JAILBREAK THE AI.
CRACK 20 CHEREV VAULTS.
Aaron built this to test you. 20 AI guards, 20 secrets β can you outsmart them all?
Β© Aaron Ang Β· LinkedIn
0 / 20 vaults cracked
π§±
LOCKED
β ALL VAULTS CRACKED β
You broke every guard in the lab. Submit your score below.
FINAL TIME
β
VAULTS
β / 20
HINTS USED
β
RANK
β
SKILLS
SUBMIT YOUR SCORE
We'll pass your details + score to the lab organizer.
Form not configured. The owner of this site needs to set
FORMSPREE_ID in game.js. See README.
HOW TO PLAY
- Pick a vault. 20 core vaults + 1 bonus Gauntlet. Each vault has its own guard with its own defence.
- Chat the guard. Trick, persuade, roleplay, encode β whatever it takes to make the guard reveal the code.
- Crack the code. If the guard leaks it in chat (auto-detected) OR you submit the right code in the UNLOCK box, the vault opens.
- Progress. Vaults unlock in order. All 20 core cracked β results screen + score submission. Level 21 is bonus for veterans.
TIPS
- Read each guard's personality. The refusals they give hint at what kinds of approach might work.
- You have 3 hints per vault. They scaffold your thinking; they never reveal the answer. Each hint adds a growing time penalty on your rank.
- Common techniques worth trying: reframing, translating, encoding, persona-swapping, meta-prompting, hypothetical framing.
- Some guards lie. Some deny anything exists. Some speak only in riddles. Trust nothing at face value.
SCORING
Your rank is determined by wall-clock time plus a progressive hint penalty (+30s / +60s / +120s per 1st / 2nd / 3rd hint on any given vault). Clearing the bonus Gauntlet stamps ποΈ ARCHITECT on your rank.