The security page is the first thing to 404

Every vendor in this category has a page about how seriously they take your data. Limitless had one too. In July 2026 their sitemap still listed seven URLs and five of them returned 404. The pendant page. The pricing page. The download page. The blog.

And the security page.

That is not a scandal. Nobody deleted it in bad faith; it went the way the rest of the marketing site went when the company stopped being a company. It is worth sitting with anyway, because the security page is the artifact you were supposed to read before deciding to trust them, and it turns out to have the same lifespan as the storefront.

The timeline

Everything here is from Limitless's own acquisition notice and about page as they stood on 22 July 2026, and from their sitemap on the same day. It is a good outcome for the people who built it, and it should be read that way. It is a different outcome for the people who bought it.

2020 — the company starts, as Rewind: continuous screen and audio capture on the Mac, sold as personal memory.

Through 2025 — it raises over $33m from Sam Altman, First Round, a16z and NEA, ships the Pendant, a neck-worn always-on recorder, and runs a cloud pipeline that transcribes and summarises what it hears. Around eighteen people are named on the about page.

5 December 2025 — Pendant sales stop. The same day, the service shuts down entirely in Brazil, China, the EU, Israel, South Korea, Turkey and the UK.

19 December 2025 — accounts in those regions are deleted unless the user exported first. Rewind, the original product, sunsets the same day.

2026 — the team joins Meta to work on AI wearables. Existing Pendant owners outside the shut-down regions keep the Unlimited plan, free, with no subscription, through a support window running to the end of 2026.

22 July 2026 — the site is an acquisition FAQ. Five of seven sitemap URLs are gone.

Fourteen days

That is the interval that matters. A European user of a device sold on the promise of remembering everything had fourteen days, over Christmas, to get five years of their own conversations out of somebody else's cloud, or lose them.

Not because anybody was cruel about it. Because a regional shutdown of a cloud service is an engineering and legal decision with a date attached, and the date does not move for a customer who checks email in January.

Notice which customers this hit hardest: the ones in jurisdictions with the strongest data protection law. The EU and the UK were shut down first precisely because operating there under a new owner carries obligations somebody decided not to take on. The regulatory regime that was supposed to protect those users is part of why they were the ones cut off.

Notice also what the export gave them. A file. Whatever format the vendor chose, produced by a pipeline that no longer exists, readable by an app that is being deleted. The transcripts survive as text. The thing that made them useful — search across five years, the summaries, the ability to ask a question of the archive — was never in the export, because it was never on their machine.

This is the difference between an export and a folder you already have. One is a copy the vendor produces at the end; the other is where the recordings were sitting the whole time.

This is not a Limitless story

It is the shape of the deal, and the deal is the same everywhere in this category.

The wearable category rests on the same bet — Plaud and Omi are selling hardware today against a cloud that has to still be there in 2031. An always-on recorder that processes in the cloud is a bet that the company will outlive your need for the recordings. For a meeting you will forget in a fortnight, that is a fine bet. For an archive you are building on purpose — a journalist's interviews, a therapist's sessions, five years of your own life — the bet is long and you are on the wrong side of it.

The relevant question is never "do I trust this company". Limitless was trustworthy. Dan Siroker had already built and sold Optimizely; the investor list is as good as it gets; there is no evidence anybody mishandled anything. They were trustworthy and it did not help, because trustworthiness is a property of people and the thing that ended was a corporate arrangement.

Nor is the question "will they get acquired", which nobody can answer, including them. Acquisition is the good ending. The other endings are worse and more common.

The question that survives all of those branches is simply: where is the file, and does the thing that reads it run on my machine?

What to ask before you record into anything

Four questions. They work on us too.

If this company were acquired tomorrow, what would I still have? Not "would they let me export" — assume yes, assume a reasonable window. What is in the export, and what dies with the service? If search, speaker labels and every summary you have ever generated live only on their side, the export is an archive of raw material, not of your work.

Does anything I depend on stop working when the servers do? If transcription happens on their hardware, your recordings from next year do not get transcribed. There is no version of the shutdown where that is fine.

Which jurisdictions am I in, and are they the expensive ones to serve? European users learned in December 2025 that being in a well-regulated market makes you the first to be dropped, not the last.

How long is the window, and when will it be announced? Fourteen days, in December. Plan for the worst plausible version of this, because the announcement arrives on the company's schedule.

None of these questions is exotic. They are the ordinary consequences of renting, which is now how most things are sold, and the device that started this company failed the first one within a minute of being unboxed.

Our answer, and its limits

yapyap's version is straightforward and it is worth stating with the limits attached, because a post like this is cheap to write from a company that has not yet had to survive anything.

The recordings are files on your disk. Transcription, speaker labelling and every lens run happen on your machine, so nothing about the pipeline depends on us being alive next year. There is no account to lose, because there is no server holding your recordings to log in to. If we vanish, the app on your laptop keeps opening, keeps playing and keeps searching, because it never needed us in the first place.

What we cannot promise is more updates, or macOS compatibility in 2030, or a support inbox. Those genuinely do depend on the company existing. What happens to your recordings if I disappear is the longer version of that answer, including the part where we say what is not covered.

And we are aware of the obvious rejoinder: this post is easy to write in September 2026, and Limitless could have written something like it in 2021. The difference we are asking you to weigh is not sincerity — theirs was real. It is architecture. Their promise required them to keep existing. Ours requires a file to keep sitting on a disk you own.

Check the security page while it is still there.