In partnership with
What the People Building AI Actually Read
The people building AI aren't only reading about AI. Sam Altman's pick for getting more done is Zero to One, and it has nothing to do with machine learning.
Shortform breaks down books like The Beginning of Infinity and Thinking, Fast and Slow into the key ideas, the context behind them, and what to do with them.
Get the ideas behind the breakthroughs in about 15 minutes, then decide which ones deserve your whole weekend.
LONGEVITY LATEST ISSUE 34 · 28 OCTOBER 2026
LONGEVITY LATEST · DEEP DIVE
The Scaffolding Problem
Why ageing skin makes less collagen, which interventions plausibly help it rebuild, and how to read a peptide label.
By Christian Thomsen · Companion to Issue 34 · 28 October 2026 · ~6-minute read
I expected the cells to be the problem. Ageing skin makes less collagen, so surely the fibroblasts that make it are simply worn out. That’s only part of what the data showed.
In 2006, James Varani and colleagues at the University of Michigan grew fibroblasts from the skin of people aged 18 to 29 and people over 80. In a dish, the older cells still produced about two-thirds as much type I procollagen, the raw material of collagen fibres: 56 against 82 nanograms per millilitre. Then the team looked at the same kind of cells in place, inside the skin. In young skin, 78% of a fibroblast’s surface was attached to collagen fibres. In old skin it was 58%, and the cells were spread out only half as far.
Some of the shortfall is the cell. A lot of it looks like what the cell is holding on to. That idea, which the Michigan group calls fibroblast collapse, is the most useful lens I’ve found for skin claims. It doesn’t replace trials, but it helps explain why some interventions are more plausible than others.
How skin loses its scaffolding
The dermis, the thick layer under the surface, is mostly type I collagen woven into a dense mesh. Fibroblasts live inside it, gripping the fibres and pulling against them, a little like a tent held taut by its guy ropes. That tension is itself a signal. A stretched fibroblast makes procollagen and keeps the collagen-cutting enzymes, matrix metalloproteinases or MMPs, in check.
Collagen in skin is unusually long-lived, with an estimated half-life of about 15 years, so damage piles up. Sugars form cross-links that stiffen the fibres. Ultraviolet light switches on MMPs that cut them; Gary Fisher’s group reports that a moderate dose of UV can raise MMP levels hundreds of times over and cut collagen production by roughly 80%. The fragments can’t be fully cleared, and fibroblasts can’t grip them.
From there the loop runs itself:
Fragmented fibres give fibroblasts less to hold.
Collapsed fibroblasts lose tension, make less procollagen and more MMP-1.
More MMP-1 fragments more collagen, and the cycle tightens.
A caveat worth noting: the collapse model comes mainly from one research group, and the University of Michigan holds patents on MMP inhibitors for skin ageing. The cell and biopsy data are solid; the model that ties them together is an interpretation, though a widely accepted one.
Ageing skin isn’t only short of collagen-making cells. It’s short of collagen worth gripping.
What each intervention does to the loop
Read through that lens, the interventions graded in Issue 34 mostly aim at one of three things: stop the cutting, restart the building, or re-stretch the cells. Where each one acts is a separate question from how well it works.
Intervention | Where it acts on the loop | Issue 34 grade |
Daily sunscreen | Stops the UV-driven surge in collagen-cutting enzymes before it starts. | A · prevention |
Retinoids | Restart procollagen and hold back MMPs. Biopsies show new collagen I forming. | A · repair |
Hyaluronic acid filler | Physically re-stretches fibroblasts; procollagen rose around injection sites in one small study. | Not graded |
Strength training | Unclear. One trial suggests blood-borne signals; elasticity improved, thickness barely changed. | C · watch |
Collagen powder | Fragments reach the blood and stimulate fibroblasts in the lab; modest effects in trials. | C · optional |
Peptide serums | Must first get through the skin barrier, which size makes harder for larger peptides. | C · unproven |
The filler result is the neatest test of the model. When Fisher’s group injected cross-linked hyaluronic acid into sun-damaged forearm skin, fibroblasts around it were visibly stretched and switched on type I procollagen within a month, and the effect lasted at least three. The filler isn’t a drug; on the Michigan group’s reading, the stretch did the work. It’s one group’s small study, and it says nothing about lasers, which work differently.
Hughes et al. Ann Intern Med (2013) · Griffiths et al. NEJM (1993) · Huang & Lee. Dermatol Pract Concept (2025) · Nishikori et al. Sci Rep (2023) · Fisher, Varani & Voorhees. Arch Dermatol (2008)
The size problem
The outer layer of skin exists to keep things out, and it’s good at it. In 2000, two Amsterdam dermatologists, Bos and Meinardi, pointed out that the compounds known to penetrate skin well, from nicotine patches to topical steroids, almost all weigh under 500 daltons. Their “500 Dalton rule” is a rule of thumb, not a verdict: formulation, carriers and damaged skin all shift it. It doesn’t rule an ingredient out. It tells you how hard the job is.
Tretinoin, about 300 daltons: comfortably small, which is part of why it works.
GHK-Cu (copper peptide), about 400: small enough to have a chance.
Palmitoyl pentapeptide-4 (Matrixyl), about 800: a fatty-acid tail is added to help it through, which tells you the barrier is a real problem.
Collagen in a cream, about 300,000 for the intact molecule: it moisturises the surface and goes no further.
That’s why topical peptide evidence is plausible but thin. The most-cited human data for GHK-Cu are six small studies, mostly 12 weeks long, gathered in a review co-written by Loren Pickart, the chemist who discovered GHK. Several were never fully published, and some had no comparison group. None of that proves they don’t work. It means nobody independent has properly checked.
Collagen powder takes the other route, through the gut. In 2005, Japanese researchers showed that after people ate gelatin hydrolysate, short fragments, mostly prolyl-hydroxyproline, appeared in their blood within an hour. In the lab those fragments stimulate fibroblasts. In trials, the two latest reviews disagree on hydration, elasticity and whether funding explains the gains: the 2025 review found no benefit in independent trials, the 2026 review found independent and sponsored trials similar. Both find nothing reliable for wrinkles.
Bos & Meinardi. Exp Dermatol (2000) · Pickart & Margolina. Int J Mol Sci (2018) · Iwai et al. J Agric Food Chem (2005) · Myung & Park. Am J Med (2025) · Batool et al. J Cosmet Dermatol (2026)
The frontier
Two threads worth watching. First, senescent cells: a Boston University study, publicised in May, found that five days of a topical senolytic cleared ageing cells from the skin of old mice, and 80% of their wounds had healed by day 24 against 56% in untreated mice. Mice only, and senescent cells also help normal wound repair, so timing will matter.
Second, training. The one skin trial found clear gains in elasticity with both cycling and weights, but its dermal-thickness result was small and only marginal between groups, there was no inactive control, and three authors work for a cosmetics company. It needs repeating in men, in other populations and against a group that doesn’t exercise before anyone says the gym builds dermis.
What this means for you
If I were designing a skin routine from this evidence, it would be boringly simple. Protect the scaffold you have with sunscreen every morning, SPF 30 or higher with a 4- or 5-star UVA rating. Give your fibroblasts a reason to rebuild with a retinoid at night, started slowly. Lift twice a week for the proven reasons, and treat any skin benefit as unconfirmed. Eat enough protein, from whatever source you like.
The most exotic thing in that routine is the tretinoin, first approved, for acne, in 1971. For wrinkles, it’s still the benchmark newer creams should be tested against.
Protect the scaffold. Then give the cells something to pull on.
Longevity Latest grades the evidence on interventions like these every Wednesday: what works, what’s premature, and what’s just marketing. The weekly issue carries the full grades and the practical plan.
Sources and further reading
Evidence reviewed through 24 October 2026. The fibroblast figures come from cell culture and skin biopsies; the collapse model is the Michigan group’s interpretation, and the university holds related patents. Molecular weights are approximate. GHK-Cu human studies are small and partly unpublished. The strength-training thickness change was not significant within groups. The senolytic result is in mice. Educational content only; retinoids should not be used in pregnancy or when trying to conceive.
© 2026 FrontWave Media Ltd · Longevity Latest