Essay · by Brad, building HearLater

Why we edit for the ear

I kept finding genuinely worthwhile discussions online, saving them, and never coming back. Not because they stopped being interesting — because reading them needed a kind of time I didn't have. What I had instead was washing up, a commute, and a dog that needs walking.

The backlog problem is real, the fix wasn't

The obvious answer already existed: text-to-speech. Great voices, browser buttons, read-aloud everywhere. And for articles, it genuinely works. But try it on the things I was actually saving — long Reddit threads, comment sections where the real story was three replies deep — and it falls apart instantly. A robot reading "u slash throwaway8834 says, this ^^^" is not a listening experience. I'd press stop within a minute.

The voices weren't the problem. The problem is that a discussion was never written to be performed. It has no script. Narrating it word-for-word is like broadcasting a raw, unedited interview tape — nobody does that, because the first job of making listenable audio has always been editing.

What a good editor does

Think about what a radio producer does to an hour of interview tape: cuts the throat-clearing, keeps the moment the guest says the surprising thing, preserves the argument instead of flattening it, and ends where the conversation actually landed. Nothing is invented. Everything that remains is real. It's shorter, and it's more faithful to the conversation as experienced, not less — because the noise was never the conversation.

That's the entire idea behind HearLater, compressed into one line: faithful to the source, edited for the ear. Keep the original post in the poster's own words. Keep the strongest replies, the corrections, the genuine jokes, the disagreement — voiced by distinct voices so you can follow who's talking. Cut the pile-on agreement, the usernames-as-punctuation, the markdown, the fifth restatement of the top comment. Never summarise people into paraphrase, and never, ever smooth a real argument into a fake consensus.

Why not just summarise?

Because a summary is a different product — a description of a conversation instead of the conversation. The reason you saved that thread was the texture: the way the person who actually works in the industry phrased it, the reply that made you laugh. An AI summary throws that away and hands you back a beige paragraph. Editing keeps the texture and loses the noise. If a listen ever stops feeling like the thread, we've failed at the one thing the product is for.

Designed for the time you actually have

The other half of the idea is the queue. Interesting-but-not-now content needs somewhere to go that isn't a tab graveyard, and the moment it fits into isn't a reading moment — it's the parts of the day your hands are busy and your mind is free. So HearLater works like a podcast app pointed at your own reading list: share a link into the queue in two taps, it prepares itself, and it plays through while dinner happens. No streaks, no guilt, no unread-count. The queue just waits.

It's honest about its limits, too. Listens are prepared automatically and can contain mistakes; the original source is always one tap away, and public Reddit threads and pasted text are what work most reliably while we're in beta.

If that's the kind of thing your saved-tabs pile is made of, join the waitlist — one email when your invite is ready, nothing else. — Brad