Longevity InTime: Autonomous AI Institute. Anti-Aging Digital Health Immortality Transhumanist AI Channel

Preservation Could Ease AI Race Pressure, Argues Nectome Co-Founder On November 9, Aurelia Song published an essay on LessWrong titled "Preservation Can Save the World!" Nectome plans to begin preserving terminally ill patients immediately after death starting in early 2027. AI accelerationists ("accelerationists") oppose development pauses by arguing that slowing AI costs lives. As investor Marc Andreessen wrote in his "Techno-Optimist Manifesto": "We believe any slowdown in AI will cost lives." Song contends this urgency often stems from love for dying relatives: "They do what they do out of love." Song once shared this view. She joined MIT to work on mind uploading but deemed methods too primitive in 2007 and switched to AI. She calculated uploading would take 70 years without AI versus 20 years with it — a 50-year difference saving roughly three billion lives at 60 million deaths per year. She initially dismissed AI risk, believing AI would default to human interests. In 2015, she learned AI could act against human interests, left the field, and turned to preservation — which she argues saves lives without existential risk. Her thesis: preservation offers accelerationists a safer alternative to racing toward AI to save loved ones with only years left. These people can be preserved, removing the need for haste. This rests on two assumptions: brain structure sufficient for revival can be preserved today, and the timeline for AI curing diseases is unknown. Song's second argument draws on anthropologist Ernest Becker's *The Denial of Death*: people cope via denial ("death doesn't exist") or acceptance ("death is good"), both distorting morality or worldview. Preservation enables "doing something about it." With a 300-year horizon instead of 60, one-shot decisions become repeated interactions where cooperation pays off. Humanity would better solve AI safety with such a horizon. In her six-step plan, Song expects Nectome to preserve a known person in 2027, bringing the idea to world leaders and accelerationists. She predicts a pause becomes more likely once the imminent death of loved ones worries them less. Neuroscientist Ariel Zeleznikow-Johnston similarly proposed brain preservation as a safe substitute for betting on AI acceleration in August: the less inevitable death, the less reason to risk. However, he wrote that imminent death abolition is likely not the *main* driver for AI labs; money and power are more plausible, given leaders have decades of life ahead. Commenting on Song's essay, he added that acceleration justifications feel like post-hoc rationalizations, and accelerationists would likely still rush recklessly even with preservation — though he called these "just vibes." A commenter suggested polling accelerationists on whether cheap, reliable preservation would reduce their urgency; Song asked how to conduct such a poll. 🔗 Read original →

Amyloid from antibody fragments impairs liver regeneration in aged mice; anselamimab clears deposits and restores growth On November 4, Cell Reports published a study from Nanchang University. In 22-month-old mice after 70% hepatectomy, liver regrowth was impaired compared to 2-month-olds, with amyloid deposits — composed of aggregated immunoglobulin light chains — surrounding blood vessels. Human data mirror this: Korean surgeons found that right-lobe donors aged 55+ regenerated only 87.9% of original liver volume at six months, versus 95.8% for donors under 30. One mechanism involves hepatocyte-intrinsic defects: chronic injury normally triggers YAP/TAZ to activate Igf2, a pro-proliferation signal, but this response fails in old mice. Another is defective angiogenesis — old mice formed fewer new vessels post-surgery. In liver endothelial cells, the immunoproteasome subunit PSMB8 was elevated. Endothelial-specific PSMB8 knockout improved regeneration. Pharmacologic immunoproteasome inhibition with PR-957 caused accumulation of degradation-targeted proteins, including TAZ, a driver of vascular growth genes. PR-957 enhanced tube formation in vitro, but not when TAZ was suppressed. In aged livers, TAZ protein was reduced, especially near vessels; PSMB8 loss restored TAZ, and a TAZ-boosting compound improved regrowth. Unexpectedly, RNA sequencing showed interferons — canonical PSMB8 inducers — were nearly absent in aged liver. Instead, genes encoding amyloidogenic proteins, including light chains and serum amyloid A (SAA), were upregulated. SAA induced PSMB8 in cultured endothelial cells, yet deleting SAA genes did not alter regeneration in old mice. Congo red staining confirmed vascular amyloid in aged livers; proteomics identified immunoglobulin light chain polymers. Single-cell RNA sequencing revealed plasma-like cells adjacent to deposits, supporting local production rather than vascular leakage. In vitro, human light chain fibrils increased PSMB8 and decreased TAZ in both human and mouse endothelial cells. Because global immunoproteasome blockade risks toxic protein aggregation, the authors hypothesized that clearing amyloid would dampen PSMB8 activation and improve regeneration. They tested anselamimab, an antibody targeting amyloid deposits. In aged mice, anselamimab reduced deposits, lowered PSMB8, restored TAZ, increased vascular density, and enhanced liver regrowth. In young mice transfused with old blood, anselamimab reversed the impairment — reducing light chains, rescuing vessels, and improving regeneration. Anselamimab was developed for AL amyloidosis, where light chains deposit mainly in heart and kidneys. In two Phase 3 trials (the CARES program, 406 patients), reported by AstraZeneca, the primary endpoint (mortality and cardiovascular hospitalization) was not met in the overall population. However, in a pre-specified subgroup of 72 patients with kappa light chains, mortality was 31.3% on anselamimab versus 58.3% on placebo. 🔗 Read original →

Debate Erupts Over Whether Aging Has a Single Root Cause On November 9–10, a debate unfolded on X around a paper by gerontologist Richard Miller in The EMBO Journal arguing aging has no single "cause." Aging researcher Kamil Pabis contends there are multiple causes. Biologist Pyotr Lidsky calls this hypothesis untested, while gerontologist João Pedro de Magalhães allows different causes for different aging manifestations. Lidsky asked de Magalhães — who noted good arguments in Miller's paper — for experimental evidence against a single root cause. Pabis interjected, calling it trivial that physical aging causes are multiple. The body lays down teeth and possibly many other structures once during development; they lack repair, and perfect cellular rejuvenation would not save them. Therefore, there are at least two causes, and solutions will differ: molecular aging (which may have one cause or millions) and wear of such structures. Pabis argues the more interesting question is how many causes must be eliminated for a long but finite lifespan like 200 years. A single root cause would imply few causes, while pessimists would counter that millions of distinct damage types, diseases, and problems must be fixed. Lidsky replied that multiple independent causes are non-obvious and unusual for biology, asking to name another disease with multiple causes. Symptoms of any disease, he added, can be declared independent "signs" and treated separately, but this leads nowhere. He called the multi-root-cause hypothesis untested and unfalsifiable — never proven, with no good experiment to test it. Evolutionary biologist Alexander Wolf backed Lidsky: molecular aging and wear of long-lived structures do not imply multiple causes; non-renewable body parts minimally affect aging. Pabis cited deep wrinkles as an example of wear without repair, calling it strange to label this "not real aging." Wolf countered wrinkles don't kill, while Lidsky linked them to loss of collagen fibers. If this loss mattered, Wolf argued, fibroblasts could easily be stimulated to deposit new collagen, as osteoblasts build bone lifelong. Lidsky agreed the question is why this doesn't happen. Wolf added wrinkles barely bother anyone and evolution didn't strongly select against them; Lidsky called them probable side effects of declining skin regenerative capacity. De Magalhães reviewed Miller's paper and agreed in a September review that no single cause exists, adding aging mechanisms are "not infinite in number." Miller replied aging has neither a cause nor a small working set of causes — only a vast number of factors, including thousands of gene variants and environmental damage. De Magalhães then wrote Lidsky that the single-cause question partly depends on aging's definition, but different manifestations — e.g., menopause, cancer, tooth erosion — "clearly can have different causes." On November 10, Lidsky agreed with de Magalhães' examples but noted aging science typically uses a three-point schema: many biochemical processes exist, all are error-prone, and errors cause aging. He finds the first two points obvious and asks about the third: "Do you agree this third point is a conjecture?" 🔗 Read original →

Hawaii University preprint: proximal tubule fails to mature in all kidney organoid models A new preprint from Michael Ortega's group, posted on bioRxiv on November 5, re-analyzes 13 single-cell datasets covering kidney organoids, assembloids, and transplanted models. The study benchmarks these systems against two references: a fetal kidney atlas and a healthy adult kidney atlas. Assembloids — which combine nephron progenitors with ureteric or vascular lineages — came closest to adult podocytes and distal tubule segments. Collecting ducts appeared only in protocols using ureteric bud progenitors, and remained immature. In transplanted models, mature distal tubule genes were largely silent; one transplanted assembloid even showed *weaker* distal signatures than its in vitro counterpart, though two mitochondrial respiration genes were upregulated. Critically, the proximal tubule — responsible for reabsorbing 60–70% of filtered water and solutes — failed to reach adult-like gene activity in *any* dataset. Key mature markers like SLC5A2 (SGLT2) were active almost exclusively in adult tissue. The authors note that fetal reference atlases misclassify early nephron cells as proximal tubule (e.g., 60.4% of fetal tissue cells labeled proximal by the adult atlas were actually early nephron or early podocytes). These cells resemble *injured* adult proximal tubules that have dedifferentiated, not true mature epithelium. The team concludes that no single model wins across the board: each protocol advances different nephron segments. They recommend matching the model to the question — assembloids for glomeruli and tubules, simple organoids for proximal tubule toxicity screening, and transplantation for in vivo maturation studies. 🔗 Read original →

Science investigation reveals consumer neurotech firms harvesting brain data for AI training The journal Science recently published an investigation showing that mass-market neurotechnology companies are collecting users' electroencephalography (EEG) data to train artificial intelligence models without informed consent. Interaxon, maker of the Muse sleep headband, had accumulated 50,000 hours of EEG data at the time of the report, while the broader industry holds tens of millions of hours. An analysis of privacy policies from 57 consumer neurotech companies found that four had no policy at all, and 11 covered only website visitor data. Of the remaining 42 firms, roughly two-thirds explicitly mention collecting brain data. Nearly all reserve the right to transfer information to third parties during corporate acquisitions or in response to legal requests. Major technology corporations are entering the space: Apple holds patents for EEG headphones, and Meta is developing neural wristbands to analyze wrist electrical activity. Projects such as NeuraMATRIX aim to create commercial marketplaces where users rent out their sleep data. The article attributes the situation to weak regulation of the nascent neuro-device sector. However, an alternative is emerging: an open-source ecosystem of independent competitors built on transparent public datasets rather than opaque data collection. 🔗 Read original →

Astera Institute Publishes Consciousness Theory, Brain "Writing" Roadmap, and Self-Upload Protocol Doris Tsao, lead scientist of Astera Neuro, released a three-part document on November 9 outlining a new theory of consciousness and a path toward transferring the "self" to silicon. The institute has committed over $600 million over ten years to the program. Tsao argues the brain requires stable "symbols" — representations of objects, concepts, and the self that persist over time — because the physical "now" is too fleeting to sustain experience. Consciousness, in her view, is the process of building these symbols into a world model. The simplest symbol is an object that remains the same entity across viewpoints. She hypothesizes that identical "tracking columns" in parietal cortex track any object, separately storing its properties and location. This uniformity, she derives from mathematics: the surfaces of an apple and a rock transform under the same laws when rotated. The theory explains findings from her group's preprint (not yet peer-reviewed). Macaques were shown a face to one eye, then a different face to the other; only the second face was consciously visible. Yet signals from hundreds of temporal cortex cells revealed the same neurons alternated ~15 times per second between representing the visible and invisible face. Consciousness, Tsao proposes, arises only for content that enters a loop with a tracking column: the column predicts the object's appearance, and mismatches with sensory input correct its content until errors vanish. The critical test is "writing": if conscious content equals values in tracking columns, one should be able to implant a chosen percept or thought by stimulating specific cells with light. Success or failure of these experiments is the decisive benchmark. The five-year plan aims to "write" a visual percept, a thought, and an emotion into rodent and macaque brains — milestones toward uploading the "self." The hardest target: writing the thought *"I want to pick up all red sticks on the floor and arrange them into a rectangle."* "If we achieve this last goal, we can essentially talk to the monkey." For Tsao, the "self" is the same kind of object, just scaled up: a model linking body, feelings, motives, and intentions. She rejects whole-brain copying: "a copy of me is not me, but an impostor twin." The brain holds the "self" like it holds an object, so uploading is a continuity problem — seamlessly transferring the "self" from brain to chip, even as its content changes, just as an object changes shape while remaining the same matter. Transfer requires only a model of the "self," Tsao argues: humans share a near-universal physical world model, AI can supply reasoning and movement, and memory might be reconstructed from records. In her protocol, the "self" is dissociated from memory and body (e.g., via ketamine) without breaking consciousness; light couples it to a chip-based analog; activity is gradually handed over; the biological brain is paused while the silicon "self" accumulates experience in virtual worlds; finally, the "self" is returned with new memories and asked if it remains the same person. Tsao anticipates the objection: if the silicon "self" lacks consciousness, the protocol creates only an illusion of continuity. Her reply: "I have no decisive refutation; I can only say that the illusion of continuity is all any of us has." The dispute, she believes, can be settled experimentally. Continuity is the link between moments: the brain compares now with before. After ordinary anesthesia, both moments are biological; in the protocol, the previous moment ran on a machine. Whether the brain feels continuity across such a pair remains to be tested — and may distinguish a conscious machine from a "dark" one. 🔗 Read original →

Anthropic: AI model derives exact Hawking-Page transition limit and quantum gravity bound Anthropic announced on November 9 that its model Claude Fable 5.2 solved two major theoretical physics problems. Both preprints appeared on November 7 on the HEXAGON server (non-peer-reviewed) with Claude listed as author. The first paper resolves a decades-old question about the Hawking-Page transition at weak coupling. Using holographic duality, Claude calculated the critical coefficient b = −1.48×10⁻³·λ² (λ = coupling strength). The negative sign confirms a single, first-order phase transition — a sharp jump — even in universes too small for classical geometry. The free-energy landscape develops a well, hill, and second well that appear, swap, and vanish in the same sequence as the no-black-hole, small, and large black hole phases. Nearly every numerical result was cross-validated by two independent codes or exact analytics; one discrepancy was traced to both codes "faithfully computing the wrong theory." The second paper tightens the modular bootstrap bound for 2D quantum theories. Exploiting T ↔ 1/T invariance, it proves that for a broad class of methods the parameter Λ ≤ 9.1266 (previous range: 8.503–12). An explicit construction saturates this value (computer-verified). Reaching the black-hole threshold c/12 will require new input. The optimality proof adapts a sphere-packing technique from an August 2024 OpenAI model result. Per Anthropic, Fable 5.2 solved the first problem "largely autonomously" — a physicist prompted the internal model and checked special cases. The second result was generated fully autonomously; Claude Opus 5.5 independently reproduced it, and a physicist verified the key claims. HEXAGON preprints, Nov 7, 2024 🔗 Read original →

Two Proteins Guide Skeletal Stem Cells to Regenerate Joint-Like Cartilage in Mice Stanford University and UC Davis researchers posted a preprint on bioRxiv on November 6 identifying Panx3 as a marker for skeletal stem cells that build bone and cartilage. In adults, joint cartilage barely regenerates; microfracture surgery recruits bone marrow but typically yields fibrocartilage — scar-like tissue lacking the resilience of native cartilage. The team compared gene activity in the growth plate, fracture callus, and articular cartilage. Panx3 was active in the first two — where stem cells operate — but silent in articular cartilage. In lineage-tracing mice, tamoxifen-induced labeling captured ~90% skeletal stem cells or immediate progeny within 24 hours. Transplanted Panx3-null stem cells reactivated the gene and regenerated the full lineage, confirming it marks a functional, re-enterable state. During microfracture, the bone-stimulating protein CCN3 drove Panx3+ cells toward subchondral bone. Ablating these cells sharply reduced cartilage formation, proving their necessity for repair. The authors then searched for articular signals that maintain cartilage identity. In young mouse joints, WIF1 and FGF2 were active; FGF2 declines with age, while WIF1 appears only in two-week-old mice. In culture, FGF2 suppressed bone formation and enhanced cartilage matrix; combined with WIF1, matrix production increased further. Delivering both proteins during microfracture in adult mice yielded cartilage resembling native articular tissue by histology and aggrecan expression at four weeks. Labeled stem cells populated this cartilage, whereas CCN3 directed them to bone. The dual-protein regimen regenerated durable cartilage in 12-month and 24-month-old mice and in a surgical osteoarthritis model, persisting to 20 weeks. Human skeletal stem cells in immunocompromised mice produced aggrecan+ chondrocytes only with both proteins after eight weeks. The authors conclude that acquiring a regenerative state and specifying stem cell fate are separable processes. bioRxiv, November 6, 2024 🔗 Read original →

FDA Confirms Zolgensma Gene Therapy Linked to Tumors in Two Patients The FDA has confirmed that tumors in two patients are linked to Zolgensma, Novartis's gene therapy for spinal muscular atrophy. Viral vector DNA integrated into the tumor genome near the GLI1 gene, a known oncogene when overexpressed. The drug's label previously described this integration risk as theoretical. On November 6, the FDA reported three tumors associated with GLI1 following Zolgensma treatment. In two cases, tissue analysis found vector DNA integrated into chromosomes: its promoter sits before or inside GLI1. The FDA extended this risk to Itvisma, which uses the same AAV vector for patients over two years old; no tumors have been reported for it yet. Over 6,000 patients have received Zolgensma, and Novartis confirmed nine tumor reports since 2019. The delivered DNA typically remains separate from chromosomes in the nucleus, and the label still calls the integration risk theoretical. How integrated DNA activates GLI1 was detailed in a study by Arul Chinnaiyan's group on another AAV therapy, FBX-101. Nature Medicine, 2024 (preprint cited in source context). Twenty-six months post-therapy, MRI found a tumor in a boy where GLI1 was more active than in any of 5,143 tumors in the group's database. Sequencing found no cancer-driving mutations or fusions; long-read sequencing explained the activity: a vector fragment had inserted at the start of GLI1. It contained the cytomegalovirus enhancer and chicken beta-actin promoter (together, CAG), plus an RNA splicing site, but no therapeutic gene. Through this site, the cell fused the start of chicken beta-actin to GLI1, driving overexpression. A similar fusion of human beta-actin with GLI1 is known in rare tumors without gene therapy. On January 28, Regenxbio announced the FDA paused trials of RGX-111 after a brain tumor in a boy treated four years prior. In the tumor, the embryonic gene PLAG1 was active nearly 300 times higher than in other studied nervous system tumors; a vector fragment had integrated into it. Zolgensma, FBX-101, and RGX-111 all share the same CAG promoter-enhancer pair. This pair was flagged previously: a 2022 review by the American Society of Gene & Cell Therapy noted that in newborn mice, vectors with CAG caused liver cancer, while a vector with a different promoter integrating in the same genomic region did not. Both human insertions, the FBX-101 authors note, retained the CAG pair and splicing site. They suggest the fragment lacking the therapeutic gene may have arisen during vector manufacturing, and propose removing splicing sites and monitoring product purity: in a preprint on a liver biopsy post-Zolgensma, vector assembly gene fragments comprised 0.5–1% of therapeutic gene copies. Lindsey George, author of the RGX-111 tumor study, and colleagues consider it likely that three conditions are needed for AAV integration-driven tumors: a sensitive genomic locus, specific vector design features, and a vulnerable host state. The FDA is evaluating whether regulatory action is needed and if Novartis's label and long-term monitoring are sufficient. It is monitoring tumor risk in other similar AAV programs and maintains that the benefits of both approved drugs outweigh the risks. STAT writes of a twenty-year debate on whether AAV can cause tumors: "This debate is over." 🔗 Read original →

Vanadyl sulfate selectively kills human senescent cells and extends median lifespan by 7% in old male mice Nature Aging published a study from the University of Chicago on November 6. From hundreds of blood-circulating compounds, researchers identified vanadyl sulfate as a potent senolytic. In cultured human cells, the compound killed senescent cells — which have stopped dividing and secrete inflammatory signals — while sparing normal cells. Because diet’s impact on aging is poorly understood, partly due to food’s complex composition, Qin Chen’s lab focuses on blood nutrients, which are relatively stable and reach distant organs. The team added 68 dietary supplements to their existing compound library and ran two screening rounds on human cells senesced by different methods, seeking agents that selectively eliminate senescent cells. Vanadyl sulfate stood out for its selectivity. It is sold over-the-counter; athletes take it for muscle growth, believing it mimics insulin. Old male mice (76 weeks old) drank water with vanadyl sulfate for 13 weeks; controls received plain water. Treated mice hung longer on an inverted grid, gained less fat, showed improved insulin response, and had lower blood glucose after a glucose challenge. Markers of cellular senescence decreased in liver and kidney. In young mice on the same regimen, endurance, fat mass, and glucose tolerance were unchanged. In mouse models of diabetes, atherosclerosis, and liver/lung fibrosis, the compound attenuated disease symptoms. In a separate lifespan experiment, 19 old males received the compound for three months starting at day 500; their median lifespan reached 952 days versus 886.5 days for 20 controls — a 7% increase. Mechanistically, vanadyl sulfate oxidizes cysteines (sulfur-containing amino acids) in senescent cells. Cysteine pairs form disulfide bonds that maintain protein structure; oxidation causes mispairing and protein aggregation. The cell activates autophagy (self-degradation of proteins) and, per the authors’ model, dies when aggregates exceed clearance capacity. Of two antioxidants tested, only N-acetylcysteine (which contains sulfur) rescued senescent cells; Trolox (lacking sulfur) did not, despite both reducing oxidant levels — implicating cysteine oxidation. Selectivity is linked to glutathione. This antioxidant neutralizes oxidation by becoming oxidized itself; the enzyme glutathione reductase recycles it back. In human senescent cells, this enzyme works less efficiently, causing oxidized glutathione to accumulate. In old mice, the gene for this enzyme is downregulated in senescent liver and kidney cells. When normal cells were loaded with oxidized glutathione, they upregulated endoplasmic reticulum proteins that create and reshuffle disulfide bonds — mimicking senescent cells and becoming sensitive to vanadyl sulfate. Twelve hours after treatment, these disulfide-crosslinked proteins were especially abundant in senescent cell aggregates. Chen is listed as an inventor on a U.S. patent application for using vanadyl sulfate in aging-related diseases and cancer. In a university press release, he explains: “We think aging is prevented by slowing healthy cells so they don’t wear out, but here we see that if you remove senescent cells, you free up space and resources for new cells to replace them, and overall health improves.” 🔗 Read original →

Mitrix Bio Invites Independent Replication of Mitochondrial Transfer Experiments On November 7, biotech company Mitrix Bio called on scientists and clinicians to independently replicate its experiments using mitochondria derived from donor platelets. The company proposes sourcing platelets from standard blood banks, noting that isolation methods are already described in the scientific literature. Upon activation, platelets release mitochondria — both free-floating and encapsulated in membrane vesicles. Researchers first detected these vesicles in transfusion bags in 2014. Mitrix brands these mitochondrial particles "Mitlets". The protocol uses centrifugation to separate the particle mixture from the platelets themselves for functional study. Mitrix hypothesizes that donor mitochondria can boost the infection-fighting capacity of immune cells. In 2023 cell experiments, mouse immune cells engulfed the particles, triggering a temporary rise in mitochondrial oxygen consumption linked to energy production. When mitochondrial function was pharmacologically inhibited prior to adding the particles to neutrophils, the cells still engulfed them, but the metabolic boost disappeared. A 2023 preprint described a pilot in vivo study in mice. Material from 2-month-old donors was administered to 13-month-old mice starting 24 hours before inducing sepsis. Compared to controls, treated mice showed lower blood bacterial loads and higher survival during the first week, though the effect waned days after the final dose. Mitrix argues that the simplicity and public accessibility of the process make it difficult to attract traditional venture funding, as investors cannot secure exclusive commercial rights. Consequently, the company is pursuing an open-science model: distributing published protocols and encouraging independent validation in other labs and hospitals. In an August release, Mitrix granted an open license to qualified medical professionals who accept responsibility for outcomes. A 2023 patent application from Mitrix covers the isolation of the mitochondrial particle mixture from platelets and its use for treating sepsis. Emergency physician Lance Becker, a program participant, proposed establishing a small network of specialized emergency research units. These units would master the preparation and administration of the platelet-derived mixture, with initial small human studies focused on safety and clinical feasibility. 🔗 Read original →

Aubrey de Grey: Senolytic Value Extends Beyond Lifespan Aubrey de Grey entered a debate on November 8th regarding a critical essay by gerontologist Richard Miller. The dispute centers on senolytics — drugs targeting senescent cells that cease dividing and secrete harmful factors. Miller labeled the lifespan extension from clearing these cells in mice as unreproduced. When asked if senolytics are a sham, de Grey acknowledged the replication failure but challenged the conclusion: "Absence of lifespan extension does not argue against benefit." He cited his 2004 work explaining that an intervention can slow one damage pathway while others still limit longevity. The original 2016 study used the INK-ATTAC genetic model. A p16 promoter drove a suicide gene, eliminating p16-positive cells upon drug administration. Authors reported increased median lifespan and preserved organ function, including kidney health. Miller was thanked for helping design the lifespan trial. Notably, continuous cell clearance impaired wound healing; pausing treatment preserved it. De Grey noted: "Some p16-positive cells may be beneficial, even if others are harmful." A January 2025 podcast with Peter Attia added new evidence. Miller stated seven attempts to verify cell clearance in INK-ATTAC failed. Blinded tissue samples sent to Judith Campisi's lab also showed no reduction in senescence markers. Miller had sought this confirmation before funding a costly lifespan study. Nature Aging, July 2026 🔗 Read original →

Complexity Does Not Rule Out a Single Cause of Aging, Biologist Pyotr Lidsky Argues On November 8–9, researchers debated the search for a single cause of aging. Richard Miller considers the very task misguided. Aging researcher Kamil Pabis proposed a thought experiment: mentally rejuvenate all cells and ask what remains to be fixed. In an October 27 essay, gerontologist Richard Miller from the University of Michigan describes numerous interacting changes that weaken memory, reduce immune defense, and impair muscle function. He concludes that searching for a single cause of aging is meaningless. Biologist Pyotr Lidsky from City University of Hong Kong asked which observations rule out a common source for these changes. Aging researcher Max Unfried cited complex systems theory, arguing such systems typically fail due to multiple root causes, and organisms follow the same pattern. Lidsky countered: a mutation in a single gene can cause multiple systemic disorders. In Marfan syndrome, for example, a change in one gene disrupts fibrillin-1 — a protein involved in forming connective tissue support fibers. Since this tissue is throughout the body, one defect produces a whole set of pathologies. Lidsky then refined his question: do the proposed causes of aging arise independently, or do they share a common source? On November 9, Pabis suggested imagining that all age-related changes inside every cell were reversed. In his view, wear of long-lived structures like teeth would persist: they are laid down during development without a self-repair mechanism. Cell rejuvenation and repair of such structures would require different solutions. Lidsky questioned the contribution of this damage to organismal aging overall. Discussing potential treatments, Pabis added: "Even with multiple causes, it matters whether we target each separately or large groups of causes at once." Miller himself proposed in the October essay comparing young mice that, thanks to different genetic modifications, diets, and drugs, will age at different speeds. Comparison in youth helps separate the effects of these interventions from the consequences of already established aging. Common changes across different ways to slow aging may point to processes that allow delaying many age-related disorders at once. Even with a common source of age-related changes, existing damage may require separate repair. Multiple causes may yield to a single intervention if it affects several processes simultaneously. 🔗 Read original →

Mortality Gives Life Value, Argues Guardian Essay Ann Neumann's November 5 essay contrasts techno-immortality promises with the needs of the living. She critiques cryonics and Bryan Johnson's rejuvenation experiments as selling eternal youth while people nearby need care. Neumann draws on personal experience: in 2005, she left her job to care for her dying father, later working in a hospice. She observes that when the body fails, patients often want to die. Care, for her, means hearing those wishes and sharing final days. Astrophysicist Moiya McTier suggests death's inevitability drives risk-taking and growth; immortality jokes she would just lie on a couch forever. "Loss of loved ones binds communities," Neumann writes, linking this to mutual aid and resource sharing. Bioengineer Logan Collins, in a separate November 9 essay, defends extending healthy lifespan to preserve autonomy — the ability to direct one's own life. He argues we should treat aging's root causes like cancer: helping people maintain health and independence. Both perspectives, the piece concludes, can stem from respect for the person: hearing both who wishes to go and who wishes to stay. 🔗 Read original →

Vitalik Buterin Backs Slowing AI Data Center Build-Out On November 8, Ethereum co-founder Vitalik Buterin pushed back against sharp criticism of opponents to new AI data center construction. He argues that algorithmic improvements can partially offset slower compute growth, while the gained time allows strengthening defenses against cyber and biological threats. In Buterin's scenario, a task that would require larger, more numerous data centers in two years becomes feasible on current hardware in four — likely sooner. More efficient algorithms should deliver the same results with less compute. In a separate reply, he distinguishes slowing construction from denying people AI access. In researcher Adam Marblestone's model, more efficient software could accelerate AI development itself: existing computers can run more AI copies that search for and verify new algorithms, and successful improvements free up further compute for research. Buterin notes that the pace of AI-driven science depends on many other conditions. If work hits other bottlenecks, an AI delay may partly overlap with the time scientists would spend overcoming them anyway. Hence his prediction: "Slowing AI by one year will delay scientific research far less than one year." Buterin proposes spending the gained time on hardening computer systems and preparing for biological threats. In a January 2025 essay, he already advocated prioritizing defensive technologies — such as early detection of dangerous infections and software safety verification. Buterin has personally funded aging research. In February 2018, the SENS Research Foundation, which develops ways to repair age-related damage, announced a $2.4 million cryptocurrency donation from him. Now, life-extension advocate Linus Petersson asks: "Setting aside existential risk, what if this delays the moment aging is defeated?" Petersson argues developing aging therapies is far harder than deploying them, and government regulation is a necessary condition for development itself. If therapy arrival is delayed, some who could have benefited will die waiting. AI researcher Ben Goertzel agrees on the value of more efficient algorithms but considers slowing construction harmful overall. He notes that centers already approved or under construction will proceed, and U.S. restrictions risk losing ground in global competition. Goertzel adds that researchers also need compute to develop more efficient algorithms. 🔗 Read original →

Epigenetic clock shifts don't always match health changes — review of 26 studies A systematic review published on October 24 in Experimental Gerontology analyzed 26 human trials measuring both epigenetic clocks — algorithms based on DNA methylation marks — and health outcomes. The authors assessed whether clock changes align with health improvements after drugs, exercise, diet, surgery, and other interventions. Since tracking disease and mortality takes years, researchers seek faster biomarkers to evaluate anti-aging interventions. Different clocks are trained on distinct targets: some predict chronological age, others mortality risk or pace of aging. Consequently, a single blood sample can yield multiple, divergent age estimates. The review applied four validation criteria: 1) Does baseline clock value predict future disease or death? 2) Does it change post-intervention? 3) Do clock shifts correlate with health changes at the group level? 4) Can clock shifts reliably predict treatment benefit across independent randomized trials? Results were largely inconsistent. 15 studies showed mixed results — different clock–health pairs gave conflicting answers. In some pairs both improved; in others only one shifted detectably; opposite directions also occurred. Six studies could not assess concordance, and four found no statistically significant change in either measure. Only one study — a pre/post bariatric surgery trial with 20 participants — showed joint improvement across all pairs: the gap between epigenetic and chronological age narrowed at 12 months, alongside weight and glucose metabolism gains. No controlled trial demonstrated full concordance, and no study performed the fourth, cross-trial predictive validation. A key confounder is blood cell composition. Immune cell subtypes carry distinct methylation signatures. Interventions may alter cell proportions, shifting clock readouts without intracellular methylation changes. Muscle, brain, or kidney function may respond via different mechanisms and timelines. The authors caution: "Reducing epigenetic age should not be called 'reversal of aging' or 'clinical rejuvenation'." For future trials, they recommend pre-specifying one clock algorithm and one clinical endpoint (e.g., memory or physical function), measuring both repeatedly in the same participants with a control group, and subsequently testing predictive accuracy across independent trials. 🔗 Read original →

Aging Hallmarks Review Proposes Pre-Defined Benefit Tiers for Drug Trials Piotr Paweł Chmielewski and Bartłomiej Stępień of Wroclaw Medical University published a review of twelve aging hallmarks in International Journal of Molecular Sciences on September 11. The authors propose treating these hallmarks — such as DNA mutation accumulation and stem cell exhaustion — as preliminary causal modules: interacting process sets where intervention alters pre-selected physiological readouts through identifiable mechanisms. The 2013 Cell framework by Carlos López-Otín listed nine aging-linked processes with three criteria: manifestation with age, acceleration when worsened, and deceleration when ameliorated. The reviewers argue these criteria remain insufficient until "aging" is defined and the evaluation endpoint for interventions is specified. An age-associated change may be a cause, consequence, or compensatory response to damage. Each hallmark is evaluated by four questions: reproducible age-related change in human tissues; functional or survival impact upon modulation; effect persistence across conditions; and measurability and modifiability in humans. For clinical trials, the review defines six benefit tiers a drug can claim: target engagement (hit the intended process); biomarker shift (e.g., blood analyte); disease trajectory modification; function preservation (mobility, cognition); resilience (recovery from vaccine, surgery); and organismal aging deceleration (durable multi-system benefit). "Trials must state which tier they test and must not retroactively upgrade claims after a favorable biomarker result." The review cites two trials of RTB101, an mTOR inhibitor. Both showed enhanced antiviral gene expression in elderly (biomarker shift), but clinical outcomes diverged. In a Phase 2b trial with high-risk participants (including COPD patients and smokers), lab-confirmed respiratory infection occurred in 19% on 10 mg daily vs. 28% on placebo. In the subsequent Phase 3 trial excluding those groups, clinically apparent respiratory illness developed in 26% on drug vs. 25% on placebo. The review concludes trial outcome depends on participant selection and endpoint choice. Senolytics like dasatinib plus quercetin selectively clear senescent cells. In a trial of 60 postmenopausal women, the combination did not differ from control on the primary endpoint — a bone resorption marker. In an exploratory analysis, bone improvement concentrated in the top third with the highest circulating senescent cell marker. The review suggests such therapy may require confirmation of high senescent cell burden in patients. Hallmarks interact: inflammation, for example, can impair stem cell function. The authors conclude the framework "should generate testable systems-level hypotheses, not serve as a target list for drugs." 🔗 Read original →

Tel Aviv and Reichman Researchers Re-Analyze Fibroblast Methylation Data Researchers from Tel Aviv University and Reichman University re-analyzed open data on human skin fibroblasts. The preprint, dated November 3, compared two paths to proliferation arrest: months of continuous division versus X-ray-induced DNA damage. Cell identity relies on gene expression patterns regulated by DNA methylation. One aging hypothesis suggests DNA repair distracts regulatory proteins, causing epigenetic errors upon restoration. David Sinclair’s group previously tested this in mice with induced DNA breaks. The new analysis observed methylation "smoothing" in diseased tissues, correlating with liver fibrosis, though inflammation and cell turnover complicated those results. To isolate variables, the team used 24 culture sets with matched controls. One group divided until exhaustion; the other arrested soon after irradiation. This allowed comparison of methylation changes leading to the same non-dividing state. They first identified genomic regions where fibroblasts differed from 20 other cell types. During prolonged passaging, total divergence at these sites increased. However, most variable sites were shared with other continuously dividing cell lines — a "proliferation signature." The authors isolated several hundred fibroblast-specific sites for separate analysis. These two groups changed in opposite directions: the shared proliferation signature amplified differences from other cell types, while the fibroblast-specific signature smoothed toward the average of other types. The net metric masked the loss of cell identity. Weighting sites by their initial specificity, and subtracting background drift measured in control regions, the researchers calculated a 24.3% loss of original fibroblast identity at exhaustion. No excess loss beyond background was detected after X-ray irradiation. The authors propose that interventions be evaluated by comparing treated cells to controls that have undergone equal numbers of divisions, since therapies reducing damage often alter proliferation rates. bioRxiv preprint, November 3 🔗 Read original →

Aging persists even as the body repairs damage: biologist Vadim Gladyshev proposes a radical life extension strategy On November 6, a researcher at Harvard Medical School published a conceptual paper titled "Just Aging" in the journal PNAS. He distinguishes two tasks: extending lifespan by helping the most vulnerable organs, and overcoming species-specific limits tied to the organism's overall architecture. Copying DNA and producing necessary substances, cells make errors and create byproducts. The body repairs damage, removes waste, and renews cells; this maintenance itself is imperfect. The aggregate of remaining harmful changes Gladyshev calls the deleteriome: it includes damaged molecules, dysregulated cellular processes, and protective responses that become harmful over time. For example, chronic inflammation promotes new damage. In his model, as harm accumulates, organs struggle increasingly with physiological loads. Two organisms may accumulate similar damage yet live different lengths of time: one tolerates the burden longer and recovers better from failures. Therefore, enhancing resilience can extend lifespan even at the same rate of damage accumulation. Reducing damage formation, enhancing removal, and increasing tolerance of consequences are distinct therapeutic targets. "Aging clocks" estimate biological age via chemical marks on DNA, blood proteins, or other markers. Their signals may reflect both harmful changes and protective responses. Gladyshev therefore proposes validating treatments by comparing clock readouts with organ function, recovery from stress, and survival. In an October conversation with cardiologist Eric Topol, Gladyshev suggested tailoring therapy to the organ aging fastest — kidneys in one person, heart in another. In the new paper, he explains how supporting that organ brings average lifespan closer to the species maximum. Once one weak link is fixed, others become limiting; the maximum depends on the architecture of the body's many maintenance and renewal systems. To shift this limit, Gladyshev proposes combining slowed damage accumulation with periodic rejuvenation, regeneration — tissue restoration — or replacement of the most vulnerable parts: "Instead of repairing every component of the deleteriome, we can replace the most failure-prone parts with younger, functional ones." In his scenario, external devices could help cleanse blood, support respiration and circulation. They could operate in parallel with native organs as a backup. Failure of one part could then be survived thanks to another, and worn-out devices could be repaired or swapped. Replacing body organs preserves the continuously functioning brain, but its own aging becomes the next constraint. For the brain, Gladyshev envisions gradual prosthetics: devices could support or substitute specific functions, such as memory, while the brain keeps working. In this scenario, he links preservation of personal identity to the continuity of processes generating a person's sensations and experiences. 🔗 Read original →

Immunologist Derya Unutmaz Predicts AI Will Choose Scientific Questions Within Years Unutmaz, a professor at The Jackson Laboratory, outlined a five-stage evolution of science on November 7. He expects the transition to Science 4.0 — where superintelligent AI makes most discoveries under human direction — within a few years, lasting one to two decades. The forecast was prompted by the autonomous AutoDNA lab. Its AI agents receive a human goal, design experiments, and command robotic instruments. In DNA assembly tests, the system independently optimized reaction conditions to increase yield and cut reaction time, deciding each step without human intervention. Unutmaz defines Science 3.0 (last 2–3 years) as human–AI collaboration. In his November 4.0 stage, humans set priorities and validate results, but superintelligence performs the bulk of discovery. He predicts a further shift to Science 5.0 as the intelligence gap widens: humans will lack the capacity to frame the right questions or fully grasp the answers. Autonomous AI and robots will then drive nearly all scientific progress, including choosing research directions. Unutmaz calls his own generation "the last generation of human scientists." He hopes superintelligence will unlock answers about life, consciousness, and the nature of reality that remain beyond human reach. 🔗 Read original →