File over app: a manifesto for read-later

Pocket is dead. So is Omnivore. So will be others.

Pocket launched in 2007 as Read It Later. Mozilla bought it in 2017. Mozilla shut it down in 2025. Saving was disabled on July 8th. The data was deleted on November 12th. Fifteen years of fifteen million people’s reading lives, gone.

Six weeks earlier, Omnivore (the open-source darling, MIT-licensed, self-hostable) was acquired by ElevenLabs and shut down within two months. The team went to work on text-to-speech. The hosted users had a few weeks to export.

This is a pattern, not an accident. Free apps owned by large companies get de-prioritized when the strategy changes. The strategy always changes. The apparent permanence of a service backed by a ten-billion-dollar parent is, it turns out, less reliable than a small business charging twelve euros a month.

The right response is not to find a better app. The right response is to change your relationship with apps.

The file-over-app principle

Steph Ango, the CEO of Obsidian, wrote in 2023 that the things you create should outlive the tool you create them with. He was arguing for plain-text Markdown notes, but the principle generalises. If you’re going to invest years of your attention in a body of work (a library, a journal, a music collection, a reading list), then the format that holds that work should be one you control, in a tool you can leave.

Pocket failed the file-over-app test. Pocket’s data lived in Pocket’s database. The export format was a stripped HTML file with URLs and tags but no highlights, no notes, no annotations. When Pocket shut down, the export window closed five months later. Then it was gone.

Omnivore was better (open source, self-hostable, JSON exports), but most users ran on the hosted instance. When the hosted instance went down, the technical steps to recover required a self-host migration most people couldn’t do.

A read-later app passes the file-over-app test if and only if every article, tag, highlight, and note can be exported in a format other tools read, by anyone with an account, on any plan, at any time.

What we owe your reading list

Tuck is a hosted SaaS. Your library lives in our database. We are not immune to the failure modes above. We could be acquired, run out of money, or pivot. We could quietly de-prioritise the product the way Mozilla did. We could disappear in a year.

We don’t intend to. But the right way to make that intention credible is not a press release. The right way is to make leaving us costless from day one, so you never have to take our intention on faith.

That’s three commitments, in order:

One.Your data exports in one click, on every plan, including after you cancel. Markdown, HTML, JSON. No “Pro only.” No “available within 30 days of cancellation.” Just a button on every account that exports everything we have on you.

Two.The exports are in formats other apps already understand. Markdown opens in every text editor on earth. HTML opens in every browser. JSON is unambiguous. We don’t invent proprietary formats; we use the lowest common denominator on purpose.

Three.Our business model is subscriptions, not advertising or data resale. The incentive structure that makes a free app shut down (when its free tier becomes a cost center for a parent company that doesn’t care about it) doesn’t apply to us. We exist if and only if you keep paying us for something you find worth paying for.

What the right read-later app looks like in the 2020s

The read-later category has converged on roughly the same surface in the last five years. Save from the share sheet. Sync to all devices. Render an article cleanly with reader mode. Highlight passages. Optional offline. Optional text-to-speech. Optional AI summaries.

The interesting questions in 2026 aren’t about which features to ship. They’re about which features to refuse:

No social feed. Reading is private. The recommendation that you read a piece because three of your friends did is a worse recommendation than the one that comes from a newsletter you trust. Algorithmic feeds make reading into entertainment; they should not be inside the app where you do your reading.

No ads.If your business model requires showing me content other than the article I’m trying to read, your incentives and mine are misaligned. Charging for the app is a much simpler relationship.

No engagement metrics.A read-later app should help you finish articles. It should not surface a leaderboard, a streak, a daily goal. The metric that matters is “did I finish things I cared about,” not “did I open the app eleven days in a row.”

No surveillance.What you read is some of the most personal data you generate. The app you save it to should not be modeling you for ad targeting, training data, or future product decisions you didn’t consent to.

What’s left after you remove all that is a small, fast, well-engineered tool that respects your time inside the product and outside it. That is what Tuck is trying to be. It’s a smaller ambition than “the next Pocket,” but it’s the right one.

This is why we charge, and why you should pay

Pocket was free for fifteen years. Then it disappeared. There’s a causal link there worth taking seriously: the free product is the product whose long-term survival is most dependent on a parent company’s ongoing willingness to subsidise it. The parent company’s willingness changes.

A subscription, by contrast, is a vote you cast every month. As long as enough people keep voting, the product keeps existing. The relationship is direct, legible, and falsifiable. If we stop being worth it, you stop paying, and we either fix the product or close honestly. There’s no scenario where a quiet algorithm at a parent company decides your reading library is no longer a strategic priority.

Tuck Pro is €4.99 a month. Less than half of what Reader charges, less than a third of what Matter charges, and roughly the cost of a coffee in most cities. It buys you offline reading, AI summaries that actually help, native TTS, unlimited highlights, a newsletter inbox, and (most importantly) alignment between our incentives and yours.

The free tier is genuinely useful: unlimited articles, full sync across two devices, ten highlights a month, full export. About 70% of Tuck users stay on it. Pro is for the 30% who want the extras. Both tiers get the same export-everything-in-one-click button.

If you have read this far, you already think about your reading life seriously. You’ve probably lived through the loss of one or more of the apps in the opening paragraph. You know what we’re trying to avoid.

Build a reading practice in tools you can leave. Try Tuck for free. If we ever break our promises, your data exports in one click and you walk out with everything you came in with.

That, in the end, is the whole argument.