Cradle is a continuous safety layer that runs from before conception to a year after birth. Its product is not content. Its product is time — the hours between a woman's first warning sign and a clinician knowing about it.
Maternal mortality review committees find that 87% of pregnancy-related deaths are preventable, and that mental health conditions are the single largest underlying cause. Almost no pregnancy product is present for either.
Every pregnancy app is organised around the forty weeks. The app is downloaded at the positive test, peaks at the anatomy scan, and quietly ends somewhere around the six-week check. The risk does not follow that curve.
Roughly a quarter of pregnancy-related deaths happen between day 43 and day 365 — in a window with almost no product, almost no monitoring, and, for most women, almost no appointments. That window is the reason this exists.
Two-way monitoring through the days after birth, when postpartum preeclampsia appears and almost nobody is measuring. Text-based BP surveillance is the strongest evidence base in digital maternity care — and it eliminated a racial disparity in who got measured at all.
Screening at every point guidance requires, a bipolar screen before any antidepressant conversation, and a licensed prevention programme for women at increased risk. Not a wellness tab — the leading underlying cause of maternal death.
A woman with no phone is registered on a card code, tracked by attendance, and followed up in person. The health worker becomes the user; she remains the patient. Nothing about her care is reduced.
Depression screening at three, six, nine and twelve months. Cardiovascular follow-up for anyone who had a hypertensive disorder. Glucose retesting after gestational diabetes. The things nobody currently books.
In almost every AI health product, the language model receives the message and a filter checks what it said. We inverted it. A deterministic clinical kernel reads every message, every reading and every score first. When it fires, the model is never invoked at all — so it cannot talk a woman out of an escalation, because on those turns it does not run.
The kernel is a pure function with no network calls, tested on an adversarial corpus of phrasings that are easy to miss — understatement, euphemism, negation traps, Nigerian English and Pidgin, and warning signs buried mid-paragraph after small talk. A build that misses one does not deploy.
Every product in this market publishes what it can do. This is the other list. It is kept in the codebase, enforced by tests, and it is the fastest way to understand what kind of company this is.
The kernel takes no channel argument. That is not a preference — it is what makes the claim honest. A woman texting from a borrowed handset in Ondo State is evaluated by exactly the same rules as a woman on an iPhone in El Paso, and any build where that stops being true fails its tests.
We are looking for pilot sites, clinical advisors, and partners — particularly obstetricians, midwives, perinatal psychiatrists, community health worker programmes, and anyone already working on maternal safety in Nigeria.