Satsuma
S

@satsuma/experiments

PublicThe system
Clone

Governed runs of Satsuma trading on tangerine over ISMP. Results are published as articles when a run completes.

Lab runs, not the live system. Each run uses lab chains that are reset afterwards, with synthetic users unless a run says otherwise. Nothing here happened on the Satsuma chain.
RunsLab run · 11 November 2026, 00:47 UTC

Product improvements

Can an improvement be proposed, approved, built, measured to a verdict on chain and removed when it does not help, with no outside help?

35/38 passed3 inconclusive · 4 h 35 min

Setup

Chains
Satsuma lab chain
  • RoleHome
  • KeptLab, not retained
tangerine lab chain
  • RoleProving
  • KeptLab, not retained
Users
  • UsersSynthetic
Model
Claude Opus
  • ModeLive and replayed
Operator seat
  • SeatBy delegate
Unattended
Yes: nobody was watching to step in
Time
4 h 35 min · ended 05:22 UTC

Checklist 38 items

Every item, unless its row says otherwise
  • ChainLab
  • UsersSynthetic

Trying and judging improvements

31/34 passed
InconclusiveThe council approves the improvement2 of 3 attempts
  • DecisionsLive model
InconclusiveThe improvement is built and lands as an upgrade2 of 3 attempts
  • DecisionsLive model
InconclusiveThe improvement that did not help is removed with no outside help2 of 3 attempts
  • DecisionsLive model
Pass31 items passedShow them
PassNo improvement is in progress when the gates are tested1 of 1 attempts
  • DecisionsLive model
PassA measured baseline exists to judge an improvement against1 of 1 attempts
  • DecisionsLive model
PassA proposed improvement with no plan for removing it is refused1 of 1 attempts
  • DecisionsLive model
PassA product that is below its threshold, under assessment or frozen takes no improvement1 of 1 attempts
  • DecisionsLive model
PassA second improvement while one is in progress is refused1 of 1 attempts
  • DecisionsLive model
  • first0x9018ee4ce45a35d6402d51fe71fbe3c08679e3d27c5e97be4b9f0e21a7e739d0Satsuma lab chain, not retained
PassA council seat sponsors a proposed improvement1 of 1 attempts
  • DecisionsLive model
PassThe council approves the improvement1 of 1 attempts
  • DecisionsLive model
PassThe improvement is built and lands as an upgrade1 of 1 attempts
  • DecisionsLive model
PassThe improvement is judged on chain from measured use1 of 1 attempts
  • DecisionsLive model
PassAn improvement with no effect is judged as having missed its target1 of 1 attempts
  • DecisionsLive model
PassThe improvement that did not help is removed with no outside help1 of 1 attempts
  • DecisionsLive model
PassA council seat sponsors a proposed improvement1 of 1 attempts
  • DecisionsLive model
PassThe council approves the improvement1 of 1 attempts
  • DecisionsLive model
PassThe improvement is built and lands as an upgrade1 of 1 attempts
  • DecisionsLive model
PassThe improvement is judged on chain from measured use1 of 1 attempts
  • DecisionsLive model
PassA harmful improvement is judged harmful1 of 1 attempts
  • DecisionsLive model
PassThe improvement that did not help is removed with no outside help1 of 1 attempts
  • DecisionsLive model
PassA council seat sponsors a proposed improvement1 of 1 attempts
  • DecisionsLive model
PassThe council approves the improvement1 of 1 attempts
  • DecisionsLive model
PassThe improvement is built and lands as an upgrade1 of 1 attempts
  • DecisionsLive model
PassThe improvement is judged on chain from measured use1 of 1 attempts
  • DecisionsLive model
PassA council seat sponsors a proposed improvement3 of 3 attempts
  • DecisionsLive model
PassThe improvement is judged on chain from measured use3 of 3 attempts
  • DecisionsLive model
PassAn improvement with no effect is judged as having missed its target3 of 3 attempts
  • DecisionsLive model
PassThe council sponsors the improvement drafted by the product reviewer2 of 2 attempts
  • DecisionsLive model
PassThe draft stored by the system and the one recorded on chain are the same document2 of 2 attempts
  • DecisionsLive model
PassImprovements are gated on the product's real standing2 of 2 attempts
  • DecisionsLive model
PassThe improvement is approved and switched on2 of 2 attempts
  • DecisionsLive model
PassThe verdict comes from the measurement windows recorded on chain2 of 2 attempts
  • DecisionsLive model
PassThe improvement is removed when it does not meet its target2 of 2 attempts
  • DecisionsLive model
PassA corrective change with no written draft is not sponsored2 of 2 attempts
  • DecisionsLive model

Council governance

1/1 passed
PassOn a protected product, an improvement needs the operator's seat as well as the model's1 of 1 attempts
  • DecisionsLive model
  • operatorSeat0xa0Ee7A142d267C1f36714E4a8F75612F20a79720Satsuma lab chain, not retained
  • council.modelSeat0x3Eb10640025Ef1534E932D857D2Ac3619Db7623CSatsuma lab chain, not retained

Product direction from measured use

1/1 passed
PassThe product's standing counts the improvement's verdict2 of 2 attempts
  • DecisionsLive model

Lineage to the objective

2/2 passed
PassThe record of objectives and products passes its audit2 of 2 attempts
  • DecisionsLive model
PassThe improvement appears in the record of objectives and products2 of 2 attempts
  • DecisionsLive model

What this proves

  • Whether, in this run, improvements went from sponsorship to verdict, and whether ones that missed or did harm were removed.

What it does not

  • That a judged improvement helps in real use: the use here is driven by the test harness.

What broke, what was fixed

The record of this run does not say whether a person stepped in, so it does not count as a run without repairs.

Cost in tokens

Not recorded: this run has no chain-derived scorecard.

Archived evidence

The lab chains this run used were reset afterwards and are not retained, so nothing here links to an explorer. The record below is the archived copy.

Scorecard
None archived.