Pocket is gone and Obsidian won't save your reading backlog by itself

Pocket shut down in July 2025. Omnivore went before it. Somewhere around 35 million people who had been saving articles “to read later” suddenly had a pile with nowhere to live.

If you spent any time in the Obsidian community after that, you saw the same advice everywhere: move it into your vault. Use the Web Clipper. Own your data. And that advice is right. “File over app,” as Steph Ango puts it, anything with a server eventually shuts down, and the people who kept their reading inside their own files didn’t lose anything. The migration is the correct move.

But I went looking through the threads where people talked about actually doing it, and something kept coming up that the migration guides skip over.

Pocket was already a graveyard.

That word isn’t mine. “To-read graveyard” is what people call it themselves. The numbers back it up: across read-it-later apps, the share of saved articles people actually go back and read sits below five percent. You saved a thousand things with good intentions and read maybe forty. So when you export that pile and import it into Obsidian, you haven’t solved anything. You’ve moved the graveyard into a nicer building. The pile dies twice.

Here’s what the existing fixes get wrong. One camp says declare bankruptcy, delete the whole thing, start clean. That treats the pile as garbage. The other camp says organize it, folders, tags, a careful taxonomy. That just delays the problem, now you have a well-labeled graveyard. Neither one answers the only question that matters: out of these thousand saved things, which ones were actually worth my time?

I think the pile is misunderstood. A backlog of things you chose to save isn’t clutter, it’s a map of what you were curious about over months or years. The problem was never organization. It’s triage. Which to read, which to skip, which themes keep showing up that you didn’t notice while you were saving one article at a time. The useful question isn’t did you read everything. It’s what did you miss that mattered. (That framing I owe to MoAutomates, who said it better than I would have.)

That’s the part I tried to automate. Once your saved articles live in Obsidian as notes, the metadata is right there in your vault. So I had AI read across the whole pile, the titles, the sources, the descriptions, and pull it into a single synthesis note: what’s worth reading, what recurring themes are buried in there, what you can let go. Not a tidier list. A verdict on the pile.

I packaged that into an Obsidian plugin, Reading Inbox Synthesizer, if you want the finished version. On privacy, to be exact about it: there’s no server, no account, no telemetry, and your data only goes to the AI provider you configure yourself. That’s the whole point of doing this inside your own files instead of someone else’s app. The subscription tools that do something similar run on their servers, which is the same arrangement Pocket had right up until the day it closed.

But the plugin is the footnote. The thing worth taking is the reframe. When your read-it-later pile moved to Obsidian, you fixed the ownership problem. You did not fix the graveyard. Don’t just store the pile somewhere safer. Make it tell you what was worth reading, or you’ll have moved a thousand articles you’ll never open from one place you ignore them to another.