Why Real Press Exists

The Recipe That Started It All

One day not too long ago from today, I was searching for a recipe for dumplings, har gow specifically. The dim sum place near me recently closed down and I really missed the fresh har gow the aunties and uncles served so I thought I'd try to make my own. However I kept getting frustrated at all of the recipe blogs I found, as it became clear that a lot of them had AI generated images. So I became really skeptical, and ended up going to a food youtuber (Chinese Cooking Simplified) I really admire and followed their recipe.

That interaction left a mark on me though.

Through a series of similar interactions on social media, I wanted to see if I could help with the issue of flagging AI content. I ran into some questionable LinkedIn posts, and another questionable Reddit post. So I took my shot at building my first real application, Real Press.

What Real Press Does

Real Press is a search engine that rates every entry with a score/classification from Human all the way to AI. We run a single processing pipeline that deals with text, images (and soon video) to give a single composite rating of Human to AI. And every result has the analytics to back up the claim, and you can verify them yourself. Read more in our about section.

A composite score may be unsure, but if you dig in you can see if the text was Likely Human and the images are AI, which rendered a classification of Likely AI. This applies to all the URLs we can see - recipe blogs, e-commerce stores, zillow listings, social posts. You can check it all (assuming we don't get blocked - working on becoming a verified bot).

Why Independence Matters

Real Press is self funded so we can maintain our independence. We want to earn your trust, so we are not influenced by funders or donors. Frontier model companies shouldn't be trusted to grade their own work, so we hope that you'll trust us.

We don't take money from venture capital, private equity nor PACs - no one. We also don't sell your data or metadata, and we also collect very little about you as a user. Check out our privacy policy to learn more or use the developer settings on our site to see the calls we are (and are not) making. Our only income is from user subscriptions, and hopefully soon some enterprise contracts to hook into our rich API.

Why Transparency

If we're going to ask you to trust our scores, we need to be very transparent and authentic. That means showing you our real numbers, even when they're small, and showing you our mistakes or revisions. It means letting you flag scores you disagree with. It means publishing quarterly transparency reports that show exactly how we're doing.

This transparency exists on every record we have. You can see a comprehensive breakdown on every entry in our database if you are a paid user. If you choose to remain a free user you can see the composite rating. All users can see the score history and if anything changed and why.

Where We Are

It's a small team against an impossibly large problem. We've indexed about a thousand articles so far, and we're working to scale to the useful web which is about ~22M (roughly 22,000 times larger than the time of this writing).

Our detection isn't perfect, and honestly it never will be, because this is fundamentally an adversarial problem. But we're doing everything we can to bring you transparency at scale, and we'll be honest about our accuracy along the way.

Thank you for being here this early. It means a lot.

Erik (a real human)