CCatalyst
No edge yet
Evaluator Pack

Review Catalyst without trusting the UI

All numbers on this page are generated from the public receipt, forward ledger, Playbook proof, and cycle index. The evaluator JSON is checked by verify:all.

Updated Jul 24, 14:18
Receiptdcf78add...4d7cb5Open proofLedger665 cyclesOpen proofPaper log70 tradesOpen proofLocal gateverify:allOpen proof

What judges should see first

Catalyst is an autonomous trading agent that knows when not to trade. Open the latest public receipt first: it shows Qwen's current thesis, the governor's approval or veto, and the checksum-backed record for that cycle.

Open latest receipt ->

Sponsor stack is load-bearing

Bitget data feeds the market scan, Qwen generates the thesis, Agent Hub contributes context, and GetAgent proves the Playbook run. Remove those pieces and the loop is not Catalyst.

Open GetAgent proof ->

Honest boundary

Execution evidence is paper/backtest only. The claim is a reproducible perception -> decision -> risk -> receipt loop, not real-capital returns or guaranteed profit.

Open paper log ->
Autonomous cycles
665
from forward ledger
Discipline alpha
+26.7%positive
simulation shadow
Live ledger win rate
62.9%
70 governor-approved paper trades
GetAgent backtest
70.0%
20 backtest trades
Paper trades
70
Track 1 log rows

Latest decision

Low conviction defaults to NO_TRADE.

PairSOLUSDT
Qwen actionNo Trade
GovernorNo edge yet
Checksumdcf78add...4d7cb5
Open receipt JSON ->

Metric source split

Two win rates appear because they measure different proof surfaces, not conflicting totals.

GetAgent backtest70.0% / 20 trades
Live ledger62.9% / 70 approvals
Runpbrun-b8eb40e05307
Modesimulation shadow
Open evaluator JSON ->
Track 1 paper trading log

Approved paper trades with required fields

Download CSV ->Open JSON ->

Derived from committed approval receipts only. Balance and quantity are normalized paper-account values, not real capital.

TimePairDirectionPriceQuantityBalance changeAfterReceipt
Jul 21, 14:27ETHUSDTLong1845.30.008861+0.75140.519a18cb0...c5dd70
Jul 21, 10:36SOLUSDTLong75.00.25016+0.84139.8a4fd2fa2...90943f
Jul 21, 10:35ETHUSDTLong1844.00.009697+0.94138.91509405a...e33014
Jul 21, 06:45SOLUSDTLong74.90.237111+0.88138.006d04292...e2edba
Jul 21, 06:44ETHUSDTLong1846.50.009298+0.84137.1e5a75eba...eb35be
Jul 21, 03:50SOLUSDTLong74.90.211761+0.71136.31ea1aa5c...0c289e
Jul 16, 06:34ETHUSDTLong1789.20.009296-0.33135.6bdda39d6...830175
Jul 16, 03:42ETHUSDTLong1780.80.009788+1.39135.9843b3391...58acdb
Public artifacts

Open the proof directly

ArtifactPurposeLink
Evaluator proof JSONOne-file summary generated from all public proof surfacesOpen ->
Latest receiptCurrent autonomous decision, checksum, thesis, council, and governor stateOpen ->
Forward ledgerAutonomous cycle history and simulation-shadow discipline metricsOpen ->
Per-decision CSVDownloadable rows for hand-checking discipline alphaOpen ->
Paper trading log JSONTrack 1 paper-trade rows with timestamp, pair, direction, price, quantity, and balance changeOpen ->
Paper trading log CSVSpreadsheet-friendly version of the Track 1 paper-trading logOpen ->
GetAgent Playbook proofCompleted GetAgent API run proof; no public rank claimedOpen ->

Local verification

Run the same gate used by the scheduled cycle workflow. The evaluator pack fails if it is hand-edited or stale relative to the source JSON.

Commandnpm ci
Commandnpm test
Commandnpm run build
Commandnpm run verify:all

Honest caveats

Simulation shadow, not real-capital PnL.
No profit guarantee and not financial advice.
Agent Hub fallback rows are labeled and do not count as live convergence.
The latest cycle can approve, veto, or stand down; the proof is the recorded decision path.

Verifier coverage: latest receipt checksum, decision-pair proof, discipline-alpha drift, calibration drift, council drift, paper-trading-log drift, evaluator-pack drift