Why a cancel button
Timelock encryption as it exists today is a one-way ratchet. You pick a date, you encrypt, and the release will happen, because the release value is a public beacon the rest of the world depends on too. That is right for sealed bids and auction commitments. It is wrong wherever “I am still here” is the actual message: inheritance, escrow, a dead drop, a disclosure a court should be able to stop, a key you may want to take back.
CRT makes cancellation the primitive. Each switch gets its own keys, held by independent operators, so one switch can be cancelled without touching any other — and because the keys are per switch and deleted at the end, a compromise later costs one switch, not the archive.
What CRT is — and is not
It gives you
A set of public keys to encrypt to, bound to one on-chain cell and one deadline. Later, the matching private keys, published on chain. A single recovery file an heir can open offline in twenty years.
It never touches
Your data. CRT does not encrypt, store or move anything of yours. Applications do that — the first is dead-man-switch, which encrypts a message or files in your browser and packs them into the recovery file.
It cannot
Stop a threshold of colluding operators (you choose how many, and from where). Prove that an operator deleted a key. Replace a lost wallet. Make any post-quantum claim.
Built on CKB Service Runner
A CRT node is not a new piece of software. It is a CSR runner — a CKB full node with the CSR sidecar — that has enabled the CRT service. CSR brings the parts a key-holding network is hardest to get right: operator registration with a 10,000 CKB Nervos DAO stake, a public ranking by an independent monitor, a sandbox with encrypted per-service storage, and paid jobs that any registered runner can pick up.
CRT adds one lock script, one small program for the sandbox, and a client library. The script makes a release impossible to commit before the deadline and impossible after a cancel; the program generates, holds, publishes and deletes one key per switch; the client builds the transactions, splits your secret across the nodes and writes the recovery file.
What you have to trust
- The chain
- CKB consensus decides time (median of the last 37 blocks) and whether the Alive Cell still exists. Nothing else is a clock.
- Fewer than t operators
- You pick n runners and a threshold t (5-of-8 by default), optionally with a mandatory branch such as your own node. Any t can decrypt after release; any t could collude before it. The app shows the concentration of your choice by operator, network and country.
- Not the website
- This site and its app build unsigned transactions and read public data. They hold no keys, submit nothing, and can be replaced by a bare CKB RPC.
- Not the recovery file’s source
- The file checks itself against the hash anchored in the Node Cells when opened, and shows MATCH or MISMATCH before anything else.
What it costs
| 8 runners, one year, defaults | CKB | Comes back? |
|---|---|---|
| Runner tips (1 keygen + 365 daily heartbeats, each runner) | ≈ 1 470 | no — this is what pays the operators |
| Organiser fee (optional, set by the app, at least 61 CKB when set; zero on cancel). Example: dead-man-switch, 70 | 70 | no |
| Cells on chain (Alive Cell, 8 Node Cells, 8 job cells) | ≈ 4 440 | yes — reclaimed when the switch ends |
About $15 spent and $44 parked at $0.01 per CKB. Longer terms scale the tips; cells stay the same. Cancelling is free. Full table in SPEC §11.
Status
The protocol specification (v0.2, CSR edition) is complete and frozen in its numbering. The lock script, the service, the client library and this site’s app are being built spec by spec; each ships to testnet when its acceptance tests pass. Mainnet follows CSR’s own mainnet gate for the service runtime. Follow the repository, or ask on the CSR Telegram.