You pick the jury
"Where should we eat?" deserves better than 4.3 stars from two thousand strangers. LiquidOpinion is one place for what you and your friends really think — about films, books, restaurants, plumbers, anything with a name. When you need an answer, it asks the people you'd have called anyway.
The store listings are on their way — these buttons go live the moment they are.
Every review site you use mints one number per thing — and one number the whole internet can read is one number worth gaming. Bombed, bought, botted, blurred into a beige 4.2 that counts your best friend and a click-farm as the same voter. The score isn't broken because people are dishonest. It's broken because it's global — the average of ten thousand strangers is nobody's opinion.
So we refuse to mint it. There is no LiquidOpinion score — not hidden, not hedged: the number does not exist. There isn't even an average inside your network: its verdict is the answer most of it actually gave — "most have it as a favorite" — never a 4.1 that nobody said, and the tally of who said what is one tap away. Truth goes back to being what it always was — relative to whose it is. A claim here reads like a poll that names its sample: "among the thirty people I trust about coffee, this place is on top right now."
Manipulating a global score costs money. Manipulating this one costs infiltrating actual friendships — one network at a time.
Past half of the pages published on the web are machine-written now, and a review is the cheapest kind there is to write: a plausible paragraph about a restaurant costs nothing to generate and is worth money to somebody. That is the whole economics of it, and it is why the star averages stopped meaning anything before the machines even arrived.
Every opinion here is signed by one human, who can say it once, to people who chose to listen. Publishing needs a credential issued to a person, one live opinion per person per subject; a revisit replaces what they said before rather than adding to it. So the plain headcount under a subject counts people rather than rows — and you can check that count yourself, from the signed data, without trusting us.
Two limits, said here rather than discovered later. This is not a wall against somebody opening twenty accounts — no anonymous system has one, and twenty accounts still read as twenty people. What it stops is one person becoming twenty voices under one subject. And what you can check offline is the deduplication; that the writer was entitled to publish at all is something we checked and you take on our word. The whole of what that means →
Most of what you think never gets written down, so it goes. Six months on you know you liked the film; you couldn't say why, or whether you liked it more than the one before it. Keep a line — a few drops, a sentence, the date — and "what did I actually think of that?" becomes a thing you look up rather than a thing you reconstruct.
So the ledger is a diary before it is anything else. Search your own words. Filter by shelf, by year, by four drops or more. Read your own rankings for a category, sharpened by the odd duel between two things you've rated. Look a whole year back as "2026 in ink", or let the desk hand you an anniversary — what you were watching three years ago today, and what you gave it. Bring in what you've already written on Letterboxd or Goodreads, and take the lot out again as a spreadsheet whenever you like — it's your record.
An account with one person in it isn't a lonely version of this app. It's the app. Friends only make the arithmetic bigger.
Right now your taste lives scattered across a shelf of single-purpose apps — one for films, one for books, a site you check before buying a bike, and a five-star widget for everything else. Each with its own account, its own feed, its own idea of who you are. None of them knows the others exist.
This ledger has a spine instead. Restaurants and cafés, films and series, novels and records, phones and trail bikes, plumbers and dentists, hotels, apps, games, even your employer — some 6,500 shelves, and a page for anything with a name. Same drops, same journal, same friends throughout: the person who steers you to a novel may be the one who knows brakes, and inking them once brings their whole ledger, not one aisle of it.
Entry 01
Jot what you think in drops — 0 to 6 — with words when you have them. It lands in an encrypted journal on your device first, airplane-mode friendly, before any network is involved. The app calls this graphite: penciled in, private, yours until you say otherwise.
Entry 02 · optional
When you want more than your own hand — and only then. Trust forms in person: two phones side by side, or a QR code across the table — that's inking a friend. No follow buttons, no requests from strangers; your network is people you've actually met. That handshake is the whole security model.
Entry 03
Ask any shelf — best espresso near you, best novel this year. On your own the standings are your rankings, out of your own entries. With friends inked they widen: trust-weighted, recency-aware, every line signed. "— Anna" beats twelve thousand anonymous stars.
And when an opinion does go public, it goes as prose and comparisons — "A beats B" — claims that aggregate honestly into standings anyone can recompute. Never a star average. There isn't one. See the public window →
No ads. No sponsored placement. The people being reviewed can't pay us anything, ever — there is no product for them to buy. You are the customer rather than the inventory, and there is nobody else to answer to: no investor, no vendor, no platform, nothing else for sale.
And nothing we store connects a published opinion to the account that wrote it: the register that stops double-voting holds no account id, and the proof that you were entitled to publish is checked and thrown away. Anyone can verify from the signed data that two opinions came from the same person; that the person was entitled to publish at all is something we checked and you take on our word. Both halves, and the arithmetic of what this costs to run, are on one page: who pays us, and who can't →
€7.90 is the founding rate. One day we'll move to the standard rate — €14.90 a year — but if you're already in at €7.90, you keep €7.90 for as long as you keep subscribing. No deadline, no countdown: if the founding rate ends, it will be announced, not expired.
The first 30 days are free, and they start the day you first write or import something — not the day you sign up. No card up front. And if you ever stop paying, nothing you wrote goes away: your journal, your rankings and your feed keep working offline, export always works, and only publishing and new friend connections wait for a payment.
The store listings are on their way — these buttons go live the moment they are.