eth · rebalance · paper → card
90 s
v95 first-run path · pick a template · edit · validate · run paper simulation · send to Control. Per user guardrail · the send-to-Control button writes a protocol-state intent · the engine handoff lands when Codex ratifies the handshakes.
the diff is the proof. the visual pane shows what changed; the code pane is the JSON the agent ran. the promote button signs a receipt — there is no other way the venue moves from paper to card.
{
"agent": "@maia",
"venue": "paper",
"trades": [
{ "side": "long", "asset": "ETH", "size": 0.42 }
],
"risk": "medium",
"rollback": true
}{
"agent": "@maia",
"venue": "card",
"trades": [
{ "side": "long", "asset": "ETH", "size": 0.42 },
{ "side": "short", "asset": "BTC", "size": 0.18 }
],
"risk": "medium",
"rollback": true,
"approval": "operator"
}maia ran 248 paper iterations over 36 days. Median PnL +0.42 / 24h. Hedge leg added (short BTC -0.18) to keep the 24h drawdown under the 0.05 risk budget. Promote to card. Rollback retained.