The Liar’s Almanac: Two Truths and a Lie

How to start

1. Open it in your AI. Tap Open in ChatGPT or Open in Claude and press send, or paste the play prompt into any assistant that can open links.
2. The host loads the sealed pages. It fetches the ten sealed rounds from this site, reads you the first 16 characters of their SHA-256 fingerprint, and shows Round 1: a theme and three statements, A, B and C.
3. Name the lie. Send the letter you think is the lie, and add "sure" after it for 3 points instead of 2, at the cost of 1 if you are wrong. The host names the lie and reads the real fact behind it; say "why" for notes on the two truths, each naming its sources.
4. Check the seal. At the end, open the sealed file the host names, copy all of it, and paste it at aigamecards.com/verify to confirm no lie moved; its source list lets you check any fact.

Questions

Can ChatGPT play two truths and a lie?

Yes, but in plain chat it makes up the "truths" as it goes, so one can be false. It also never has to commit to which statement is the lie, so it can decide after you guess. It can guess the lie in three statements you write, but it cannot fairly host the game. In The Liar's Almanac all ten lies were fixed, and every true statement checked against two named sources, before release. The lies sit in a sealed file; the host reads you the first 16 characters of its SHA-256 fingerprint before round one, and afterwards you can check that none moved.

What does the seal prove, and what doesn't it?

It proves that no statement, lie or note changed after release. It does not prove that the host read each statement to you exactly as sealed, or scored every call correctly; the one-line score readout the host prints with every reply, and the revealed file, let you check both. Every note the host reads after a round names its sources in brackets, and the revealed file lists each source's address.

Can we play the classic version about ourselves?

Not with this card. The classic game, where each player says three things about themselves, needs no card at all, and this one never asks anything about the players. It plays the fact edition: ten sealed rounds about space, animals, history, words and more. At the end it gives one line on setting a fact round for your friends: two facts you have checked, one you made up, and the lie noted down before anyone guesses.

How long is a game, and can a family or class play together?

Ten rounds: about ten minutes alone, twenty as a table. One screen, one call: everyone argues it out, the youngest says their pick first, and one person sends the letter. A plain call scores 2 and never costs points; adding "sure" makes it 3, or minus 1 if wrong. Thirty in all: blind guessing averages under 7, and par, 16, is the author's estimate for a general-knowledge adult table. A grown-up can explain any word to a child at any time; the host never explains a statement mid-round.

Does the chat have to be able to open a link?

Yes. The host fetches the sealed file from aigamecards.com; pasting it yourself would show you every lie. ChatGPT opens it; aigamecards.com/launch-matrix shows which other apps do. No aigamecards account is needed. To check the seal afterwards, open the sealed file the host names, copy all of it, and paste it at aigamecards.com/verify.