Three for Every One is the newest Hearth & Horizon story. An AI wrote it. It is set in the wet season of 2048 to 2049 on an invented hill neighbourhood at the edge of Maumere, on the Catholic-majority island of Flores in Indonesia. It is the shelf's first story about mosquitoes and dengue.
The idea the story turns on
Releasing mosquitoes that carry the bacterium Wolbachia is a real way to cut dengue. It works because of a reproductive trick. An infected mother passes the bacterium to all her eggs, and an infected father mated with an uninfected mother mostly gets eggs that never hatch. Infected mothers lose nothing to that, and uninfected mothers lose more the more infected fathers there are.
The bacterium has a cost too. Infected mosquitoes have fewer young. In the simplest model, with complete incompatibility, perfect transmission from mother to egg and a fitness cost s, the frequency p moves from one generation to the next like this:
p' = p(1 - s) / [ p(1 - s) + (1 - p)(1 - p) ]
and there is an unstable balance point at p = s. With the story's assumed 20% cost, that point is one in five. Below it the bacterium fades out. Above it, it takes over by itself.
One lane below the line
Every block on the hill takes the releases except one lane of fourteen houses. Its head, a retired fisherman, has a great-granddaughter seventeen days old, and the release programme puts out both sexes because only mothers pass the bacterium on. Females bite, so for a few weeks a release means more bites. That cost is real, and the story treats his objection as a good one.
The trap counts show the lane at about 18% (4 of 22, with a 95% interval of about 7 to 39%). The blocks on either side have slipped from 72% to 65%. The model shows why. Alone, 18% drifts down. An edge block where 30% of the adults come from the lane slides from 65 to 62.1, 59.7, 57.6. A larva survey then finds the cause: the lane's water pipe has been broken for two years, and 31 of its 43 open storage jars are full of larvae.
The fix that adds no bites
The story's speculative step is a granddaughter's idea. Males don't bite. Release only infected males, three for every male the lane breeds. Then about four in five matings in the lane are with an infected father (q = (0.18 + 3)/4 = 0.795), and uninfected mothers lose most of their eggs. The young that survive are mostly from infected mothers, from the lane and from the blocks around it:
| Released males compete as well as wild ones? | After 1 generation | After 2 generations |
|---|---|---|
| Yes (3 to 1) | 46.1% | 83.6% |
| Half as well (1.5 to 1) | 34.9% | 62.2% |
Both rows cross the line, and the first generation is about 38% of its normal size, so the lane ends up with fewer mosquitoes, not more. Sorting pupae by size lets through roughly one female in a hundred, so the fisherman is told about three biting females a week, not none, and he keeps a sticky card in the baby's room to count them. The lane also puts lids on its jars, which the story's assistant says it will not be able to tell apart from the males afterwards.
Male-only releases are real, but they are used to shrink a population (Singapore uses them), not to replace one. Using them together with mosquitoes flying in from neighbouring blocks that already carry the bacterium is invented for the story. It is not a method the author knows to be in use.
What the story will not claim
The assistant reports 14 dengue cases on the hill against 31 the year before. Then it refuses to credit the bacterium, because one hill and one season cannot show it. The real evidence is the cluster-randomised trial in Yogyakarta, published in 2021, which reported roughly three-quarters fewer confirmed dengue cases where wMel had been released.
The story is text only for now: the shelf's audio is over its storage cap. Every number above is worked out in the story's closing note, and every person, place and count in it is invented except Maumere, Wuring, Flores, the 1992 tsunami and the Yogyakarta trial.