What we've changed, what we've got wrong, and what we're still figuring out.
Last updated: April 2026 | Maintained by: Next of Kin
Most companies publish polished “About” pages and hide their mistakes. We'd rather do the opposite. This page is a public record of things we've changed, things we've got wrong, and things we're still working out.
We think honesty about our limits builds more trust than silence about them.
These are genuine open questions. We don't have complete answers yet and we're not going to pretend we do.
Does the AI actually help carers?
We believe it will. We have one founding contributor (Evelyn) and one founding profile (Ben). We have not yet run a rigorous test of whether the AI's answers are genuinely useful to a real carer in a real situation. That test is our most urgent priority. We will publish what we find — including if the results are disappointing.
How do we handle PHI responsibly?
Families will upload sensitive information — some of it clinical — even when we ask them not to. We are in the process of formalising data processing agreements with our AI providers (Anthropic and OpenAI) to ensure that content is handled appropriately. This is not yet complete. We will update this page when it is.
What happens when a caregiver dies?
The handover scenario — where a caregiver is gone and a new carer needs access — is the core reason Next of Kin exists. We have a QR code access model and a plan for long-lived access codes left with solicitors or trustees. We have not tested this in a real handover situation. It remains theoretical until we do.
Is the interview framework the right one?
Our seven interview domains were developed through our founding conversation with Evelyn. They represent our best current thinking about what carers need to know. They will almost certainly change as we learn from more families. We will update the content policy when they do.
This log records material changes to how Next of Kin works — new features, changes to data handling, corrections to published content, and decisions we've revisited.
| Date | What changed | Why |
|---|---|---|
| April 2026 | Deployment architecture changed from Cloudflare Pages to Cloudflare Workers | A misconfiguration meant the domain was never pointing at the Pages project. Fixed after investigation. |
| April 2026 | Auth method changed from magic links to email + password | Magic links proved unreliable in testing. Switched to email + password for stability. |
| April 2026 | Upload categories moved from hardcoded UI to database table | Allows new categories to be added without a code deployment. |
| April 2026 | Waitlist form rebuilt with progressive save | Previously, partial form completions were lost if users dropped off before submitting. Now the record is created on Step 1. |
| April 2026 | Terms & Conditions added as an acceptance gate before first upload | Users must now read and accept T&C before uploading any material. |
When we publish something inaccurate — in a blog post, on a product page, or in a generated answer — we correct it and record it here.
| Date | What was corrected | Impact |
|---|---|---|
| No corrections recorded yet. |
We decided not to use magic links for authentication.
We originally planned magic links (passwordless login via email). They proved unreliable — links expired, emails were delayed, and early users couldn't get in. We switched to email and password. We may revisit passwordless auth when the underlying infrastructure is more stable.
We decided not to send newsletters on a fixed weekly cadence.
The action plan originally called for a weekly newsletter. We changed this to “when there is something worth saying.” We didn't want to fill inboxes with filler just to meet a schedule.
We decided not to build a native app yet.
A web app that works well on mobile is sufficient for early testing. Native iOS and Android apps are on the roadmap but will be built after the prototype has been validated.
We decided not to make the AI answer from general knowledge.
An early design question was whether the AI should supplement family knowledge with general care guidance when it couldn't find an answer in the knowledge base. We decided against this. The value of Next of Kin is specificity. A general answer is worse than an honest “I don't know” in this context.
If you find something on the Next of Kin website or app that looks wrong — a broken feature, an inaccurate statement, something that doesn't match what we say we do — please tell us.
The most helpful report includes:
Write to: hello@nextofkin.ai
We would rather fix a mistake quickly than defend it.
This page will only be useful if we maintain it honestly. We commit to updating it when things change, when we get things wrong, and when we change our minds. If you notice it going stale, that's worth telling us too.