Karl Pfleger argues aging therapies should be judged by multi‑disease benefit, not lifespan

On 18 September 2026, investor Karl Pfleger responded to Professor João Pedro de Magalhães’ question about how to judge anti‑aging therapies if not by lifespan. Pfleger proposed an alternative criterion and noted that the only organization testing such combinations in practice, the LEV Foundation, ran an experiment where two of three approaches failed and a larger follow‑up remains unfunded due to lack of money.

On 9 August 2026, de Magalhães wrote that humanity is not approaching victory over aging, citing that over the past century the only reliable way to extend mouse lifespan has been caloric restriction. Pfleger replied on 10 August that measuring progress by a single best therapy is mistaken, because aging comprises at least seven distinct molecular breakdowns—such as senescent‑cell accumulation—that require different treatments, just as different cancers need different drugs.

On the same day, de Magalhães disagreed, insisting lifespan remains the only reliable metric; otherwise symptomatic relief could be mistaken for anti‑aging effect. On 7 September Pfleger countered with an analogy to the 1960 Moon landing, arguing that demanding immediate success ignores the needed development time. On 17 September de Magalhães acknowledged the need for combination therapies but asked how to prove they work—or don’t.

The next day Pfleger answered: a therapy validates the “geroscience hypothesis” if it treats, alleviates, or prevents several age‑related diseases at once, even if it does not extend life. Individual interventions may show no effect alone; benefit appears only when combined. He cited the LEV Foundation’s RMR1 experiment on 1,000 mice, which showed a single dose of repair does not last longer than one year. Of the three interventions, two were defective—one “broken,” the other “partially broken.” A second, larger trial with eight interventions still seeks funding.

Pfleger lamented that only one group is testing such combinations, calling it regrettable, and noted that structural incentives discou

🔗 Read original →

Food and anti-parasitic treatment slow epigenetic aging in wild mice

Scientists experimentally tested on wild woodland mice that feeding and anti‑parasite treatment slow the epigenetic clocks that measure aging via DNA methylation marks. A team from Edinburgh and Altos Labs built DNA‑methylation clocks for the woodland mouse and transferred them to a wild population near Edinburgh. In the September 16 version of the study, mice that were simultaneously fed and treated for intestinal parasites showed a significantly slower clock rate than untreated animals of the same chronological age.

Wild animals’ ages are rarely known precisely because they are not caught at birth; body mass, tooth wear or telomere length give only rough estimates or require killing the animal. Epigenetic clocks solve this: with age, methylation changes at specific DNA sites predict age. Sarah Wolf and colleagues first trained the clocks on 74 laboratory woodland mice of known age using 177 such marks. The clocks predicted age with an error of 5.45 days on the training set and 14.64 days on independent validation. This is an order of magnitude more accurate than universal pan‑mammalian clocks trained on 18

🔗 Read original →

NUS Reveals PROMETHEUS Pilot Protocol Personalizing Anti‑Aging Supplements by Gerotype

On 18 September Professor Andrea Meyer presented the protocol and eight‑week results of the PROMETHEUS trial at the Academy of Healthy Longevity webinar, National University of Singapore. The study is one of the finalists in the XPRIZE Healthspan competition.

Gerotype combines genomic, epigenetic, microbiome and clinical data to capture uneven organ ageing; it determines both who is enrolled and which supplements each participant receives. Participants aged 50–80 received a uniform base package of sleep, nutrition, whey protein, creatine, counselling and combined physical‑cognitive training.

Supplements were added only when a specific ageing biomarker fell below the 50th percentile for age and sex: urolithin A for low muscle strength, nicotinamide mononucleotide for low endurance, and a multivitamin for weak cognitive performance. At week 4 all measures were repeated and doses were adjusted or supplements changed if there was no response, mirroring clinical drug titration.

Enrollees were drawn from the lower 75% of endurance and NIH Toolbox cognition scores, intentionally excluding the top performers; 20 participants completed the program, with an average age 61. Over 80% adhered to the exercise regimen and took most supplements, and the protocol proved safe.

After eight weeks cognitive scores rose significantly — stronger than expected — muscle strength and mass increased significantly, and visceral fat decreased. Peak oxygen consumption (VO₂) did not change significantly, while the CD4/CD8 T‑cell ratio, a marker of immune ageing, shifted significantly toward the 1.5–2 range.

XPRIZE Healthspan offers a $101 million prize; NUS was named a finalist, while ten other teams received $1 million grants for coordinated trials. Meyer’s team is preparing a 12‑month study with 180 participants that will add omega‑3, vitamin D, zinc and coaching to test whether the pilot effects scale up.

In parallel, Meyer and co‑author Ukhvat are designing a trial frameworks that adapt interventions in real time using hierarchical outcomes, responding to new measurements as they arise.

🔗 Read original →

Virtual Biotech AI Agents Predict Lung Cancer Target, Later Confirmed by FDA Breakthrough Therapy

A virtual biotech company composed of tens of thousands of AI agents analyzed lung‑cancer data available only up to January 2025 and identified the protein B7‑H3 as a therapeutic target, proposing a strategy against it without seeing newer data.

In August 2025, the antibody ifinatamab deruxtecan from Daiichi Sankyo and Merck received FDA breakthrough‑therapy status for the same target, showing a 48% response rate in patients with small‑cell lung cancer.

The work was peer‑reviewed and published on Science, September 17 2026; an analysis of 566 patients revealed that high B7‑H3 activity correlates with a 62% higher risk of death compared to low activity, and spatial mapping showed B7‑H3 accumulating in cancer‑associated fibroblasts that push immune cells out of the tumor microenvironment.

Because no druggable pocket was found on B7‑H3, the team suggested an antibody‑drug conjugate to deliver toxin directly to B7‑H3‑positive cells; Virtual Biotech grew from the earlier Virtual Lab system that had produced COVID nanobodies outperforming human ones, and its agents now read papers and databases directly through the shared Paperclip filesystem.

In a second case, the MOONGLOW trial of vixarelimab for ulcerative colitis was re‑examined; agents found that non‑responders had elevated OSMRβ in fibroblasts before therapy, recommending patient selection based on fibroblast OSMR levels, and noted that Virtual Biotech’s fibroblast‑focused targets are 48% more likely to reach market than broadly active targets.

🔗 Read original →

Antimicrobial K21 extends worm lifespan via macrophage mitophagy

K21 is a synthetic antimicrobial compound sold under the brand FiteBac and has been incorporated into FDA‑approved dental products for over ten years. It is used to treat burns and diabetic wounds, shows broad‑spectrum activity against bacteria, viruses and fungi, and does not damage membranes or harm human cells. Researchers hypothesized that its wound‑healing benefits arise from enhanced macrophage activity.

When human macrophages were treated with K21 and infected with the intracellular pathogen Enterococcus faecalis, bacterial uptake remained unchanged but intracellular survival dropped dramatically. The study was published 17 September in Life Science Alliance. Single‑cell RNA sequencing revealed that K21 stimulates mitochondrial fission and subsequent mitophagy by stabilizing the DRP‑1 protein, independent of oxygen deprivation or membrane damage — a process the authors term mitohormesis. In contrast, smokers with a SIRT3 defect in alveolar macrophages show impaired mitophagy and reduced bacterial clearance.

Feeding Caenorhabditis elegans bacteria expressing K21 induced the same mitochondrial fission/mitophagy in the worm intestine and significantly extended lifespan compared with controls (p<0.0001). No changes were observed in feeding, development, body size or egg count, ruling out caloric restriction. Although mitophagy‑mediated lifespan extension in C. elegans is usually linked to the atfs‑1 gene, K21 prolonged life in atfs‑1 mutants, indicating a distinct pathway. Eight genes showed concordant regulation in humans and worms, providing molecular evidence of a conserved mechanism.

The authors conclude that K21’s enhancement of human macrophage antimicrobial defense and its lifespan‑extending effect in worms stem from a single intervention in mitochondrial quality control, linking infection resistance and longevity. Lifespan was measured only in worms; in humans the assay focused on macrophage antimicrobial activity. Both project leaders received grants from FiteBac, which markets K21. K21 remains principally a topical antimicrobial; the mitophagy activity and worm lifespan extension represent newly discovered properties.

🔗 Read original →

DNA methylation variability adds independent aging signal boosting mortality prediction

Researchers from Simon Fraser University, the US National Institute on Aging, and Dalhousie University examined DNA methylation in blood from 1445 Canadians aged 45–85. Besides the usual sites where average methylation shifts with health, they searched for loci where the spread of methylation between people of the same age changes, discovering a largely non‑overlapping set of regions.

A biomarker built from both the average level and the variability predicted all‑cause mortality more accurately than either signal alone and outperformed the Horvath, Hannum, and GrimAge clocks in two independent cohorts. The study appeared September 17 in Journal of Gerontology: Series A.

The concept grew from the team’s 2024 work showing that many physiological traits are linked to an optimal middle value, with deviation in either direction worsening health. Similar patterns had been noted for telomere length variability in healthy centenarians and for rising methylation variability in tumours, indicating a loss of epigenetic stability.

Combining 789 CpG sites into a composite score, the biomarker surpassed each individual signal and five established tools—frailty index, Horvath, Hannum, PhenoAge, and GrimAge v2—in a hold‑out Canadian sample, where none of those tools showed a significant mortality link. Its heritability ≈41% indicates that lifestyle and environment drive most of the variation.

The same advantage held in the independent Baltimore Aging Study (728 participants); the biomarker correlated with smoking and inflammation markers such as C‑reactive protein and interleukin‑6. Among five age‑related diseases tested, it significantly predicted only higher risk of chronic obstructive pulmonary disease after a stricter significance threshold to guard against false positives.

A comparable finding emerged recently from the Generation Scotland cohort, where sites showing age‑related increases in methylation variability alone forecast mortality. While epigenetic clocks have traditionally read methylation on a single axis—average level—this work reveals a second, independent axis—variability—and shows that ignoring it incurs a measurable loss in predictive precision.

🔗 Read original →

Stanford shows fore‑ and hindbrain arise from distinct progenitor cells

Researchers from Stanford Medicine, Caltech, and UCSF tracked hundreds of progenitor cells in mouse embryos, finding that descendants stayed almost exclusively in either the fore‑/midbrain or the hindbrain stream. Of the 494 clones examined in 16 embryos, mixing between the two streams was virtually absent.

At the 7.5‑day stage, the genes Otx2 and Gbx2 become active in different zones of the neural tube. Labeling cells with active Gbx2 using a short 12‑hour pulse showed that their progeny contributed only to the hindbrain, without mixing with Otx2‑lineage cells.

Attempts to reverse progenitor fate with signaling cues failed, both alone and in combined culture. Chromatin analysis revealed that the two cell types already had distinct DNA accessibility patterns by the 2‑day stage, locking in their developmental programs.

Using the newly identified hindbrain progenitor, the team generated human hindbrain motor neurons that release acetylcholine and fire action potentials. In the body, these neurons control facial, tongue, and throat muscles for swallowing, speech, facial expression, and eye movements.

Loss of such neurons is likely to impair swallowing in spinal muscular atrophy and amyotrophic lateral sclerosis, leading to choking, pneumonia, and sometimes death. The same developmental split helps explain diffuse pontine glioma—the leading cause of cancer death in children, with a five‑year survival of about 1%. Co‑author Michelle Monje received the Brain Prize in 2025 for discovering that healthy neurons form synapses with glioma cells and electrically stimulate their growth.

The progenitor

🔗 Read original →

Scientists Reverse APOE4‑Driven Blood‑Brain Barrier Leak by Targeting Fibronectin

Although the APOE4 variant has recently been linked to longevity, it also triggers a dangerous side‑mechanism that scientists have now deciphered and learned to reverse. In Alzheimer’s disease the blood‑brain barrier becomes leaky because of pathological accumulation of the protein fibronectin. Normally fibronectin acts as a “scaffold” for wound healing, but under the influence of APOE4, inflammation and amyloid plaques astrocytes overproduce it.

Excess fibronectin clogs blood vessels, blocks chemical signals (including VEGF receptors) and destroys tight‑junction contacts, letting blood

🔗 Read original →

Anthropic Confirms Own Wet Lab for Claude‑Driven Robot Experiments

On 18 September 2026, Eric Cauderer‑Abrams, head of Anthropic’s life sciences division, told Reuters that the company owns and operates its own “wet” laboratory in the San Francisco Bay area. The facility lets Claude direct robots to run experiments with minimal human involvement, though a human operator remains required for safety. A company spokesperson said the lab focuses on fundamental biology, not drug development.

Previously Anthropic was known mainly for providing tools to external labs; in April it bought startup Coefficient Bio for $400 million in stock. In June it launched Claude Science, a workbench for researchers, and announced preclinical programs for diseases that are not profitable for pharma. In August Claude designed 1320 mini‑proteins for 15 targets, and two external contract labs verified the binding while Anthropic itself did not touch the test tubes.

Also in August Anthropic opened the Model Hardware Standard, letting AI agents remotely read and adjust external instruments; in one

🔗 Read original →

First thing after installing a neural interface: say 'I love you'

What should you do after installing a neural interface? Control drones, operate robotic arms, browse the internet, or hack megacorporation servers?
Many imagine futuristic feats right away.

But the first step is simpler: tell your significant other 'I love you'.
Only after that should you move on to piloting drones, hacking servers, or any other high‑tech activity.

🔗 Source: @solid_state_humanity

Madrid Longevity March to Demand Recognition of Aging as Medical Target

On 1 October, International Day of Longevity, the International Longevity Alliance will stage a march from Spain’s Royal National Academy of Medicine to the City Hall palace of Cibeles in Madrid. The procession will pass Puerta del Sol and the Congress of Deputies, with four stops where participants will read the Madrid Longevity Declaration of 2023.

The march coincides with the International Longevity Summit, both events led by summit organizing committee chair José Cordeiro. According to Longevity.Technology, 18 September, the route symbolically links academic medicine, parliament, and municipal authority, turning a scientific argument into a political demand for recognizing aging as a medical target and securing its funding.

The 2023 declaration notes that germ cells, cancer cells, and certain hydra and jellyfish are biologically immortal, and it cites the 2012 Nobel Prize awarded to Japanese biologist Shinya Yamanaka for showing that mature cells can be reprogrammed to a stem‑cell state. It asserts that rejuvenating the organism is now merely a matter of time, quoting Aubrey de Grey’s 2011 line — “The first person who will live to 150 years has already been born” — and adding David Sinclair’s view that “Aging is a disease, the most common one, and it needs to be treated aggressively.”

José Cordeiro, an MIT engineer and prominent transhumanist, authored the 2018 bestseller *The Death of Death*, translated into eight languages, which argues that with sufficient funding death could become optional by 2045. He calls the march “the biggest public longevity march in history” but stresses that “This is only beginning,” explaining that the goal is not to celebrate aging as it exists today but to end the diseases that make aging feared.

Organizers point to Eurostat data showing life expectancy at birth in the Madrid community rose from 84.3–85.7 years (2014–2024), keeping Spain among Europe’s longest‑lived populations. International Day of Longevity was established in 2013 by activist Ilya Stambler and had previously been marked by local lectures and petitions in Italy, Brazil, and Pakistan. The growing size of the International Longevity Summit in Madrid — now hosted by the Royal Academy of Medicine — has finally given the movement critical mass, taking the conversation from conferences and labs onto the streets, as noted by Longevity.Technology.

🔗 Read original →

Two‑Decade Review Links Nuclear Defect to Progeria’s Multi‑Level Damage

Professor Karim Jabali of the Technical University of Munich published a twenty‑year review of Hutchinson‑Gilford progeria on 17 September in Mechanisms of Ageing and Development. The paper synthesises research from 2002 onward into a single causal chain that connects a nuclear defect to cellular, tissue, and organismal decline.

The disease stems from a single LMNA mutation that prevents removal of a farnesyl tail from progerin, a truncated lamin A protein. Because the tail remains attached, progerin sticks irreversibly to the inner nuclear membrane, distorting the nucleus. Cells appear deformed under microscopy but divide and function normally until a heat shock reveals their inability to recover, unlike healthy cells.

Nuclear pores assemble correctly after division but gradually clump with age, and progerin accumulates over time, especially in vascular cells where it persists for weeks. This selective buildup explains why the heart and arteries suffer most despite ubiquitous LMNA expression. Similar low‑level progerin appears in ageing skin and in kidney‑disease‑damaged vessels, linking the defect to normal ageing processes.

At the cellular level, nuclear damage overwhelms the protein‑clearance system, and weakened clearance accelerates further progerin buildup—a vicious cycle. Sulforaphane from broccoli and rapamycin can restore this clearance and ease cellular symptoms, though they do not eliminate them.

Tissue‑level damage creates a second cycle: senescent cells release inflammatory signals that drive neighbouring cells to age faster. The approved drug baricitinib dampens this signal; combined with lonafarnib it extends mouse lifespan more than either agent alone. Lonafarnib, the only progeria therapy with proven survival benefit, was approved in the United States in 2020 and acts at the nuclear level by weakening progerin’s membrane attachment.

Jabali proposes adding baricitinib and sulforaphane to target tissue and cellular levels while awaiting more radical tools. Genome editing that corrects the LMNA mutation—or an approach that bypasses the mutation—has raised mouse lifespan 2.4‑fold, but such treatments are not yet available to children. Key nuclear and cellular experiments come from her lab (active since 2002); tissue and organismal data derive from independent studies. Jabali earned her doctorate under Nobel laureate Günter Blobel, who discovered cellular protein‑targeting signals, and her work on nuclear‑pore clumping extends that legacy. She concludes: “HGPS is not a copy of physiological ageing but a genetically programmed system that shows how persistent nuclear stress progressively erodes the organism’s conservative adaptive networks.”

🔗 Read original →

Senolytic drugs restore motor cortex function in ALS mouse model

On September 18, Neurobiology of Disease, September 18 published a study from the University of Missouri showing that signs of cellular senescence appear early in the brains of mice carrying the TDP-43 Q331K mutation, before full ALS disease onset. The researchers used this mouse model to test whether removing senescent cells could alter disease progression.

Fifteen weeks of treatment with the senolytic combination dasatinib and quercetin (D&Q) — first tried in humans in 2019 for pulmonary fibrosis — was administered to a cohort of the mice. In earlier pilot studies in elderly people at

🔗 Read original →

HA-Tau nanospheres capture inflammatory proteins and restore memory in aged mice

The team from Duke University and UCLA described HA‑Tau — hyaluronic acid conjugated with taurine and formed into nanospheres of about ~200 nanometers — in a preprint dated September 18, 2026 preprint, September 18, 2026. One injection of these nanospheres into old mice after a bone fracture lowered inflammatory proteins in the blood, preserved the blood‑brain barrier, and returned spatial‑memory performance to normal levels.

In 20–22‑month‑old mice (equivalent to elderly humans) the nanospheres administered 30 minutes after a tibial fracture captured roughly ~548 different proteins from plasma, most of them belonging to inflammatory pathways. Levels of IL‑6, IL‑1β and TNF‑α dropped, barrier‑strengthening proteins remained, dye leakage decreased, and glial activation in the hippocampus — the memory centre — nearly returned to baseline, so memory test scores matched those of non‑operated animals.

Molecular docking showed that a partial substitution of taurine for hyaluronic acid — around ~43% — creates new cytokine‑binding sites while strengthening existing ones; full substitution abolishes both effects due to steric hindrance. This optimal degree was chosen by direct testing on blood proteins, where it bound best. Taurine’s controversial reputation in geroscience — highlighted by a Science 2023 article linking low taurine to aging and lifespan extension in mice, though later human data failed to confirm the link — is repurposed here simply as a hook for cytokines.

Tests on plasma from elderly donors and human brain‑vascular cells reproduced the same effect: the nanospheres captured IL‑6, IL‑1β and TNF‑α, which signal through the NF‑κB switch, and made the barrier less permeable than without them. This cytokine node has previously been tied to cognitive impairment in living people by Canadian data.

The nanospheres barely enter the brain and accumulate mainly in the liver, causing no detectable liver damage within the first day; the brain is protected because the substance that intercepts the inflammatory signals never needs to cross the blood‑brain barrier.

🔗 Read original →

Glycocalyx‑edited stem cells reverse osteoporosis in small trial

The approach uses a patient’s own mesenchymal stromal/stem cells (MSCs) harvested from bone marrow, which are bioengineered to restore surface expression of the molecule sLeX. This glyocalyx modification re‑equips the cells with the ability to home to bone marrow after intravenous infusion, turning them into a “living drug”

🔗 Read original →

Naive CD8+ T cells pre‑mark mitochondrial genes with H2A.Z via GABPα and IL‑7 for rapid response

Naive CD8+ T cells mark mitochondrial genes with the histone variant H2A.Z under the direction of GABPα and the background signal IL‑7, allowing them to launch a full effector response within hours of encountering their specific pathogen. This preparatory state weakens with age but can be restored, and the same mechanism enhances anticancer CAR‑T therapy.

The study was conducted by a group from the Chinese Academy of Sciences and published in Science Advances, 18 September. A year earlier the same lab showed that H2A.Z primes memory CD8+ T cells for rapid action; the new work extends this finding to naïve cells that have not yet met a pathogen.

A naive CD8+ T cell is a killer lymphocyte that conserves resources while waiting for its target. Upon recognizing the appropriate threat it has only a few hours to proliferate, produce effector proteins, and remodel its metabolism. How this readiness is kept on pause for years had remained unclear; the answer lies in DNA packaging.

H2A.Z is a histone “spool” around which DNA is wound to fit two meters of genome into a microscopic nucleus. Where H2A.Z resides, chromatin is looser and genes are easier to activate. Mice lacking H2A.Z in T cells show diminished naïve and effector populations, and the remaining cells poorly control a model bacterial infection and melanoma tumor growth.

The defect is specific: without H2A.Z, mitochondrial respiratory‑chain genes activate far less strongly upon stimulation, and adding pure ATP to the culture partially restores activation capacity—indicating that an energy shortage keeps the cells idle.

Mass‑spectrometry identified the depositor of H2A.Z on these genes as GABPα, known since the 1990s as

🔗 Read original →

China adds xenotransplantation, universal cell therapy, and organ growing to 2030 medical industry plan

On September 18 ten Chinese agencies led by the Ministry of Industry and Information Technology released document No. 210 – the medical industry development plan for 2026‑2030. Among its ten targets are industry revenue of 3.5 trillion yuan (~$523 billion) and at least a quarter of global first‑in‑class drugs originating from China.

The plan calls for technological breakthroughs in xenotransplantation (animal‑to‑human organ transplants), universal cell therapy with off‑the‑shelf products, creation of complex regenerable organs, and synthesis of artificial living systems (building functional biological systems from scratch). It also urges expansion of next‑generation genome editing and neuro‑interfaces that link the brain to computers.

Continuing the previous industrial cycle, China launched 230 innovative drugs during 2021‑2025 and ranked second worldwide in drugs under development, per the ministry’s explanation. The plan links this effort to accelerating population aging: by end‑2025 323.38 million people aged 60+ (23 % of the population), according to the National Bureau of Statistics National Bureau of Statistics.

In the traditional Chinese medicine section, anti‑aging agents are listed among new consumer directions alongside elderly care, rehabilitation, and health management. For rare diseases it speeds development of cell and gene therapies – e.g., hemophilia, ALS, and Gaucher disease.

In the first half of 2026 licensing deals for Chinese innovative drugs reached $106.3 billion, versus $135.7 billion for all of 2025. The plan connects these industry scales to concrete technology targets through 2030.

🔗 Read original →

FDA Accepts Reasonable Expectation of Effectiveness Package for Loyal’s LOY-003

On September 16, the FDA accepted an interim submission for Loyal’s LOY-003 that includes a “reasonable expectation of effectiveness” (RXE) determination. This marks the third RXE granted to separate Loyal programs, according to the company. The RXE step evaluates whether a future drug can deliver the claimed benefit.

LOY-003 is a daily tablet designed for large and giant breeds that aims to extend their lifespan. Loyal links the shorter life of these dogs to the hormonal signals GH and IGF-1, which influence body growth. The tablet is intended

🔗 Read original →

Calorie restriction primes mice for fasting hours before the stomach empties

Researchers from Roman Kondratov’s lab at Cleveland State University compared calorie restriction (CR) with a single fast‑feed‑fast cycle (FRF) that delivered the same total amount of food and the same total fasting time. Mice on CR received 70% of their normal ration in one daily meal, ate it within two hours, and then fasted for about 22 hours. The FRF group underwent one 22‑hour fast, a two‑hour feed, then another 22‑hour fast.

In the CR mice, metabolic switches to a fasting mode began many hours before the stomach was empty—insulin fell already after two hours and mTOR activity dropped after six hours, roughly 12 hours before the stomach became empty at around the 18‑hour mark. In the FRF mice, insulin and mTOR changes tracked stomach emptying closely, occurring only after the stomach was nearly empty at about six hours.

The study showed that only CR improved glucose tolerance and kept liver fat low, whereas the FRF mice displayed sharp post‑meal glucose spikes, no improvement in glucose tolerance, and a progressive accumulation of liver fat that peaked around the 18‑hour fasting point.

These results were published 17 September in Cell Reports. The work also builds on earlier findings that functional circadian clocks are required for CR‑mediated lifespan extension, and that the circadian protein PER2 can bind mTOR to tune its response to fasting—a mechanism highlighted in a separate September study.

The 2017 Nobel Prize in Physiology or Medicine was awarded to Jeffrey Hall, Michael Rosbash and Michael Young for elucidating the molecular mechanism of circadian clocks.

🔗 Read original →

TH1834 protects human heart cells from reperfusion injury after heart attack

TH1834 blocking TIP60 protected human heart cells and mini‑tissues grown from stem cells from damage during reperfusion after a heart attack — an effect previously demonstrated only in mice.

Rest

🔗 Read original →

Claude autonomously GPU‑optimizes 36 open‑source biomolecule tools, boosting speed up to 4.1×

On September 17 Anthropic released the Anthropic technical report showing that its research model Claude autonomously optimized code for 36 versions of more than 30 open‑source scientific programs that predict protein 3D shape, design molecules, and analyze genomic data. Two staff members with biology knowledge but no GPU‑coding experience led the effort. Speed and accuracy were validated on thousands of predictions, and all code was made open.

The programs spend most of their compute on triple‑segment comparisons; doubling a molecule’s size raises the cost eight‑fold. Acceleration comes from rewriting low‑level GPU kernels. Claude wrote a set of FlashPairformer kernels that

🔗 Read original →