The Other mRNA

On June 16, 2021, CureVac reported interim Phase 2b/3 results for CVnCoV, its mRNA COVID-19 candidate. The number was 47 percent. Not against a placebo of saline complacency — against the same virus, in the same year, using the same headline technology. It missed the trial's own prespecified success criteria, and it landed below 50 percent — the efficacy floor the WHO and regulators had treated as the minimum bar worth clearing. The vaccine that was supposed to be the third member of a triumphant family quietly failed to reach the bar the family had set.

Here is the strange part, and it is worth sitting with: nothing about the category had stopped working. Messenger RNA still did what it does — instruct your cells to manufacture a spike protein and rehearse the immune system against it. The platform was intact. What varied was the implementation, and the implementation varied by enough to span the distance between "miracle" and "not approvable."

The differences were almost embarrassingly specific. Pfizer's and Moderna's mRNA was chemically modified — its uracil swapped for pseudouridine, a tweak discovered years earlier that lets the synthetic strand slip past the body's inflammatory alarms and stay legible long enough to do its job. CureVac bet on unmodified mRNA, dosed lower precisely because raw mRNA provokes more reaction. And the trial ran across ten countries while thirteen variants circulated, so the virus the vaccine was measured against was already drifting away from the one it was designed for.

Three knobs — a chemical modification, a dose, a moving target — and the outcome moved by tens of percentage points. The word stayed constant. The reality underneath it did not.

And where you draw the line between "platform" and "implementation" is itself a choice, not a fact of nature. You could just as easily insist that the chemical modification is the platform — that the thing people mean when they say "mRNA works" already includes the trick that makes it work — in which case CureVac wasn't running the same platform at all, but an earlier, losing branch of it. Draw the boundary either place and the point survives: the noun "mRNA vaccine" hid the very decision the outcome turned on.

This is the thing the headline tense hides. "The mRNA vaccines worked" is true, but it is a sentence about two specific products, not about a category. The platform didn't succeed; particular configurations of it succeeded, after years of unglamorous optimization that the triumphant phrasing erased. We tend to compress that work into the name of the thing once the thing works — as if the success lived in the noun rather than in the thousand decisions stacked beneath it.

And that erasure has a face. The pseudouridine modification that separates 95 percent from 47 percent traces back to Katalin Karikó and Drew Weissman, who spent the better part of two decades sidelined and under-funded precisely because the implementation labor wasn't the kind of thing the field knew how to value — until the noun finally won, at which point the category collected the credit. So the compression isn't only an epistemic error about where success lives. It is also a quiet machine for deciding who gets erased: let the category absorb the achievement, and the people who did the patient tuning vanish into the name of the thing they built.

It is a pattern that outlives this vaccine. We say "machine learning works" or "fusion is coming" or "the platform is proven" and let the category absorb credit that belongs to specific implementations — and we are surprised when the next instance, built on the same principle, lands at 47 percent. Technologies don't transfer their success by inheritance. A platform is a space of possible outcomes, not a guarantee of one, and the space can be enormous. The same physics that produces a 95 percent vaccine can produce a 47 percent one a few parameters away.

CureVac's result wasn't a refutation of mRNA. It was a reminder of where the achievement actually lived — not in the discovery that mRNA could teach cells to make a protein, which was the easy part, but in the patient tuning that turned a mechanism into medicine. The triumph was never the category. It was the implementation, all the way down. The map said "mRNA vaccine." The territory said: it depends entirely on how you build it.