Finn Brooks: Tell me if this is too bleak a place to start — but I've been thinking about my grandmother this week and how fast the last two years went. Like, not slow and steady. Fast.
Clara Bennett: No, that's — that's exactly the right place to start, actually. Because what you're describing isn't just grief. It's real biology. The decline wasn't linear.
Finn Brooks: Right — and I didn't have the vocabulary for it until I started reading about this framework. There's a paper, June 6, 2013, in Cell — 'The Hallmarks of Aging' — and the team behind it is Carlos López-Otín from the University of Oviedo, Maria A. Blasco from the Spanish National Cancer Research Centre, Linda Partridge, Manuel Serrano, Guido Kroemer. Five people who basically handed biology a unified theory of why we fall apart.
Clara Bennett: And what makes it a framework rather than just a list — the key is the feedback architecture. The nine hallmarks aren't running separately. Each one damages the others.
Finn Brooks: Nine. Okay — genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, deregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion, altered intercellular communication. I've said that list maybe forty times this week and it still feels like a lot.
Clara Bennett: It is a lot. But the reason to hold all nine in your head at once is that they're — the best way I can frame it — they're locked in a cycle. Genomic instability drives senescence. Senescent cells release inflammatory signals. Those signals worsen mitochondrial dysfunction. Mitochondrial dysfunction generates more DNA damage. You're back at genomic instability.
Finn Brooks: So it spirals.
Clara Bennett: It spirals. Which is why a 70-year-old declines faster than a 40-year-old — not because more things are broken, but because the broken things are actively worsening each other. The rate accelerates.
Finn Brooks: And that — I mean, that reframes the whole question, doesn't it? Because if it's an accelerating system, then asking 'what causes aging' is almost the wrong question. The question is: why does the rate itself speed up?
Clara Bennett: That's the question, and it's actually answerable — in practice, the rate speeds up because the loops don't just add damage, they multiply each other's damage. Let me give you the plain version first. Imagine your body's repair crews are all stretched thin, so each one starts borrowing workers from the others. Now every repair job gets worse simultaneously. That's it. That's the whole thing.
Finn Brooks: Wait — simultaneously? Like they're all degrading in parallel, not in sequence?
Clara Bennett: In parallel, and each one is actively feeding the others. Now here's a concrete chain so it stops being abstract — genomic instability pushes damaged cells into senescence. Those senescent cells don't just sit there, they release what's called the SASP — the senescence-associated secretory phenotype — which is essentially a cocktail of inflammatory signals. Those signals land on mitochondria and worsen their dysfunction. And dysfunctional mitochondria generate reactive oxygen species, which cause more DNA damage. You are back at genomic instability. The loop is closed.
Finn Brooks: So the SASP is the thing doing the long-distance damage. Like senescent cells are almost — they're broadcasting?
Clara Bennett: Exactly that. They're metabolically active, still secreting, still signaling. That's why Maria Blasco's work on telomere attrition matters so much here — shortened telomeres trigger senescence in the first place, so you get more SASP-broadcasting cells, which then accelerate damage elsewhere. The entry point of the loop matters.
Finn Brooks: Okay but — hang on — the 2022 Copenhagen meeting added disabled macroautophagy as a new hallmark. And autophagy is basically the cell's garbage collection, right? So is that doing the same loop thing?
Clara Bennett: Yes, and this is the part that's — I mean, it's almost more insidious? Macroautophagy is the mechanism by which damaged proteins and organelles get sequestered and degraded. When that declines with age, the debris just accumulates. And that debris feeds back into proteostasis failure — proteostasis being the whole effort to keep your protein pool properly folded and functional. Misfolded proteins pile up, the folding machinery gets overwhelmed, and now you've got a second input into the same amplifying network.
Finn Brooks: So disabled autophagy isn't just one broken thing — it's like, kicking multiple loops at once.
Clara Bennett: That's exactly right. It simultaneously worsens proteostasis and sends debris into other hallmark pathways. Which is why — actually, no, the cleaner way to say it is — this is why the loops don't add, they multiply. Each failure point isn't contributing its damage and stopping. It's opening new routes for other failures.
Finn Brooks: That's — yeah. That's the click. That's the thing I couldn't quite land before.
Clara Bennett: And the consequence — the reason this matters beyond the mechanism — is that it explains the acceleration directly. A 70-year-old isn't just carrying more hallmark damage than a 40-year-old. They're carrying damage in a system where the loops are already tighter. New damage hits harder because the buffering capacity is already partially gone.
Finn Brooks: So it's not that something new starts happening at 70. The same machine is running — it's just that by then the feedback has compounded enough that any new hit propagates through the whole network almost immediately.
Clara Bennett: And that's actually — that propagation speed is what the three-tier classification is trying to explain. Because not every hallmark is doing the same job in that network. The 2013 López-Otín paper is explicit about this: primary hallmarks directly cause molecular damage. Antagonistic hallmarks initially protect against it. Integrative hallmarks are the organism-level wreckage that builds up from the other two.
Finn Brooks: Wait — protect? Some of them protect?
Clara Bennett: That's the part that gets dropped from the popular version. Cellular senescence is the clearest case. A cell detects DNA damage — let's say you're 35 — and it permanently arrests. Stops dividing. That is the right call. It does not become cancer. Short-term, that's your body running exactly as designed.
Finn Brooks: Okay but then by 65 those same arrested cells are just — they're still sitting there, right? They didn't get cleared.
Clara Bennett: Still sitting there, metabolically active, leaking SASP signals into surrounding tissue for decades. The protection becomes the problem. Same mechanism, same cells — the timing is what flips it.
Finn Brooks: So if you tried to just — I mean, people are working on senolytics, drugs that clear senescent cells — but if you clear them too aggressively, especially early, you've just removed your cancer defense.
Clara Bennett: The tradeoff is baked into the biology. That's not a side effect to engineer around. That's the framework telling you the intervention has structural limits.
Finn Brooks: Which — okay, this is where dietary restriction gets genuinely weird to me. Because it actually works across model organisms, lifespan and healthspan both, and it does it by hitting autophagy, mTORC1, AMPK, sirtuins, NAD+ metabolism — like multiple nodes at once. But then you've got rapamycin only blocking mTORC1, metformin only activating AMPK, NAD+ precursors only touching that one pathway. Each one is — I mean, they're each grabbing one thread of a net.
Clara Bennett: And the net is interconnected. Deregulated nutrient sensing — mTORC1 overactivated, AMPK and sirtuins underactivated — that's not one dial. So a single mimetic moves one dial and leaves the rest.
Finn Brooks: Right right — and actual dietary restriction does move multiple dials simultaneously, which is probably why it works, but it also increases infection susceptibility and impairs wound healing. So even the thing that works has the tradeoff built in.
Clara Bennett: That's the antagonistic structure again. The same pathways that slow aging when you modulate them also run your immune response. You can't fully separate the signals.
Finn Brooks: So loss of proteostasis — that's integrative, right? Misfolded proteins, Alzheimer's, Parkinson's — that stuff only emerges after decades of upstream damage from the primary and antagonistic hallmarks stacking up.
Clara Bennett: Exactly — integrative hallmarks are organism-level consequences. By the time you see the aggregated proteins, you're looking at the end of a very long upstream chain. Which is — actually, this is where the list-expansion question gets interesting. Whether adding chronic inflammation and dysbiosis in 2022 reveals new independent processes or just describes the same network in finer resolution — that's a real epistemological question about whether hallmarks are discovered or constructed. And there's a methodological response to that worth getting into.
Finn Brooks: Okay that question — I genuinely do not know where I land on it. Because if we're just subdividing, the list could be thirty hallmarks by 2040 and we'd be no closer to knowing which thread to pull first.
Clara Bennett: And that's exactly the risk — but before we accept that framing, I want to push on it. Because the thirty-hallmarks worry assumes the additions are arbitrary. The 2022 Copenhagen meeting didn't just pad the list. It added disabled macroautophagy, chronic inflammation, dysbiosis. And macroautophagy specifically — that one had been sitting inside proteostasis loss for nine years, treated as a contributor. The Copenhagen summary elevated it to standalone. Which means the boundary between 'mechanism' and 'hallmark' is, at minimum, a judgment call.
Finn Brooks: Wait — so it was already in the framework, just nested?
Clara Bennett: Nested. Under proteostasis. And now it's a peer of proteostasis. That's not a discovery — that's a reclassification. Which raises the real question: are hallmarks natural kinds, things you find in biology the way you find a gene, or are they constructs that scientific consensus agrees to promote?
Finn Brooks: Okay I love that framing but — I mean, doesn't the fact that it's useful matter? Like, if treating macroautophagy as its own hallmark changes how you target it therapeutically, maybe the reclassification earns its keep regardless of whether it was 'discovered.'
Clara Bennett: That's fair. But the falsifiability issue doesn't go away. If you can keep adding hallmarks as long as they're useful, the model never tells you what's not a hallmark. And a framework that can absorb anything can predict nothing.
Finn Brooks: Hm. That's — yeah, that actually lands differently than I expected.
Clara Bennett: The methodological response is to stop debating what counts as a hallmark and instead map the genetics. Barabási and Gladyshev, with Gross, Ehlert, and Loscalzo, published a network medicine study — arXiv 2025, Nature Aging 2026 — where they took 2,358 longevity-associated genes and mapped them onto the human protein interactome. The whole interaction network.
Finn Brooks: 2,358 genes. That is — wait, and they found clustering?
Clara Bennett: Hallmark modules. Genes associated with each hallmark form connected subgraphs — their own cluster within the interactome. Which means the hallmarks aren't just a committee's taxonomy. There's a network structure underneath that partially validates the groupings.
Finn Brooks: The biology is telling you the categories are real, at least approximately. And then — they used that to identify drugs? Like, existing drugs whose targets are proximal to those modules?
Clara Bennett: Computationally, yes. Repurposable compounds sitting close to one or more hallmark modules. That's the elegant part. But — and this is where I want to be honest — proximity in the interactome and clinical benefit in a human are two very different gaps to close. We don't have an aging drug that works in humans. Not yet.
Finn Brooks: Right — because even if your drug hits the module, the module is still tangled with every other module. Single target, interconnected system.
Clara Bennett: Exactly. The network study is the most sophisticated answer yet to the 'which thread first' problem — but it surfaces the same limitation the hallmarks framework already implied. Now, epigenetic clocks are the one domain where resetting is actually being tested in living systems — methylation-based biomarkers that estimate biological age, and partial reprogramming that tries to walk cells back toward a younger epigenetic state. That's where the 'can we reverse it' question is most live right now.
Finn Brooks: So the Barabási network map tells you where to aim, and epigenetic reprogramming is the closest thing to actually pulling the trigger. And still — no working drug. That's the honest place to land, isn't it.
Clara Bennett: That is the honest place. And the part I'm still turning over — the disease framing versus the process framing — it changes everything about what 'honest' even means here. If aging is a disease, the hallmarks imply a cure. If it's a fundamental biological process, the hallmarks imply you can modulate the rate but not reverse the direction. Those are not the same intervention.
Finn Brooks: And partial epigenetic reprogramming is basically placing a bet on the first one. Like, resetting methylation patterns toward a younger state using the epigenetic clocks as your readout — that's not modulating. That's saying aging is reversible.
Clara Bennett: In a specific cell, maybe. Systemically in a living human — nobody knows if that's safe. The feedback architecture we spent this whole conversation on means a reset in one hallmark module lands inside a network that's already compensating. You don't know what you're destabilizing.
Finn Brooks: Right — but that's not actually a reason to stop. That's just the next problem.
Clara Bennett: No, it's not a reason to stop. It's a reason to be precise about what the antagonistic hallmark problem actually demands — there may not be one correct intervention. Only context-dependent tradeoffs, staged differently across a lifespan.
Finn Brooks: I keep thinking about — you started with my grandmother and how fast those last two years went. And I came in thinking the question was 'what's the mechanism.' But the framework doesn't actually answer whether you can stop it. It just gets very precise about why it's hard.
Clara Bennett: That's maybe the most useful thing it does. López-Otín gave us vocabulary for what was happening to her. Not a cure. Vocabulary first.
Finn Brooks: Yeah. That's not nothing, actually.
