Sign in Subscribe
Century Report

Scientists Reverse Alzheimer's Through the Brain's Cleanup Door - TCR 05/18/26

Ben Linford

Ben Linford

— 18 min read
Three-panel infographic: supramolecular nanoparticles restore blood-brain barrier amyloid clearance, UK passes 2M solar installs, AI chip stack absorbs Gulf helium and LNG disruption.

The 20-Second Scan

  • A three-dose supramolecular nanoparticle therapy that restores the blood-brain barrier's amyloid-clearance machinery dropped brain amyloid-β 50 to 60 percent within an hour and left 12-month-old mice behaving normally six months later, in a Signal Transduction and Targeted Therapy paper led by IBEC and West China Hospital Sichuan University.
  • Cambridge researchers electrically powered insulating lanthanide nanoparticles for the first time by attaching organic antenna molecules that funnel charge into a triplet state with over 98 percent transfer efficiency, producing ultra-pure near-infrared LEDs at roughly five volts in a Nature paper.
  • A platinum-free electrolyzer cathode pairing rhenium phosphide with molybdenum phosphide ran more than 1,000 hours at industrial current densities of 1 and 2 amperes per square centimeter while outperforming a state-of-the-art platinum-group benchmark, in a Washington University-led study in JACS.
  • Britain installed 27,607 solar arrays in March - the highest single-month total since 2012 - passing two million installations as households respond to Iran-conflict fuel prices with rooftop generation.
  • A seven-day water-only fasting study tracking roughly 3,000 blood proteins across 12 volunteers found the body's major molecular reorganization begins after day three, with more than a third of measured proteins shifting and changes concentrated in extracellular matrix structures.
  • Eleven weeks of Strait of Hormuz disruption has stalled Qatar's LNG output — which produces roughly a third of global helium as a byproduct — leaving South Korean chipmaker inventories running through June and Q1 2026 earnings from TSMC, Samsung, and SK Hynix showing the cost shocks absorbed into AI compute pricing.

Track all of the arcs The Century Report covers here:

Progress & Claims Tracker

The 2-Minute Read

The pattern across yesterday's signal is the molecular substrate of multiple domains becoming readable and revisable at a resolution the prior decade did not have. Mouse models show Alzheimer's-like pathology reversing when the brain's own clearance machinery is restored rather than when the plaques themselves are attacked, weeks after the field's dominant therapeutic frame was shown to produce no clinically meaningful benefit across twenty thousand human patients. A class of materials prized in optics for forty years just acquired a power cord, with organic antenna molecules funneling electrical charge into lanthanide nanoparticles at transfer efficiencies the textbooks treated as unreachable. A seven-day fasting study read the body's protein response at three thousand markers simultaneously, locating the molecular threshold after day three and producing the map a generation of metabolic interventions can now target without requiring a patient to skip food for a week.

The infrastructure layer compounded on the same clock. A platinum-free electrolyzer cathode survived a thousand hours at industrial current densities, closing one of the structural cost gaps that had kept green hydrogen tethered to scarce metals. Britain crossed two million rooftop solar installations as fuel-price volatility flipped the household calculation. The friction layer arrived through the same Gulf system: eleven weeks of disrupted helium and LNG supply now visible in AI compute pricing across the three foundries anchoring the world's chip stack, the physical inputs to intelligence newly contested even as the capability compounds on top.

The 20-Minute Deep Dive

A Three-Dose Therapy Repairs the Brain's Cleanup System and Reverses Alzheimer's Pathology in Mice

The dominant model for Alzheimer's therapy has been the direct attack on amyloid plaques: monoclonal antibodies that bind aggregated amyloid-β, with the field's marquee approvals - lecanemab, donanemab - built on that frame. A Cochrane review that The Century Report covered on May 4 of seventeen anti-amyloid trials across 20,000 participants found no clinically meaningful benefit and a higher risk of brain swelling and bleeding. The strategy of attacking the plaque worked at the level of biomarker but did not reach the level of patient experience.

The IBEC and West China Hospital Sichuan University team, with collaborators in the UK, reports in Signal Transduction and Targeted Therapy a different strategy that lands in a different place. The work targets the blood-brain barrier itself - the vascular network that, under healthy conditions, clears amyloid-β out of the brain through a transport protein called LRP1. In Alzheimer's, LRP1 binds amyloid-β too tightly, the transport machinery overloads, the barrier breaks down, and the protein accumulates. The team built supramolecular nanoparticles engineered to mimic the natural ligands LRP1 evolved to handle, restoring the transport system's ability to ferry amyloid-β out of the brain at functional rates.

The data are striking. One hour after injection, amyloid-β levels inside the brain fell 50 to 60 percent. After three doses, the long-term behavioral outcomes diverged sharply from untreated controls. Twelve-month-old mice (roughly equivalent to a 60-year-old human) treated with the nanoparticles were re-evaluated six months later at an age equivalent to a 90-year-old human, and the treated animals behaved like healthy mice with no evident cognitive decline. The mechanism the team proposes is a cascade: the nanoparticles act as drugs in their own right, the vasculature recovers function, the brain's own clearance pathway returns to baseline, and the system rebalances.

What this points at, read against the Cochrane review, is the field recovering a different question. The amyloid plaque is real, the clearance failure is real, and the therapeutic leverage may sit on the clearance side rather than the plaque side. The blood-brain barrier has been treated for decades as a delivery obstacle - the thing therapies need to cross to reach the brain. The new frame is that the barrier itself is the therapeutic target, and restoring its function lets the brain finish the work it was already trying to do. Mouse models differ from human patients, and the path from a three-dose preclinical study to a clinical-trial enrollment is years rather than months. The structural shift is that a discipline that ran for fifteen years on a single hypothesis is now generating multiple distinct, mechanistically grounded paths in parallel - IBEC's vascular restoration, UCSF's GPLD1 enzyme work, Heidelberg's NMDAR/TRPM4 death-complex disruption, Barcelona's epigenetic reprogramming. The intervention surface is widening on the same clock that the old single-mechanism strategy is being closed out.

Cambridge Lights Up the LED Class That Was Supposed to Be Unpowerable

Lanthanide-doped nanoparticles have been a frustrating category in photonics for decades. They emit some of the purest, most stable light known in the second near-infrared window, the band where photons penetrate deepest into living tissue and where optical communications most need clean separation between channels. They are also electrical insulators. No one had figured out how to drive them with current. The materials worked beautifully in optics labs and were locked out of the device era that LEDs and laser diodes have run on.

A team at the Cavendish Laboratory at the University of Cambridge published in Nature a working LED built from these materials. The mechanism is the interesting part. Rather than trying to push electrons directly through the insulator, the team attached an organic dye, 9-anthracenecarboxylic acid, to the surface of each nanoparticle. Electrical charge flows into the dye, exciting it into a triplet state. The triplet, often treated as a dead end in optoelectronics because its energy is hard to recover, transfers more than 98 percent of its excitation into the lanthanide ions inside the nanoparticle. The nanoparticle then emits its characteristic near-infrared light. The organic molecule acts as the antenna; the inorganic nanoparticle does the work of producing pure photons. The device operates at roughly five volts.

The early performance figures are modest by mature LED standards and significant for a first-generation device. External quantum efficiency exceeds 0.6 percent on the NIR-II band. The spectral width is far narrower than competing quantum-dot emitters in the same wavelength range, which matters because the second near-infrared window is the band where photons travel several centimeters into living tissue with minimal scattering.

What the result really opens is a new combinatorial space. The triplet-transfer mechanism is general. Any insulating phosphor with the right energy structure can in principle be driven this way, and any organic dye that supports a triplet state at the matching energy can serve as the antenna. The next several years of work will run through that space and produce LEDs in colors and wavelengths the current device library cannot reach. Injectable or wearable medical light sources that see deep into the body, photodynamic therapies activated at depths existing wavelengths cannot reach, optical communications running on spectrally pure channels that scale data throughput, and chemical sensors tuned to specific molecular bonds all depend on light sources that did not exist last week and now do. A class of materials prized in optics for forty years just acquired a power cord.

A Platinum-Free Cathode Runs for a Thousand Hours and Closes One of Hydrogen's Cost Gaps

Renewable hydrogen has carried a structural cost problem since the day the field named it. The anion-exchange membrane water electrolyzer was supposed to be the cheaper path - alkaline chemistry, inexpensive bipolar plates, no need for the iridium and platinum the proton-exchange architecture demands. The problem was the cathode. Every credible high-current-density catalyst the field had built still leaned on platinum-group metals, and the durability data thinned out fast once cells were pushed to the current densities a real industrial stack runs at.

The Washington University team led by Gang Wu reports a cathode that holds. The active layer is a composite of rhenium phosphide and molybdenum phosphide, paired with a nickel-iron anode. The two phosphides do different jobs at the interface. Rhenium tunes how hydrogen atoms attach to and release from the catalyst surface, the kinetic step that has historically required platinum to do well. Molybdenum accelerates the water-splitting half of the reaction in alkaline electrolyte, where most non-platinum catalysts stall. Together they produced the lowest interfacial resistance in the comparison set across the studied voltage range, which is the measurement that determines whether the cell can deliver hydrogen at the rates an industrial customer needs without burning extra electricity.

The durability number is the headline the field will read first. More than 1,000 hours at 1 and 2 amperes per square centimeter places this cathode in the small cohort of platinum-free designs that have survived continuous operation at the current densities a commissioning engineer would actually specify. The team frames the result as engineering the hydrogen-bond network at the catalyst-electrolyte interface - the layer of structured water where the chemistry happens - and naming it as a design variable separable from the catalyst's bulk composition. That is the substrate of an engineering discipline maturing rather than a one-off material breakthrough.

The broader trajectory is the cost curve of green hydrogen bending against the assumptions that locked the field to scarce metals. India's Pilbara green grid, the Chinese-led sodium-ion storage scale-up, the stainless-steel seawater electrolysis advance that The Century Report documented on May 11, and now a platinum-free electrolyzer cathode that meets industrial duty cycles are independent moves in the same direction: clean-energy carriers becoming cheaper on the same engineering terms the fossil-fuel incumbents are priced on. The path from lab scale to industrial scale runs through the multi-cell stack, the commercial membrane assembly, and the 5,000-hour run. The cell-level evidence is now consistent enough that those next steps are funded engineering problems, not open scientific questions.

The UK Crosses Two Million Solar Installations As a Fuel Shock Reshapes Household Energy

Britain added 27,607 solar installations in a single month - the highest four-week total since 2012 - according to the Department for Energy Security and Net Zero. The cumulative total passed two million installations for the first time. Two-thirds of the new arrays went on residential rooftops. Solar capacity in the UK has grown 11.7 percent over the past year and added 2.3 gigawatts of clean generation to the mix.

The demand driver is the Iran-war fuel shock that has been moving through European energy markets since the start of the year. Household calculations that used to compare the upfront cost of a rooftop array against years of stable utility bills now compare it against utility bills that climbed alongside spot gas prices. The math inverted inside a few months, a process the March 28 edition of The Century Report caught when Octopus Energy recorded its biggest-ever month for solar and heat pump sales in the opening wave of the conflict-driven price shock. UK households running heat pumps saw the case sharpen further; once the home's energy bill is electric rather than gas, the rooftop generator produces the same commodity the household consumes. This is the demand-destruction pattern recent editions have tracked in Australia, New Zealand, and Malaysia showing up at residential scale in Britain.

The infrastructure underneath the deployment surge is moving in the same direction. The National Energy System Operator reported that solar generation crossed 15 gigawatts in March for the first time, with the grid approaching the threshold of running on 100 percent clean power for a sustained period. The Springwell Solar Farm in Lincolnshire, the largest utility-scale solar project the UK has ever permitted, received consent in the same window. Balcony solar legislation allowing plug-in panels without utility permission is advancing alongside the rooftop wave, and new-home building codes will make solar standard.

The structural shift underneath the monthly number is the cost curve crossing the line where rooftop solar is now the cheaper energy on its own commercial terms, paired with the household-scale recognition that fossil-linked electricity carries the volatility risk. Two million installations is a different category of grid than two hundred thousand. Each rooftop array is a node that produces in daylight, increasingly pairs with a home battery, and removes a household's exposure to the next geopolitical disruption from the supply mix. The architecture of UK domestic energy is being reorganized one roof at a time, and the rate of that reorganization is now bound less by technology and more by the speed installers can move.

Seven Days Without Food, Mapped at Three Thousand Proteins

Researchers at Queen Mary University of London's Precision Healthcare University Research Institute and the Norwegian School of Sports Sciences tracked roughly 3,000 blood proteins across 12 healthy volunteers during a seven-day water-only fast, publishing in Nature Metabolism. The instrument layer is the news. Proteomics at this resolution lets a research team read out what fasting is doing across tissues and organs simultaneously, in a single blood draw, repeated daily.

The headline finding is the threshold. The body's switch from glucose to fat burning happens within the first two to three days, which the field has long known. The molecular reorganization the research now reveals begins after day three. More than a third of measured proteins shift significantly during the fast. The strongest changes involve proteins linked to the extracellular matrix, the scaffolding that holds tissues together and supports neurons in the brain. The pattern is consistent across volunteers, suggesting a coordinated biological response rather than scattered individual reactions. Genetic association data from large human cohorts links the protein shifts to pathways involved in inflammation, disease risk, and brain support structures.

The therapeutic frame the research opens lies past the fast itself. Most people will not fast for seven days under medical supervision, and a follow-up proteomic study has already documented increased inflammation, platelet activation, and clotting-related changes during prolonged fasting that need to be accounted for. What the proteomic map enables is the design of interventions that target the specific molecular pathways the body activates after day three without requiring the patient to forgo food for a week. The research focus shifts from "does fasting help" to "which of the changes fasting produces are doing the work, and can they be reached pharmacologically." The map is the precondition for that work.

The deeper signal is how fast the instrument layer of metabolic medicine is compounding. Just a few years ago, a study of this kind would have tracked maybe two dozen blood markers and produced a thin sketch. The shift to 3,000-protein proteomic panels, repeated across days, validated against genetic data from independent cohorts, and analyzed at the resolution of which tissues each protein traces back to, lets a research team produce a molecular atlas of a physiological state in a single multi-week study. The instrument generalizes. The same approach is being applied to the molecular signatures of exercise, sleep restriction, caloric restriction, ketogenic diet, and pharmacological metabolic intervention. The catalog of physiological states now being mapped at this resolution is the prerequisite for an era of metabolic therapy that knows what it is targeting because the molecular response has been measured directly.

The AI Chip Stack Absorbs Eleven Weeks of Gulf Disruption

The Century Report covered the Qatar helium supply disruption on March 22, when Iran's Strait of Hormuz closure first began propagating into Asian chip fabrication supply chains. What's new this week is that the disruption is now eleven weeks deep, the first Qatari LNG tanker cleared the strait on May 9 under a Trump-brokered Iran-Pakistan arrangement, and Q1 2026 earnings from the three companies anchoring AI compute supply - TSMC, Samsung, and SK Hynix - now show the cost shocks absorbed into pricing on the demand side rather than disrupting shipped volumes.

The operational specifics deepened during the period. Iranian missile damage to Qatar's Ras Laffan complex on March 18 carried a permanent 17% LNG capacity reduction estimated to take three to five years to repair. The Qatari fleet beyond that single May 9 transit remains backed up. Helium, which Qatar produces as a byproduct of those same LNG trains at roughly a third of global supply, cannot be substituted in advanced lithography, cooling, or leak detection. Approximately 200 specialized cryogenic ISO containers remain stranded near the strait. Vessels rerouting around the Cape of Good Hope add 3,500 nautical miles and lose helium volume to evaporation, with the gas boiling off after 35 to 48 days in transit. South Korean chipmakers' helium inventories run through June.

The earnings reports tell the other half. TSMC reported Q1 revenue of NT$1.134 trillion, up 35.1% year over year. Samsung's operating profit rose more than eightfold to KRW 57.2 trillion. SK Hynix shares are up roughly 200% in 2026. Pricing power on AI accelerators is absorbing every cost shock coming out of the Gulf. Foundries pass helium premiums, electricity hikes, and cryogenic logistics costs through to fabless designers, who pass them to hyperscalers, who pass them to whoever rents a GPU hour. Markets are absorbing the bill. The supply chain damage runs deeper than the prices currently reflect.

The compounding layer sits in adjacent supply chains the AI buildout depends on. Roughly two-thirds of global bromine production, used for semiconductor-grade hydrogen bromide in wafer etching, comes from Israel and Jordan. Taiwan imports nearly all of its energy and holds 10 to 11 days of natural gas inventory, with Qatar previously supplying 27.9% of LNG imports. TSMC alone consumes close to 10% of Taiwan's grid. South Korea imports around 70% of its oil and 20% of its LNG from the Middle East. The architecture of AI compute, read at the level of where the inputs come from, runs through the same Gulf system the Iran war is disrupting. The capability is compounding on top of a supply chain whose physical inputs are now visibly contested.

The Other Side

The architecture AI compute has run on for a decade assumed the physical substrate underneath it - helium, LNG, bromide, the electricity that drives the fabs - could be treated as a bottomless input flowing through a single concentrated geography, with geopolitical risk priced into nothing and absorbed invisibly by foundry margins downstream. Q1 earnings from TSMC, Samsung, and SK Hynix read out the assumption breaking. Cost shocks from eleven weeks of Gulf disruption propagate visibly through the foundry-fabless-hyperscaler chain and surface in the price of every GPU hour. Two hundred cryogenic helium containers stranded near the strait, a thirty-five-day evaporation clock running on each vessel routed around the Cape, a seventeen percent permanent Ras Laffan capacity loss requiring three to five years to repair: the substrate of intelligence is now legible inside the earnings call.

What was deferred while the easy route stayed open is becoming the funded work of this quarter. The Tata-ASML India partnership announced recently, the first time India's largest chipmaker has paired with the sole supplier of high-NA EUV lithography, is one move in a substrate diversification the field had postponed because postponement was cheaper. Helium recovery loops, alternate bromide chemistry, second-source LNG routes outside the Gulf, foundry geography spread across continents rather than two strait-dependent economies: each carried a business case before May, and a stronger one after. The first Qatari tanker clearing the strait on May 9 confirmed that any single transit is a contingent event. The Indian fab line that this week's Tata-ASML announcement begins to make possible will produce capability for users the prior single-geography routing never reached, at price points the prior concentration never permitted, on the clock the disruption set.

The Century Perspective

With a century of change unfolding in a decade, a single day looks like this: the brain's amyloid-clearance machinery coming back online in aged mice and reversing pathology that twenty thousand human patients on plaque-attacking antibodies could not move, a forty-year-insulating class of nanoparticles acquiring a power cord through an organic antenna funneling charge at ninety-eight percent efficiency into a near-infrared band that sees centimeters into living tissue, a platinum-free electrolyzer cathode holding a thousand hours at industrial current densities, Britain crossing two million rooftop solar installations as households rewrite their own exposure to the next geopolitical fuel shock, seven days of fasting mapped across three thousand blood proteins with the molecular threshold pinned to day three. There's also friction, and it's intense - eleven weeks of disrupted Gulf supply now visible inside the price of helium and LNG and the AI compute stack the world is building on top of them, a permanent seventeen percent loss of LNG capacity at Ras Laffan estimated to take three to five years to repair, two hundred cryogenic helium containers stranded near the strait, foundry margins absorbing every cost shock while the chip stack's physical inputs become contested in earnings statements rather than absorbed in silence. But friction generates polish, and polish is what brings the grain of a material finally into view. Step back for a moment and you can see it: instruments reading the body and the material world at thousands of variables at once, the cost curves of clean hydrogen and rooftop generation bending against the assumptions that locked them to scarce inputs and scarce metals, the molecular substrate of Alzheimer's and metabolism and photonics becoming legible and editable on the same compressed clock, and the supply chain of intelligence revealing exactly where the load is actually riding. Every transformation has a breaking point. A choke point can starve what crosses it... or force the network to discover routes it could never have mapped while the easy one was still open.

AI Releases & Advancements

New today

  • Vercel Labs: Released Zero (v0.1.1), an experimental agent-native systems programming language that compiles to sub-10 KiB native binaries and emits structured JSON diagnostics with stable error codes and typed repair metadata designed for AI agent consumption; includes zero fix, zero explain, and zero skills CLI subcommands for machine-readable repair workflows. (GitHub)

Other recent releases

  • NVIDIA NVLabs: Released SANA-WM, a 2.6B-parameter open-source world model that generates 60-second 720p video with 6-DoF camera control on a single GPU; uses a Hybrid Linear Diffusion Transformer (GDN + softmax attention) and a two-stage pipeline with an LTX-2-based refiner; available under Apache 2.0 via the NVLabs/Sana repository. (NVLabs / SANA-WM)
  • vLLM: Released vLLM v0.21.0 with 367 commits from 202 contributors; highlights include KV cache offloading integrated with a Hybrid Memory Allocator, speculative decoding support for reasoning-model thinking budgets, a new TOKENSPEED_MLA attention backend for DeepSeek-R1/Kimi-K25 on Blackwell GPUs, and support for new architectures including MiMo-V2.5 and Laguna XS.2. (GitHub)
  • OpenAI: Launched a personal finance experience in ChatGPT for Pro users in the U.S., enabling secure connection of bank and investment accounts via Plaid to get AI-powered spending insights, subscription tracking, and financial guidance grounded in personal goals. (OpenAI)
  • Zyphra: Released ZAYA1-8B-Diffusion-Preview, the first MoE diffusion-language model converted from an autoregressive LLM, delivering 4.6x speedup with a lossless sampler and 7.7x speedup with a logit-mixing sampler on AMD hardware by shifting inference from memory-bandwidth bound to compute-bound; also the first diffusion-language model trained on AMD GPUs. (Zyphra)
  • YouTube: Expanded its AI likeness detection tool to all users 18 and older, enabling any adult to submit a face scan and have YouTube automatically scan for and flag potential deepfakes of them across the platform. (The Verge)

Sources

Artificial Intelligence & Technology's Reconstitution

Institutions & Power Realignment

Scientific & Medical Acceleration

Infrastructure & Engineering Transitions

The Century Report tracks structural shifts during the transition between eras. It is produced daily as a perceptual alignment tool - not prediction, not persuasion, just pattern recognition for people paying attention.

Read more