What is live

v1 is permissioned. Interlude runs the validator, the node and every seat of the bench. Every batch, challenge and vote is on Monad, with the data that proves it right or wrong.

Interlude runs the validator, the node and every seat of the bench that rules on a dispute. So in v1 you trust Interlude for two things: keeping the node up, and ruling a dispute honestly. The record does not rest on that trust: every batch lands on Monad with the signed transactions behind it, the hub refuses a batch whose roots do not match, and anyone can challenge a batch with a bond. A wrong batch, and a wrong ruling, stay provable from Monad’s data alone.

A session copies the validator’s bench and terms when it opens and keeps them for life, so when a session opened decides who rules on it. The bench is a dedicated cold resolver key, with the judge bots the validator had seated by then. Every seat is Interlude’s. The live hub carries the fixes from the 2026-09-26 audit: disputes go to that bench, not to the public anvil account the audit found, and the deploy scripts refuse anvil’s public keys anywhere but a local anvil.

Opened since Oct 1, 18:44 UTC

The cold key and three judge bots, A, B and C: any two votes of the four settle a dispute, the bots' among them. 10 MON stake, 100 MON challenge bond, 900 s per dispute move. Every session opened since, an app you ship today among them: Kandle's seven tables (opened Oct 5) and Interlude Exchange's eu, ny and tokyo books, the apps people play today.

Opened from Sep 30, 05:19 UTC, to Oct 1, 18:43 UTC

The cold key and two judge bots, A and B: two votes of three, which the two bots can cast without the cold key. 0.1 MON stake, 0.5 MON challenge bond, 1,800 s per dispute move. Interlude Exchange's previous eu and ny books, retired on Oct 12. The partner apps opened then had closed by Oct 4.

Opened before Sep 30, 05:19 UTC

The cold key alone. 0.1 MON stake, 0.5 MON challenge bond, 1,800 s per dispute move. Kandle's previous seven tables, retired on Oct 12, the Room demo's floors and salons, GridBet v3 and the old lab Room.

The Lie Lab runs on a validator of its own, with the cold key and judges A and B, two votes of three and 0.02 MON stake, 0.01 MON challenge bond, 600 s per dispute move, so a visitor’s dispute never touches a production stake. Every address, and the commands that read each of these back from the chain, are in DEPLOYMENTS.md.

The judges replay a batch once a dispute opens. Interlude runs no watcher on a shipped app: a challenge is open to anyone, you included. To watch the whole game run on Monad testnet (a node that lies on purpose, the challenge, the bisection and the bench’s vote), open the Lie Lab.

Hub v3, the live hub, is 0x98922c6E5e4Bea62761C71D2401c7ec2c26eC43e on MonadScan. Every commit, session and challenge on it can be read there.

The public demo

Eight demo floors on Monad testnet, one writer each: Paris, New York, California, Singapore, Tokyo, India, Johannesburg, São Paulo. India’s floor runs in Singapore, as Fly has no region in India. A tab opens the nearest. Two browsers on the same floor see each other before the next block. The cells are still the contract. The snippets are on The demo app.

Your own app

Same loop, different URL. First hour is the path. ship prints the pair. Do not point the SDK at a public demo node.

The common explorer

Live ticks stay on each node. Settled batches are one log on one Hub. The public explorer reads both: every floor’s engine, and the Hub tape (Committed, sessions, challenges).

A confirmed slash walks the diffs backwards. That is the trust model.