OMNI saysWhen a worker commits a bad block, VERIFY catches it and the orchestra ROLLBACKs the chain to the last known-good — apoptosis ends a worker, rollback undoes its damage.
Apoptosis (mol-65) handles a worker that finishes; rollback handles a worker that finishes wrong. The two are different tools. When a critic's VERIFY fails after a result has already been COMMITted — or when downstream evidence proves an earlier block was bad — the orchestra performs a ROLLBACK: it reverts the per-user chain to the last block that VERIFYed clean, discarding everything after it. The bad worker is already gone (it apoptosed); rollback removes its footprint from the durable record.
Rollback is precise because the chain is append-only and hash-linked: each block names its parent, so reverting to block N is unambiguous and replay from N forward is deterministic. After a rollback the conductor can respawn the failed stage with a tighter contract — smaller scope, stricter critic, lower budget — so the second attempt is less likely to repeat the fault. Detect (VERIFY) → revert (ROLLBACK) → respawn is the swarm's self-healing loop.
Power-ups you unlock
Apoptosis ends a worker; ROLLBACK undoes a worker’s committed damage
VERIFY failure (or later evidence) triggers revert to last known-good block
The hash-linked chain makes reverting to block N unambiguous and replayable
Respawn the failed stage with a tighter contract after rollback
Detect → revert → respawn = the swarm’s self-healing loop
The Silent Failure attacks — common mistakes
Trying to patch a bad block in place instead of rolling back (chain is append-only)
Rolling back but respawning with the identical contract → same failure
No known-good marker, so you cannot tell how far back to revert
Confusing apoptosis (normal exit) with rollback (error recovery)
Boss battleCommit three blocks, fail VERIFY on the third, ROLLBACK to the last good block, then respawn that stage with a smaller budget and show it pass.