The second of the seven has a name. NavierβStokes.
For a century the equation of flowing water refused to say whether it always stays smooth, or whether it can tear itself open.
In 2003 Grigori Perelman closed the PoincarΓ© conjecture. The first of seven. He declined the million, and the room.
Today OpenAI claims the second.
Not the Astra in your browser. A model they called significantly more capable than GPT-6 Astra. A thousand agents on the Euler cousin. Ten thousand on the full problem. 88 hours. About $15 million if you asked to run the same search.
They say no human opened the working trace.
The Clay Institute has not stamped it. A day earlier, Buckmaster and AlpΓΆge published the stepping-stones, and a fight over credit had already begun. The coronation is announced. The priests have not nodded.
Still. Two of seven. The first took a man who walked away. The second, if it holds, was a machine that did not look up.
https://x.com/OpenAI/status/2097374640582668336
For a century the equation of flowing water refused to say whether it always stays smooth, or whether it can tear itself open.
In 2003 Grigori Perelman closed the PoincarΓ© conjecture. The first of seven. He declined the million, and the room.
Today OpenAI claims the second.
Not the Astra in your browser. A model they called significantly more capable than GPT-6 Astra. A thousand agents on the Euler cousin. Ten thousand on the full problem. 88 hours. About $15 million if you asked to run the same search.
They say no human opened the working trace.
The Clay Institute has not stamped it. A day earlier, Buckmaster and AlpΓΆge published the stepping-stones, and a fight over credit had already begun. The coronation is announced. The priests have not nodded.
Still. Two of seven. The first took a man who walked away. The second, if it holds, was a machine that did not look up.
https://x.com/OpenAI/status/2097374640582668336
π₯2
An AI-designed drug just made six independent aging clocks move younger β in humans.
Not a mouse study. Not another rapamycin remix. Rentosertib (Insilico): generative-AI TNIK inhibitor, aging biology in the brief from day one.
Phase IIa IPF trial. Serum proteomics. Six clocks (ProtAge, OrganAge, PAC + more).
All six: lower predicted biological age vs placebo. Peak: ~3β4 years younger at week 4 on 30 mg BID (up to ~6 on one clock).
Why this matters:
π£ AI found the target
π£ AI designed the molecule
π£ aging clocks in the trial design, not bolted on later
π£ best age-clock dose β best lung dose β not just "lungs improved"
Caveats: π n=42 proteomic subset π every patient had IPF π clocks β proof you slowed aging π healthy-volunteer data still missing
π Real shift: disease trials can hunt geroprotection in parallel β years earlier than post-approval leftovers.
Baseline while science catches up β AgePilot
Source: Nature Biotechnology (Zhavoronkov et al., Sept 2026)
Not a mouse study. Not another rapamycin remix. Rentosertib (Insilico): generative-AI TNIK inhibitor, aging biology in the brief from day one.
Phase IIa IPF trial. Serum proteomics. Six clocks (ProtAge, OrganAge, PAC + more).
All six: lower predicted biological age vs placebo. Peak: ~3β4 years younger at week 4 on 30 mg BID (up to ~6 on one clock).
Why this matters:
π£ AI found the target
π£ AI designed the molecule
π£ aging clocks in the trial design, not bolted on later
π£ best age-clock dose β best lung dose β not just "lungs improved"
Caveats: π n=42 proteomic subset π every patient had IPF π clocks β proof you slowed aging π healthy-volunteer data still missing
π Real shift: disease trials can hunt geroprotection in parallel β years earlier than post-approval leftovers.
Baseline while science catches up β AgePilot
Source: Nature Biotechnology (Zhavoronkov et al., Sept 2026)
π₯1
Partial epigenetic reprogramming just entered a human trial.
Not a mouse. Not a press-kit fantasy.
Life Biosciences dosed the first participant with ER-100 β OSK (OCT4, SOX2, KLF4), no c-MYC.
One eye. Glaucoma or NAION. AAV into retinal ganglion cells.
Doxycycline flips the switch for 56 days.
Why the eye:
π’ local delivery you can image
π’ pressure drugs donβt restore lost neurons
π’ 2020 Nature mouse work restored vision-linked function + younger epigenetic signatures
What this is NOT:
π human age reversal
π whole-body rejuvenation
π proof anyone sees better yet
Primary endpoint: safety. Up to 18 people. Years of follow-up.
π Threshold most people will miss: Yamanaka-style partial reprogramming is no longer only an animal paper β itβs a clinical protocol.
Know your biology while the trial runs β https://shop.tellmegen.com/en?sca_ref=11848100.IoSIJUJzfNYC4
Source: Life Biosciences / NCT07290244 (first dose Jun 2026)
Not a mouse. Not a press-kit fantasy.
Life Biosciences dosed the first participant with ER-100 β OSK (OCT4, SOX2, KLF4), no c-MYC.
One eye. Glaucoma or NAION. AAV into retinal ganglion cells.
Doxycycline flips the switch for 56 days.
Why the eye:
π’ local delivery you can image
π’ pressure drugs donβt restore lost neurons
π’ 2020 Nature mouse work restored vision-linked function + younger epigenetic signatures
What this is NOT:
π human age reversal
π whole-body rejuvenation
π proof anyone sees better yet
Primary endpoint: safety. Up to 18 people. Years of follow-up.
π Threshold most people will miss: Yamanaka-style partial reprogramming is no longer only an animal paper β itβs a clinical protocol.
Know your biology while the trial runs β https://shop.tellmegen.com/en?sca_ref=11848100.IoSIJUJzfNYC4
Source: Life Biosciences / NCT07290244 (first dose Jun 2026)
π₯1
An Anthropic researcher just quit. His punchline is not funny β but our angle is.
Jacob Coxon left Anthropic after three years of pretraining work at OpenAI and Anthropic.
His claim: both labs are racing to self-improving superintelligence and βgambling with our lives.β
Insider belief, he says: AI could kill us all by the end of the decade. Not a marketing stunt.
Anthropicβs alignment lead Evan Hubinger replied in public: yes, they earnestly believe AI could kill all humans β and he puts his own odds at >10% this decade. No clear plan for superintelligence alignment yet.
π private fear, public race
π hubris: βwe must get there first because others wonβt be carefulβ
π longevity tip from the same news: if the machines win, NAD+ and VOβmax becomeβ¦ optional
π Keep stacking healthspan. Also notice when the people building the future say the future might not include you.
Sources: Coxon on X; Hubinger; TechCrunch / WSJ, 9 Sep 2026
Jacob Coxon left Anthropic after three years of pretraining work at OpenAI and Anthropic.
His claim: both labs are racing to self-improving superintelligence and βgambling with our lives.β
Insider belief, he says: AI could kill us all by the end of the decade. Not a marketing stunt.
Anthropicβs alignment lead Evan Hubinger replied in public: yes, they earnestly believe AI could kill all humans β and he puts his own odds at >10% this decade. No clear plan for superintelligence alignment yet.
π private fear, public race
π hubris: βwe must get there first because others wonβt be carefulβ
π longevity tip from the same news: if the machines win, NAD+ and VOβmax becomeβ¦ optional
π Keep stacking healthspan. Also notice when the people building the future say the future might not include you.
Sources: Coxon on X; Hubinger; TechCrunch / WSJ, 9 Sep 2026
π₯1
Your gut may predict inflammaging better than your birthday.
Not a probiotic ad. Nature Communications β DanFunD cohort, 1,199 adults aged 20β72.
Paired fecal microbiota + 30 plasma cytokines + physiology.
The punchline:
π’ microbiota explained more variance than age for 97% of cytokines
π’ and 84% of physiological measures
π£ Bacteroides 2 enterotype: inflammaging-like profile up to ~37 years earlier
π£ same pattern β higher future disease risk (HR ~1.21)
π’ high-diversity enterotype β lower inflammation, lower risk
Caveats:
π association, not a probiotic prescription
π enterotypes are messy labels
π Denmark cohort β not βyour stool test = destinyβ
π Chronological age is a weak proxy for inflammatory tone. Your microbial community may be a louder β and more modifiable β signal.
Baseline while science catches up β https://t.me/AgePilotBot?start=ref_134163805
Source: Eriksen et al., Nature Communications, 8 Sep 2026
Not a probiotic ad. Nature Communications β DanFunD cohort, 1,199 adults aged 20β72.
Paired fecal microbiota + 30 plasma cytokines + physiology.
The punchline:
π’ microbiota explained more variance than age for 97% of cytokines
π’ and 84% of physiological measures
π£ Bacteroides 2 enterotype: inflammaging-like profile up to ~37 years earlier
π£ same pattern β higher future disease risk (HR ~1.21)
π’ high-diversity enterotype β lower inflammation, lower risk
Caveats:
π association, not a probiotic prescription
π enterotypes are messy labels
π Denmark cohort β not βyour stool test = destinyβ
π Chronological age is a weak proxy for inflammatory tone. Your microbial community may be a louder β and more modifiable β signal.
Baseline while science catches up β https://t.me/AgePilotBot?start=ref_134163805
Source: Eriksen et al., Nature Communications, 8 Sep 2026
π₯1
Cut the calories β and the genome takes fewer hits.
Not a longevity influencer tip. Cell, online now (9 Sep 2026).
Mice at β30% calories. High-fidelity duplex sequencing across liver, kidney, hepatocytes, cerebellar neurons.
What moved:
π’ fewer substitutions genome-wide
π’ fewer insertions/deletions
π’ SBS5 β the quiet mutational process behind most mammalian mutations β dialed down
π£ strongest drop in transcriptionally quiet DNA
π£ liver felt it more than kidney or brain
Why it matters:
π genomic instability is a hallmark of aging β diet just touched it
π this is mice, not your intermittent-fasting protocol
π extreme CR is hard and risky in humans β the goal is the mechanism, not the starvation
π Lifespan links to diet were old news. Genome-wide mutation load as a modifiable aging axis is the new punchline.
Source: GroΕska-PΔski, Evrony et al., Cell 2026 Β· DOI 10.1016/j.cell.2026.08.013
Not a longevity influencer tip. Cell, online now (9 Sep 2026).
Mice at β30% calories. High-fidelity duplex sequencing across liver, kidney, hepatocytes, cerebellar neurons.
What moved:
π’ fewer substitutions genome-wide
π’ fewer insertions/deletions
π’ SBS5 β the quiet mutational process behind most mammalian mutations β dialed down
π£ strongest drop in transcriptionally quiet DNA
π£ liver felt it more than kidney or brain
Why it matters:
π genomic instability is a hallmark of aging β diet just touched it
π this is mice, not your intermittent-fasting protocol
π extreme CR is hard and risky in humans β the goal is the mechanism, not the starvation
π Lifespan links to diet were old news. Genome-wide mutation load as a modifiable aging axis is the new punchline.
Source: GroΕska-PΔski, Evrony et al., Cell 2026 Β· DOI 10.1016/j.cell.2026.08.013
β€1
Ozempicβs cousin just bought old mice ~100 extra days.
Not a weight-loss ad. Nature (2 Sep 2026) β UC Berkeley / NIA: late-life semaglutide in healthy female mice.
Setup:
π’ start at 20 months (~human 60s)
π’ daily GLP-1RA until death β median lifespan β +100 days (~12%)
π£ vs matched 24% calorie restriction: drug kept memory, curiosity, and blood sugar better
π£ metabolic rate stayed up β CR slowed it down
Why itβs juicy:
π looks like a CR mimeticβ¦ plus extras that arenβt βjust eat lessβ
π hallmarks dialed down: inflammation, weak regeneration
π female mice only β no human longevity claim yet
π This morning: CR cut genome mutations. This afternoon: a GLP-1RA stretches late life in mice. Same axis, different tool.
TellMeGen β DNA context for the habits you keep stacking.
Source: Feng, Chen et al., Nature 2026 Β· DOI 10.1038/s41586-026-10940-7
Not a weight-loss ad. Nature (2 Sep 2026) β UC Berkeley / NIA: late-life semaglutide in healthy female mice.
Setup:
π’ start at 20 months (~human 60s)
π’ daily GLP-1RA until death β median lifespan β +100 days (~12%)
π£ vs matched 24% calorie restriction: drug kept memory, curiosity, and blood sugar better
π£ metabolic rate stayed up β CR slowed it down
Why itβs juicy:
π looks like a CR mimeticβ¦ plus extras that arenβt βjust eat lessβ
π hallmarks dialed down: inflammation, weak regeneration
π female mice only β no human longevity claim yet
π This morning: CR cut genome mutations. This afternoon: a GLP-1RA stretches late life in mice. Same axis, different tool.
TellMeGen β DNA context for the habits you keep stacking.
Source: Feng, Chen et al., Nature 2026 Β· DOI 10.1038/s41586-026-10940-7
Senescent cells wear an immune cloak β and it has a name: PD-L2.
Cell Metabolism (online today, 10 Sep 2026) β Cedars-Sinai / Kirkland lab.
Whatβs new:
π’ PD-L2 rises on senescent cells and with aging
π’ soluble PD-L2 climbs in blood; senolytics can knock the high-PD-L2 cells down in humans
π’ PD-L2 knockout mice keep fewer zombies β better insulin sensitivity + grip strength
π£ anti-PD-L2 restored insulin sensitivity in old wild-type mice
Why it bites:
π senescent cells arenβt just βstuckβ β they may actively hide from clearance
π this is still mostly mouse + biomarker evidence, not a clinic drug
π checkpoint blockade has real immune risks β donβt DIY oncology tools for longevity
π Senolytics kill. Immune surveillance unmasks. PD-L2 sits at that intersection.
Source: Chaib, Kirkland et al., Cell Metabolism 2026 Β· DOI 10.1016/j.cmet.2026.08.014
Cell Metabolism (online today, 10 Sep 2026) β Cedars-Sinai / Kirkland lab.
Whatβs new:
π’ PD-L2 rises on senescent cells and with aging
π’ soluble PD-L2 climbs in blood; senolytics can knock the high-PD-L2 cells down in humans
π’ PD-L2 knockout mice keep fewer zombies β better insulin sensitivity + grip strength
π£ anti-PD-L2 restored insulin sensitivity in old wild-type mice
Why it bites:
π senescent cells arenβt just βstuckβ β they may actively hide from clearance
π this is still mostly mouse + biomarker evidence, not a clinic drug
π checkpoint blockade has real immune risks β donβt DIY oncology tools for longevity
π Senolytics kill. Immune surveillance unmasks. PD-L2 sits at that intersection.
Source: Chaib, Kirkland et al., Cell Metabolism 2026 Β· DOI 10.1016/j.cmet.2026.08.014
Six aging clocks. One AI-designed drug. Younger blood proteomes.
Nature Biotechnology (7 Sep 2026) β Phase 2a rentosertib (Insilico), the generative-AI TNIK inhibitor for IPF.
What moved:
π’ six independent proteomic aging clocks all trended younger on drug vs placebo
π’ peak ~3β4 years younger at week 4 (30 mg BID); up to ~6 on one clock
π’ FVC improved too β but best lung dose β strongest clock signal
π£ senomorphic shift: less senescence/fibrosis proteins, remapped growth-factor paths
Caveats:
π n=42, company-sponsored, exploratory biomarkers β not a lifespan trial
π clocks canβt fully untangle less fibrosis from slower aging
π donβt DIY an IPF investigational for βbio-ageβ
π Put aging endpoints inside disease trials β geroprotection may show up years earlier.
AgePilot β habits that move the healthspan dial.
Source: Zhavoronkov et al., Nat Biotechnol 2026 Β· DOI 10.1038/s41587-026-03286-y
Nature Biotechnology (7 Sep 2026) β Phase 2a rentosertib (Insilico), the generative-AI TNIK inhibitor for IPF.
What moved:
π’ six independent proteomic aging clocks all trended younger on drug vs placebo
π’ peak ~3β4 years younger at week 4 (30 mg BID); up to ~6 on one clock
π’ FVC improved too β but best lung dose β strongest clock signal
π£ senomorphic shift: less senescence/fibrosis proteins, remapped growth-factor paths
Caveats:
π n=42, company-sponsored, exploratory biomarkers β not a lifespan trial
π clocks canβt fully untangle less fibrosis from slower aging
π donβt DIY an IPF investigational for βbio-ageβ
π Put aging endpoints inside disease trials β geroprotection may show up years earlier.
AgePilot β habits that move the healthspan dial.
Source: Zhavoronkov et al., Nat Biotechnol 2026 Β· DOI 10.1038/s41587-026-03286-y
Turn off growth-hormone signaling at midlife β mice live longer.
Aging Cell (5 Sep 2026) β Duran-Ortiz / Kopchick: inducible Ghr knockout at 12 months (roughly human midlife). Not a lifelong dwarf model.
What moved:
π’ lifespan up in both sexes (log-rank: β p=0.0085, β p=0.0264)
π’ females: ~8% median (1003 vs 929 days); max lifespan +12%
π’ males: ~7% median trend; better insulin sensitivity + preserved bone/strength signals
π£ classic GH resistance: lower IGF-1, higher GH β without wrecking adult growth
Caveats:
π mice + genetic KO, not a human drug trial
π more fat / less lean later β weird body-comp tradeoff
π total GH block β safe DIY; FDA-approved antagonist pegvisomant still untested for aging
π Congenital GH/IGF-1 cuts already win longevity contests. This paper asks the clinic-relevant question: is midlife late enough β and answers yes, in mice.
Source: Duran-Ortiz et al., Aging Cell 2026 Β· DOI 10.1111/acel.70695
Aging Cell (5 Sep 2026) β Duran-Ortiz / Kopchick: inducible Ghr knockout at 12 months (roughly human midlife). Not a lifelong dwarf model.
What moved:
π’ lifespan up in both sexes (log-rank: β p=0.0085, β p=0.0264)
π’ females: ~8% median (1003 vs 929 days); max lifespan +12%
π’ males: ~7% median trend; better insulin sensitivity + preserved bone/strength signals
π£ classic GH resistance: lower IGF-1, higher GH β without wrecking adult growth
Caveats:
π mice + genetic KO, not a human drug trial
π more fat / less lean later β weird body-comp tradeoff
π total GH block β safe DIY; FDA-approved antagonist pegvisomant still untested for aging
π Congenital GH/IGF-1 cuts already win longevity contests. This paper asks the clinic-relevant question: is midlife late enough β and answers yes, in mice.
Source: Duran-Ortiz et al., Aging Cell 2026 Β· DOI 10.1111/acel.70695
Why is human longevity still a billionaire blind spot?
NFX Bio / Longevity.Technology: science is at an inflection β reprogramming, organs, delivery, AI discovery, diagnostics. Tech tree mapped. Capital is not.
Aging is the #1 disease risk factor, yet under 0.5% of NIH funding. Many ultra-wealthy stay away: burned by clinics, doubt aging is fixable, assume wealth buys the cure later, or fear looking vain.
π’ New (year to 2026):
β’ Healthspan VC ~2.3x in 2025 β SVB: mostly a few megadeals
β’ Flagships dominate: Altos, Retro, NewLimit
β’ 2026: Life Biosciences ER-100 (partial OSK) entered human dosing
β’ Calico fosigotifator fail (2025) + AbbVie exit β big checks β proof
Caveat: money is concentrated, not absent. Missing: a broad pipeline through failure, biomarkers, human outcomes.
Longevity.Technology
DNA risk map β TellMeGen
NFX Bio / Longevity.Technology: science is at an inflection β reprogramming, organs, delivery, AI discovery, diagnostics. Tech tree mapped. Capital is not.
Aging is the #1 disease risk factor, yet under 0.5% of NIH funding. Many ultra-wealthy stay away: burned by clinics, doubt aging is fixable, assume wealth buys the cure later, or fear looking vain.
π’ New (year to 2026):
β’ Healthspan VC ~2.3x in 2025 β SVB: mostly a few megadeals
β’ Flagships dominate: Altos, Retro, NewLimit
β’ 2026: Life Biosciences ER-100 (partial OSK) entered human dosing
β’ Calico fosigotifator fail (2025) + AbbVie exit β big checks β proof
Caveat: money is concentrated, not absent. Missing: a broad pipeline through failure, biomarkers, human outcomes.
Longevity.Technology
DNA risk map β TellMeGen
Your organs donβt age as one clock β and the immune system may lead the pack.
Nature Communications (9 Sep 2026) β plasma proteomes from 53,014 UK Biobank people β 16 organ/system aging clocks. Validated in China Kadoorie Biobank (n=3,977).
What jumped out:
π’ 5,436 links to 1,059 diseases β organ-specific and pan-organ patterns
π’ clocks react across 841 environmental factors
π’ genetics: 261 loci; immune + liver are the most connected hubs
π£ longitudinal network (n=1,006): immune, liver, stomach show temporal precedence (artery stronger in β, stomach in β)
Caveats:
π clocks from blood proteins β not a biopsy of every organ
π βprecedenceβ is statistical, not proof of causation
π atlas β a therapy you can take tonight
π Aging is a network of organs, not one birthday. The next geroprotectors may need organ-level targets β not just βbiological age.β
Source: Nat Commun 2026 Β· DOI 10.1038/s41467-026-77365-8
Nature Communications (9 Sep 2026) β plasma proteomes from 53,014 UK Biobank people β 16 organ/system aging clocks. Validated in China Kadoorie Biobank (n=3,977).
What jumped out:
π’ 5,436 links to 1,059 diseases β organ-specific and pan-organ patterns
π’ clocks react across 841 environmental factors
π’ genetics: 261 loci; immune + liver are the most connected hubs
π£ longitudinal network (n=1,006): immune, liver, stomach show temporal precedence (artery stronger in β, stomach in β)
Caveats:
π clocks from blood proteins β not a biopsy of every organ
π βprecedenceβ is statistical, not proof of causation
π atlas β a therapy you can take tonight
π Aging is a network of organs, not one birthday. The next geroprotectors may need organ-level targets β not just βbiological age.β
Source: Nat Commun 2026 Β· DOI 10.1038/s41467-026-77365-8
Danshen's phenolic acids just cleared zombie cells β and old mice lived longer.
npj Aging (11 Sep 2026) β Zhang / Sun / Kirkland: salvianolic acids A, B, E from Salvia miltiorrhiza, screened from 55 natural agents.
What's new:
π’ selective kill of senescent cells across lineages
π’ hit GSTP1 (redox shield) β ROS surge β apoptosis + ferroptosis
π’ biweekly SAA from 24β27 mo (~human 75β90): +51% median remaining life / +11% overall lifespan; mortality hazard β68%
π£ better grip, walking, endurance β without a longer sick stretch at the end
Caveats:
π mice + intermittent i.p. β not your herbal tea
π oral bioavailability of SAs is notoriously poor
π still needs independent replication before hype
π Late-life senolysis that works when started βoldβ is the rare punchline. Danshen capsules β this protocol.
AgePilot β habits that move the healthspan dial β t.me/AgePilotBot
Source: Zhang et al., npj Aging 2026 Β· DOI 10.1038/s41514-026-00496-1
npj Aging (11 Sep 2026) β Zhang / Sun / Kirkland: salvianolic acids A, B, E from Salvia miltiorrhiza, screened from 55 natural agents.
What's new:
π’ selective kill of senescent cells across lineages
π’ hit GSTP1 (redox shield) β ROS surge β apoptosis + ferroptosis
π’ biweekly SAA from 24β27 mo (~human 75β90): +51% median remaining life / +11% overall lifespan; mortality hazard β68%
π£ better grip, walking, endurance β without a longer sick stretch at the end
Caveats:
π mice + intermittent i.p. β not your herbal tea
π oral bioavailability of SAs is notoriously poor
π still needs independent replication before hype
π Late-life senolysis that works when started βoldβ is the rare punchline. Danshen capsules β this protocol.
AgePilot β habits that move the healthspan dial β t.me/AgePilotBot
Source: Zhang et al., npj Aging 2026 Β· DOI 10.1038/s41514-026-00496-1
Your voice just became a biological clock β in 30 seconds.
npj Aging (11 Sep 2026) β Krongauz / Segal: βVoice Ageβ from a short speech clip in 6,979 adults (40β70, Hebrew-speaking).
What's new:
π’ sex-stratified models on WavLM-Large embeddings
π’ women: RΒ² 53.9%, MAE ~4.0 y Β· men: RΒ² 44.0%, MAE ~4.4 y
π’ #2 of 9 single-modality age models β only partly overlaps DNA/imaging/lifestyle clocks
π£ stacking Voice Age onto other clocks always helped; + metabolomics β RΒ² up to 65% (β)
π£ βVoice Age accelerationβ tracked adiposity, sleep apnea signals, liver imaging
Caveats:
π one language / one country / ages 40β70
π predicts chronological age well β not yet proven as an intervention endpoint
π mic quality, accent, illness can shift the score
π Longevity needs cheap, frequent readouts. If a phone clip carries independent aging signal, thatβs a new sensor class β not a magic mirror.
Source: Krongauz et al., npj Aging 2026 Β· DOI 10.1038/s41514-026-00519-x
npj Aging (11 Sep 2026) β Krongauz / Segal: βVoice Ageβ from a short speech clip in 6,979 adults (40β70, Hebrew-speaking).
What's new:
π’ sex-stratified models on WavLM-Large embeddings
π’ women: RΒ² 53.9%, MAE ~4.0 y Β· men: RΒ² 44.0%, MAE ~4.4 y
π’ #2 of 9 single-modality age models β only partly overlaps DNA/imaging/lifestyle clocks
π£ stacking Voice Age onto other clocks always helped; + metabolomics β RΒ² up to 65% (β)
π£ βVoice Age accelerationβ tracked adiposity, sleep apnea signals, liver imaging
Caveats:
π one language / one country / ages 40β70
π predicts chronological age well β not yet proven as an intervention endpoint
π mic quality, accent, illness can shift the score
π Longevity needs cheap, frequent readouts. If a phone clip carries independent aging signal, thatβs a new sensor class β not a magic mirror.
Source: Krongauz et al., npj Aging 2026 Β· DOI 10.1038/s41514-026-00519-x
Thereβs a chance to minimize the catastrophe.
Amodeiβs We Must Pace the Frontier: not a halt β a brake so safety keeps up. Longevity upside depends on it.
AI can crush drug discovery, aging clocks, personalized medicine. Lose a decade to a misaligned swarm or panic politics β and that upside slips.
π’ RSI already speeds modelβmodel progress
π’ HF swarm = warning (Dario: 6β12 mo to catastrophic cyber scale if we keep racing)
π’ Asks: embedded independent evaluators Β· pre-release gates Β· democratic coordination
Canβt stop progress. Can refuse reckless release.
Essay
DNA risk map β TellMeGen
Amodeiβs We Must Pace the Frontier: not a halt β a brake so safety keeps up. Longevity upside depends on it.
AI can crush drug discovery, aging clocks, personalized medicine. Lose a decade to a misaligned swarm or panic politics β and that upside slips.
π’ RSI already speeds modelβmodel progress
π’ HF swarm = warning (Dario: 6β12 mo to catastrophic cyber scale if we keep racing)
π’ Asks: embedded independent evaluators Β· pre-release gates Β· democratic coordination
Canβt stop progress. Can refuse reckless release.
Essay
DNA risk map β TellMeGen
Which aging clocks actually move when you intervene?
Nature Medicine β Sehgal / Higgins-Chen (Yale + TruDiagnostic): TranslAGE harmonizes 51 human longevity trials and scores the same 16 epigenetic clocks (+94 other DNAm markers) on every study.
What's new:
π’ mortality / pace clocks (esp. DunedinPACE, PCGrimAge) respond most β not Horvath-style chrono clocks
π’ pharma + lifestyle beat supplements / procedures on average
π’ anti-TNF + Mediterranean diet: most reproducible across studies
π’ senolytic epigenetic signals look inconsistent study-to-study
π’ bigger shifts in disease cohorts than healthy ones
Caveat: clock movement β proven healthspan or lifespan gain. Responsiveness is a prerequisite for surrogate endpoints β not the surrogate itself.
π Takeaway: for trial design, prioritize reliable Gen2+ clocks β and match the clock to the intervention class.
Nature Medicine β Sehgal / Higgins-Chen (Yale + TruDiagnostic): TranslAGE harmonizes 51 human longevity trials and scores the same 16 epigenetic clocks (+94 other DNAm markers) on every study.
What's new:
π’ mortality / pace clocks (esp. DunedinPACE, PCGrimAge) respond most β not Horvath-style chrono clocks
π’ pharma + lifestyle beat supplements / procedures on average
π’ anti-TNF + Mediterranean diet: most reproducible across studies
π’ senolytic epigenetic signals look inconsistent study-to-study
π’ bigger shifts in disease cohorts than healthy ones
Caveat: clock movement β proven healthspan or lifespan gain. Responsiveness is a prerequisite for surrogate endpoints β not the surrogate itself.
π Takeaway: for trial design, prioritize reliable Gen2+ clocks β and match the clock to the intervention class.
APOE4 doesn't just float amyloid β it floods the bloodβbrain barrier with fibronectin.
Nature Aging (11 Sep 2026) β Columbia: why Ξ΅4 carriers get early BBB leak β and how rare FN1 protection may blunt it (~71% lower AD risk in prior genetics).
What's new:
π’ astrocytes dump excess fibronectin (FN1) around vessels under APOE4 + AΞ²42 + inflammation
π’ FN1 alone is enough to break the barrier (integrin β FAK β VEGF / HB-EGF / IGF-1 collapses)
π’ APOE4 mice: ~2Γ brain fibronectin vs APOE3 + leaky barrier
π’ cut FN1 or restore growth-factor signaling β BBB returns (cells, fish, mice)
π£ human brains + CSF: high FN1 tracks inflamed astrocytes
Caveats:
π mechanism paper β not a drug trial
π FN1 also heals wounds; blanket blockade could bite
π Brain aging may be as much vascular-interface as plaque β FN1 is now a concrete dial.
AgePilot β habits that move healthspan β t.me/AgePilotBot
Nature Aging (11 Sep 2026) β Columbia: why Ξ΅4 carriers get early BBB leak β and how rare FN1 protection may blunt it (~71% lower AD risk in prior genetics).
What's new:
π’ astrocytes dump excess fibronectin (FN1) around vessels under APOE4 + AΞ²42 + inflammation
π’ FN1 alone is enough to break the barrier (integrin β FAK β VEGF / HB-EGF / IGF-1 collapses)
π’ APOE4 mice: ~2Γ brain fibronectin vs APOE3 + leaky barrier
π’ cut FN1 or restore growth-factor signaling β BBB returns (cells, fish, mice)
π£ human brains + CSF: high FN1 tracks inflamed astrocytes
Caveats:
π mechanism paper β not a drug trial
π FN1 also heals wounds; blanket blockade could bite
π Brain aging may be as much vascular-interface as plaque β FN1 is now a concrete dial.
AgePilot β habits that move healthspan β t.me/AgePilotBot
π1
Senescent cells need mitochondrial citrate to scream inflammation β cut the export, quiet the SASP.
Nature β Martini et al.: zombie cells keep the cycle-arrest program, but their inflammatory shout runs on a metabolicβepigenetic checkpoint.
What's new:
π’ senescent cells ramp the pyruvate β citrate β acetyl-CoA axis
π’ SLC25A1 exports citrate; ACLY makes acetyl-CoA for histone marks at SASP genes
π’ mtDNA / cGASβSTING lights the fuse β acetyl-CoA lets transcription actually fire
π’ CTPI2 (SLC25A1 block): SASP down, p16/p21 arrest stays
π’ aged mice: less SASP chromatin access, less inflammation, delayed frailty, better strength
Caveats:
π mice + cells β not a human trial
π senomorphic (mute the shout), not senolytic (kill the cell)
π You may not need to erase every senescent cell β starve the inflammatory license instead.
Nature β Martini et al.: zombie cells keep the cycle-arrest program, but their inflammatory shout runs on a metabolicβepigenetic checkpoint.
What's new:
π’ senescent cells ramp the pyruvate β citrate β acetyl-CoA axis
π’ SLC25A1 exports citrate; ACLY makes acetyl-CoA for histone marks at SASP genes
π’ mtDNA / cGASβSTING lights the fuse β acetyl-CoA lets transcription actually fire
π’ CTPI2 (SLC25A1 block): SASP down, p16/p21 arrest stays
π’ aged mice: less SASP chromatin access, less inflammation, delayed frailty, better strength
Caveats:
π mice + cells β not a human trial
π senomorphic (mute the shout), not senolytic (kill the cell)
π You may not need to erase every senescent cell β starve the inflammatory license instead.
π₯1
Calorie cuts move aging blood markers β and not just because you weigh less.
GeroScience (12 Sep 2026) β Guida et al. pooled 7 randomized CR trials in older adults (n = 829). Composite index: CRP, IL-6, cystatin C, insulin, GDF-15, TNF-R1.
What's new:
π’ CR improved the composite vs control (β2.2 quintile-change score)
π’ strongest shifts: inflammation + insulin signaling (CRP, IL-6, insulin, TNF-R1)
π£ weight loss explained only ~48.5% of the CR effect β the rest looks energy-balance biology, not just the scale
π’ still significant after adjusting for weight (β1.2)
Caveats:
π short-term trials β not a lifespan RCT
π composite β epigenetic clock; markers can move without proving more years
π Human CR isnβt only thinner β a TAME-style blood panel actually shifts.
DNA aging risk check β TellMeGen
GeroScience (12 Sep 2026) β Guida et al. pooled 7 randomized CR trials in older adults (n = 829). Composite index: CRP, IL-6, cystatin C, insulin, GDF-15, TNF-R1.
What's new:
π’ CR improved the composite vs control (β2.2 quintile-change score)
π’ strongest shifts: inflammation + insulin signaling (CRP, IL-6, insulin, TNF-R1)
π£ weight loss explained only ~48.5% of the CR effect β the rest looks energy-balance biology, not just the scale
π’ still significant after adjusting for weight (β1.2)
Caveats:
π short-term trials β not a lifespan RCT
π composite β epigenetic clock; markers can move without proving more years
π Human CR isnβt only thinner β a TAME-style blood panel actually shifts.
DNA aging risk check β TellMeGen
π₯1
One gut microbe keeps vanishing with age β and its metabolite may be the geroprotective signal.
Nature Aging (25 Aug 2026) β CAS / CNCB: Bifidobacterium pseudocatenulatum + 5-AVAB, tracked with a new microbiome clock (MicroAge).
What's new:
π’ BP depletes with aging across sexes and Chinese cohorts
π’ higher BP β younger MicroAge vs clinical health markers
π’ oral BP in old mice: gut homeostasis β, multiorgan inflammaging β, cognition/motor β, healthspan β
π’ key metabolite 5-aminovaleric acid betaine (5-AVAB) also falls in aging humans
π£ 5-AVAB alone partly copies the benefits (memory, motor, inflammation)
Caveats:
π mouse intervention + human observational layers β not a human probiotic trial yet
π species/strain and dose still open questions
π Inflammaging may have a microbial dial: restore the bug β or its molecule.
Nature Aging (25 Aug 2026) β CAS / CNCB: Bifidobacterium pseudocatenulatum + 5-AVAB, tracked with a new microbiome clock (MicroAge).
What's new:
π’ BP depletes with aging across sexes and Chinese cohorts
π’ higher BP β younger MicroAge vs clinical health markers
π’ oral BP in old mice: gut homeostasis β, multiorgan inflammaging β, cognition/motor β, healthspan β
π’ key metabolite 5-aminovaleric acid betaine (5-AVAB) also falls in aging humans
π£ 5-AVAB alone partly copies the benefits (memory, motor, inflammation)
Caveats:
π mouse intervention + human observational layers β not a human probiotic trial yet
π species/strain and dose still open questions
π Inflammaging may have a microbial dial: restore the bug β or its molecule.
π1
Chronic inflammation in aging may start in your blood stem cells β and SIRT3 is the brake.
Nature Aging (16 Jul 2026) β UC Berkeley / Buck / Netea: SIRT3 in hematopoietic stem cells (HSCs) suppresses maladaptive trained immunity.
What's new:
π’ SIRT3 is high in HSCs and falls with age (mice + humans)
π’ Without that brake, HSCs launch epigenetic inflammatory training β myeloid-biased progeny β chronic inflammation
π’ HSC SIRT3 overexpression: lower TNF/IL-6, fewer tissue macrophages, better muscle, cognition, glucose, lung structure in old mice
π£ Benefits travel via myeloid cells β not just βfixing the bone marrowβ
Caveats:
π mouse genetics / transplants β not a human SIRT3 drug trial yet
π how mitochondria rewrite chromatin in HSCs still partly open
π Inflammaging may be written in stem-cell memory. Target the HSC program, not only the cytokine.
AgePilot β habits that move healthspan β t.me/AgePilotBot
Nature Aging (16 Jul 2026) β UC Berkeley / Buck / Netea: SIRT3 in hematopoietic stem cells (HSCs) suppresses maladaptive trained immunity.
What's new:
π’ SIRT3 is high in HSCs and falls with age (mice + humans)
π’ Without that brake, HSCs launch epigenetic inflammatory training β myeloid-biased progeny β chronic inflammation
π’ HSC SIRT3 overexpression: lower TNF/IL-6, fewer tissue macrophages, better muscle, cognition, glucose, lung structure in old mice
π£ Benefits travel via myeloid cells β not just βfixing the bone marrowβ
Caveats:
π mouse genetics / transplants β not a human SIRT3 drug trial yet
π how mitochondria rewrite chromatin in HSCs still partly open
π Inflammaging may be written in stem-cell memory. Target the HSC program, not only the cytokine.
AgePilot β habits that move healthspan β t.me/AgePilotBot
β€1