody fully agrees on why it happens. And the competing explanations have very different implications for whether it can be solved.
The competing theories
The most commonly cited explanation is perceptual mismatch—the idea that human brains are optimized over millions of years of evolution to process human faces with extraordinary precision, and when something is close to human but slightly off, the discrepancy triggers an error signal. We’re wired to detect subtle abnormalities in faces because historically, detecting disease, deception, or unfamiliarity in the people around you was a survival advantage. A robot that’s 95 percent human-looking trips the same alarm system that would fire if you encountered a person with something wrong with them—asymmetrical facial movement, delayed eye tracking, a smile that doesn’t reach the upper face. The problem isn’t that the robot looks bad. The problem is that it looks almost right, and the remaining five percent registers as pathological.
A second explanation focuses on categorical ambiguity. Things that sit cleanly in one category—”obviously a machine” or “obviously a human”—are psychologically comfortable because the brain knows what schema to apply. Things that fall between categories—not quite machine, not quite human—create cognitive dissonance. You don’t know whether to treat it as an object or a person, and that uncertainty is inherently aversive. This explanation has roots in psychological research on disgust responses to category violations more broadly—the same mechanism that makes chimeric creatures in horror films effective.
A third theory, supported by a 2021 study published in Computers in Human Behavior, argues that the uncanny valley is driven specifically by the perception that a robot has feelings. Researchers found that humanoid robots are unsettling in part because people unconsciously attribute the capacity for subjective experience—what philosophers call phenomenal consciousness—to things that look human. When you look at a robot face and your brain automatically assumes something is going on behind those eyes, and then another part of your brain recognizes that nothing actually is, the collision between those two assessments produces the eerie feeling. The study demonstrated that “dehumanizing” a humanoid robot—explicitly telling people it has no feelings—significantly reduced uncanny valley responses, and this effect held up in a field study with hotel guests interacting with real robots in Japan.
This third theory is the one with the most interesting commercial implications, because it suggests the uncanny valley isn’t purely a design problem. It’s a framing problem.
What the latest research shows
A February 2025 study from researchers working with Nadine, a hyper-realistic humanoid robot, tested whether equipping a robot with LLM-powered conversational abilities could reduce uncanny valley effects. Eighty participants interacted with the robot and completed pre- and post-interaction surveys. The findings were promising and limited in roughly equal measure: LLM-enhanced conversations significantly reduced feelings of eeriness and increased perceptions of pleasantness and approachability. But—and this is the part that matters for anyone trying to commercialize these things—a subset of users continued to report discomfort even after extended, high-quality conversation. The uncanny valley effect shrank. It didn’t disappear.
The regression analysis revealed something counterintuitive: conversational naturalness and interestingness predicted willingness to continue interacting, but visual human-likeness did not. In other words, once the conversation was good enough, how human the robot looked stopped being a significant factor in whether people wanted to keep talking to it. The implications of that finding, if it replicates, are substantial—it suggests that investing in better AI conversation may be a more efficient path past the uncanny valley than investing in more realistic skin texture.
The Columbia Engineering lip-movement paper approaches the same problem from the opposite direction. Their argument is that facial expression—particularly lip synchronization during speech—is so fundamental to human communication that getting it right is non-negotiable. Nearly half of human attention during face-to-face conversation is directed at the speaker’s lips. A robot whose mouth movements don’t match its speech triggers uncanny valley responses even if everything else looks perfect, because the mismatch between what you hear and what you see is processed as a deep wrongness. Their robot learned lip movements through self-supervised learning—watching itself in a mirror and comparing its movements to human video—rather than being explicitly programmed. The result was lip synchronization natural enough that Lipson, a self-described “jaded roboticist,” reported involuntarily smiling back at his own creation.
The design choice nobody talks about
Here’s the thing about the uncanny valley that tends to get overlooked in favor of the more dramatic question of “can we cross it?”: most of the companies actually building commercial humanoid robots have decided not to try.
Look at the design language of the machines that are closest to deployment. Boston Dynamics’ Atlas has no face at all—it’s a headless torso with extraordinary agility. Agility Robotics’ Digit has a face that is deliberately stylized and cartoonish—two large “eyes” on a rounded head that reads unmistakably as friendly and unmistakably as not human. Figure’s robots have a visor-like head that evokes a motorcycle helmet more than a human face. Tesla’s Optimus has gone through several iterations, and each one has moved further from human facial features toward a smooth, featureless faceplate.
These aren’t failures of engineering ambition. They’re deliberate design decisions informed by the uncanny valley research. If your robot’s face can’t be indistinguishable from a human face—and no current robot face can—then the safest design choice is to make it clearly, obviously, comfortably non-human. Stay on the left side of the valley. Don’t attempt the crossing. The cartoon robot, the friendly geometric face, the abstract visor—these designs maintain positive emotional responses without risking the plunge into eeriness.
The companies pursuing hyper-realistic human faces—Hanson Robotics (creators of Sophia), Hiroshi Ishiguro’s Geminoid series—tend to be research labs and publicity vehicles rather than commercial deployment operations. Sophia is famous, but Sophia isn’t picking items in a warehouse or delivering medication in a hospital. The robots that are actually being deployed at scale are the ones that look like robots, and that’s not a coincidence.
Whether any of this matters
The pragmatic argument—increasingly popular among robotics engineers who are tired of the uncanny valley being treated as an unsolved existential crisis—is that the entire question is overrated for most commercial applications. A warehouse robot doesn’t need a face. A manufacturing robot doesn’t need to be likable. Even in healthcare and hospitality, where robots interact directly with humans, the evidence suggests that functional competence matters more than facial realism. The Japanese hotel study found that guests’ satisfaction with robotic service was unaffected by whether the robot was framed as human-like or machine-like, as long as it performed its job.
The counterargument—and it’s a serious one—is that the applications where the uncanny valley matters most are precisely the applications where humanoid robots would generate the most value. Eldercare. Companionship. Therapy. Education. Customer-facing roles that require trust, rapport, and emotional connection. If you want a robot that an elderly person feels comfortable having in their home, or that a child feels safe learning from, or that a patient trusts to assist with rehabilitation, the emotional response to its face is not a trivial consideration. It’s the consideration.
The honest answer to “will we cross the uncanny valley?” is probably “yes, eventually, for robots that are sufficiently expensive and specifically designed for contexts where crossing matters.” The Columbia lip-movement research, the LLM conversation studies, and advances in silicone skin and micro-actuator facial expression systems are all pushing the right side of the valley upward. But for the next decade of commercial humanoid robotics, the dominant strategy will almost certainly be avoidance rather than crossing—designing robots that are useful, functional, and friendly without pretending to be something they aren’t. The uncanny valley is a real phenomenon with real psychological mechanisms behind it. The most practical response, for now, is to respect it rather than try to solve it.
We cover the uncanny valley alongside the mechanical engineering, AI integration, and commercial deployment of humanoid robots across 24 lectures in our Humanoid Robots & Drones course—including why the companies spending billions on robot bodies are making very specific choices about what those bodies look like.
