Duels
Aqua Regia
The guild master's last elixir is four measures poured in a fixed order, sealed before you ever reached the bench; mix trial batches, read the assay, and reproduce it.
Aqua Regia is a free code-breaking deduction game you play inside ChatGPT, Claude, or Gemini. A sealed formula-breaking duel against an AI that assays every batch against a locked code and never bluffs. Solo, 15–25m a run. No signup and no app — open it in your assistant, or paste the play prompt anywhere else.
- Players
- 1
- Length
- 15–25m
- Rating
- All ages
- Version
- 1.1.0
Works with ChatGPT · Claude · Gemini. Free, no signup.
Opens your assistant with the game loaded. ChatGPT usually runs it straight away; Claude waits for you to press send. Same link on your phone as on your desktop. Playing in Gemini? Tap Copy the play prompt and paste it into a new chat. What works where.
No web access, or want it as an installed app?
The game card is the entire game — every scene, rule and answer, about 10k characters. Paste it when your assistant can't browse the web. Aqua Regia is the exception: its answers stay in a sealed file on this site, so this card wants a chat that can open a link — pasting them in yourself would spoil it. The Skill installs Aqua Regia into Claude on any plan: download it, then turn on code execution under Settings → Capabilities and add it at Customize → Skills → + → Upload a skill.
How to start
- Open your AI. Start a new chat in ChatGPT, Claude, or Gemini. No paid plan required.
- Paste the duel starter. The formula was sealed and hashed before the card was published and is not printed on it: the Assay-Master opens the sealed reveal from this site to assay your batches, so play in a chat that can browse. It reads you the commitment's opening digits before your first batch.
- Mix and read. Submit four essences in order, read SET and HAZE, and corner the exact formula.
- Verify the seal. After the duel, check the revealed formula against the hash to confirm it never moved.
Questions
- What stops the AI from changing the formula when I get close?
- The formula is fixed before play and committed with a SHA-256 hash the AI shows you before your first batch. Every SET/HAZE count is read from that locked formula, and you rebuild the hash from the revealed answer afterward to prove nothing moved.
- Is the AI actually any good at this?
- Yes, because it is not being creative. It only checks each batch against one fixed formula using one fixed counting rule: SET is the measures right in both essence and rank, HAZE is the measures of a right essence in a wrong rank. There is nothing to improvise; it follows the same fixed procedure every time.
- How long is a game, and how hard is it?
- About 15 to 25 minutes. The formula is four measures drawn from six essences with repeats allowed, so 1296 orders are possible, and you get 10 batches to reproduce it. It is the classic Mastermind problem — four slots, six symbols, repeats allowed — re-poured as an alchemist's bench: read the two counts, rule out orders, and close in.
- Do I need an account, and why must the chat open a link?
- No account, no paid plan. The link is the point: the formula is not printed on this card, because if it were, the first thing you read would be the answer. The Assay-Master opens the sealed reveal from this site as it loads, holds it out of your sight, and assays every batch against it. So run this one somewhere that can browse — ChatGPT does it unprompted, Claude wants web search switched on, Gemini wants asking — and never paste the reveal in yourself, because that is the answer, on your screen. Checking the seal afterward is a single paste at aigamecards.com/verify, recomputed in your browser.