Hi Product Hunt, I'm the maker of PaperSays.

I work in AI and machine learning, and keeping up with what's actually new is part of the job. Somewhere in the last two years I stopped being able to. New papers every day, in AI and in the neuroscience and physics I also care about, and most of what reached me was a headline or a thread that said more than the paper did. I got tired of finding out a week later that the "breakthrough" was a mouse study with no control group.

So I built PaperSays, mostly for myself at first.

What it does: an editor picks the research that matters in three fields (AI & machine learning, brain & mind, physics & space), and each paper becomes a short podcast episode, about nine minutes, two hosts, written from the paper's full text. Your daily briefing shows three editor picks, and the whole catalogue is open to browse. Every line is tagged SAYS (the paper), EXPLAINS (us unpacking a term or a method) or INTERPRETS (our read), with the citation, an evidence meter and a "what this paper does not show" section. The hype problem is really a who-is-talking problem.

The features came from my own annoyances with science podcasts:

- They mention a method I don't know and I can't stop to look it up while walking. So you ask the episode out loud, phone in your pocket, and the answer is limited to the paper's text and tagged the same way.

- There's no straight way to take a note mid-episode. So you say the note, and it's saved against that moment.

- A week later I've forgotten most of it. So it quizzes you on what you saved, then again later.

How it's made, honestly: a human editor picks the paper and the angle; the script is generated from the paper's full text with an LLM and checked against the paper's own numbers; the two voices are synthetic (ElevenLabs) and the app says so; a human reviews every episode before release.

Honest bits: iPhone only, English only, three fields for now, not in EU storefronts yet. Bootstrapped, no funding, launched last week.