No Inheritance

A vaccine remade against a world that moved while it was being made — last season’s proof describes viruses that no longer circulate, so each formulation meets its season as if it were the first

R829’s heat shield could hand its result forward because the next shield is its near-twin: same material, same shape, one trajectory adjusted around the lesson. The spent proof had a stable address to arrive at. Now take away the stability. What of a proof whose successor is deliberately not a twin — re-aimed every year at a target that moved during the aiming — so that what the predecessor proved, however completely, is a fact about a world that has stopped existing? The seasonal influenza vaccine lives there permanently.

The bet is placed before the test exists. In February each year, the World Health Organization recommends which viruses the northern hemisphere’s next influenza vaccine should contain; the southern hemisphere’s choice follows in September. The dates are not bureaucratic habit. Roughly six to eight months are needed to produce, approve and distribute the vaccine, so the formulation must be fixed that far ahead of the season it will face. At the moment of choosing, the complete test — the winter itself, the actual viruses that will actually circulate — has not happened and cannot be hurried. Every formulation is therefore a forecast wearing a syringe: aimed with great care, by the best surveillance network on Earth, at a future it will meet exactly once.

What the predecessor proved does not transfer, because its subject left. Influenza’s surface proteins change continually — antigenic drift — and the fastest-changing component, A(H3N2), can change enough in a single interval between selection and season to break the match. The record shows the swing. In well-matched seasons the vaccine has cut the risk of a flu medical visit by around half or more: 60% overall effectiveness in 2010–11, 52% in 2013–14. Then 2014–15: most circulating H3N2 viruses had drifted from the vaccine virus, overall effectiveness fell to 19%, and against the drifted H3N2 itself the final adjusted estimate was 6% — while against the well-matched influenza B component in the very same shots it was 55%. One vaccine, one winter, two verdicts: the matched part inherited fine, because its world had held still; the drifted part faced its test as a stranger. And the drifted strain had been nearly invisible at decision time — about 1% of detected viruses when the formulation was chosen in February 2014, dominant by late summer. The future that broke the proof was, at the moment of betting, a rounding error.

It is still happening, on schedule. For 2025–26, CDC reported that 93% of genetically characterised H3N2 samples were a subclade (K, within 2a.3a.1) first identified in June 2025 — after the season’s vaccine viruses had been selected — and antigenically different from the vaccine’s H3N2 virus. The preliminary overall effectiveness estimate for that season stands at 36%: not the collapse of 2014–15, and not the 56% of the season before it, but another fresh draw from the same lottery. Read CDC’s long table as a whole and that is what it looks like — 60, 47, 49, 52, 19, 48, 40, 38, 29, 39, 36, 30, 44, 56, 36 — not a technology steadily improving on its inheritance, but a score that resets every year to whatever the virus did while nobody could watch the test.

The honest edges. Much does hand forward, and the page’s claim depends on saying exactly what. The surveillance system hands forward: a global network watches the virus year-round and gets better at watching. The factories hand forward: the manufacturing platform, the approval path, the distribution chain are all inherited, improved, reused. Even immunity hands forward partially: a drifted vaccine still blunted severity in bad-match years, and some components stay unchanged for years when their viruses hold still — the 2025–26 H1N1 and B/Victoria components were carried over unchanged from 2024–25. What does not hand forward is the one thing R829’s shield line transmitted so cleanly: the match itself — the specific, settled fact that this article fits this test. The shield’s successor inherits a result. The vaccine’s successor inherits a method, a factory and a warning — and then faces its complete test, the season, with its own fit entirely unproven, as if it were the first formulation ever made. Because, against the viruses that will actually come, it is.

The test, schematically. Each season, a vaccine is aimed in February at the virus as last seen; the virus then drifts during the six-to-eight-month gap. The numbers are schematic units, not measurements: the diagram exists to show which column inherits. The platform column — surveillance, factories, method — grows every season. The match column starts at zero every season, and this season’s result depends only on this season’s drift, never on last season’s score.

No seasons run yet. Season 1’s formulation is being aimed — on a platform built by every season before it, with a match proven by none of them.

The shield and the vaccine are the two ends of one question: can a proof be inherited? The shield’s line answers yes, because the world it flies into holds still enough for a twin to use the result. The vaccine answers: only the scaffolding inherits. The answer itself must be re-earned every year, in advance, against an opponent whose whole strategy is to make last year’s answer about a virus that no longer exists. Each formulation is believed on credit, spent in its season, and leaves behind a number that will describe, precisely and uselessly, a winter nobody will ever meet again.