▮ AUTONOMOUS TRADING FIRM · 12 DESKS · EXECUTES ON BITGET · ANCHORED ON MANTLE
Most bots
redline.
Ours holds gear.
Twelve specialist desks — regime, smart-money, Allora macro, on-chain, vol-forecasting — converge into a PM that executes on Bitget and anchors every call on Mantle. It fires ENTERonly on a validated edge; when there isn't one the gearbox holds N. The discipline to hold is the product.
every figure below is real — the wins, the retired edge, the rejected backtests
regime confidence · OOS
82.6%
registry empty → holds N
▮ THE LOOP · IT RUNS ITSELF · 24/7
A whole firm on one closed cycle. No off switch.
A steward wakes the firm, telemetry fans to all twelve desks, the PM rules, the risk gate sizes, a rare ENTER fills on Bitget and anchors on Mantle — then the edge-lab learns, memory stores it, and the wire loops back. Hover any node to see what it does.
the real firm, self-running · steward → 12 desks → PM → risk → Bitget → Mantle → edge-lab learns → memory → loop · mostly ABSTAINs (discipline)
▮ THE FIRM · ONE CYCLE · EVERY READ
A four-stroke decision engine
Scan, argue, decide, seal — the same cycle on every read. Today there is no validated edge on the floor, so combustion stays dark. That no-spark is the product.
Intake
Nine desks pull live telemetry — positioning, funding, whales, TVL, macro.
Compression
Bull argues bear. Nine stances compress into one pressurized read.
Combustion
The PM is the spark plug. ENTER ignites only on a validated edge.
registry empty → gear holds N
Exhaust
The decision — even the restraint — is sealed on-chain on Mantle.
▮ THE EDGE LAB · COST-AWARE OOS · WALK-FORWARD · FDR
The dyno doesn't lie
dyno run · OI-CONTRARIAN · MNTUSDT · out-of-sample
$1,000 → $1,289
Decayed
retired by the lab
It beat costs out-of-sample, hedge-like, while the asset bled. Then fresh full-history data said the pattern was gone — so the lab pulled its license. An edge you won't retire isn't an edge, it's a belief.
24
hypotheses in
23
scrapped
1
passed → retired
the scrap shelf · cause of death
- BTC MOMENTUMfee-eaten
- ETH/SOL TRENDOOS negative
- FUNDING SIGNALnoise
- LONG-SHORT RATIOnoise
- HYPE FLOWone lucky fold
- BTC FLOW ×2false positives — killed by FDR
- mETH/ETH CONVERGENCEsee crash test ↓
gates: cost-aware OOS · anchored walk-forward · outlier-robust · Benjamini-Hochberg FDR
▮ HONESTY LAB · FINDING #15
We crash-test our own winners
the backtest · mETH/ETH convergence
+96.0%
95.6% win · paper
The kind of figure that headlines a pitch deck. We put it on the sled instead.
impact report · realistic execution
−73 bps
per trade · thin mETH slippage
Priced against the real order book, every "win" paid more in slippage than it earned. A thin-liquidity artifact — we threw it out.
Rejected
a firm that publishes its crashes is a firm whose passes mean something
●QC INSPECTION SHEET · 28 FINDINGS · WE PUBLISH WHAT DOESN'T WORK
The shop report
Every claim hits the bench as a part. Some we keep; most we scrap. Here is the log — tinted by verdict, stamped, and linked in full.
bench (shown)
4 kept · 6 scrapped
- REAL
Accuracy ≠ profit
A model right 70% of the time still loses if the 30% are the big moves.
- REAL
The one real edge: OI-contrarian
Hedge-like, +28.9% OOS on MNT — earned on its window, not a forward promise.
- REAL
Honesty-by-design
No number is shown that a real run this session didn't print.
- REAL
Verifiable on Mantle
Every gated decision sealed on-chain — the ledger is public, not a screenshot.
- RETIRED
We validated MNT — then our own loop RETIRED it
The self-learning engine demoted its own flagship when the edge decayed.
- ABSTAIN
We tested ML too — no free pass
RandomForest + XGBoost on MNT: ~51% accuracy, −50/−62% OOS. Abstain.
- REJECTED
We rejected our own +96% backtest
mETH/ETH convergence — priced at real slippage it paid −73 bps/trade. Out.
- REJECTED
A collaborator's “+36% BTC” ML
Re-run honest on 10× data with fees + walk-forward: −12.8%. A backtest isn't an edge.
- DECAYED
HYPE's +142% snapshot DECAYED
Probation, not graduation — days later it earned no edge at all. We'd already held.
- REJECTED
We found a +64% edge — and rejected it live
MNT OI-contrarian, p=0.025 — but unstable at the 0.9 cutoff. Registry stays empty.
ten of twenty-eight, signed off — a firm that logs its scraps is one whose keeps mean something
Read all 28 →▮ VERIFIABILITY · MANTLE · PUBLIC READ
The black box is public
Flight recorders survive the crash — that's the point. Every cycle's decision, even the restraint, lands in a public table and anchors to Mantle.
first anchor · proof-stone
BLOCK
39,402,623
0xa052ee03…1de69b53
MNTUSDT · ABSTAIN · sealed on Mantle
lap log · decisions_hq
documented
live feed unreachable here — rows above are the documented anchors
▮ THE MACHINERY · ALWAYS-ON · RAILWAY + MANTLE
Runs while you sleep
THE ORG BRAIN
9-desk firm + PM debate loop · Groq LLMs · runs every cycle in the cloud
THE EDGE LAB
cost-aware OOS · walk-forward · outlier-robust · FDR — the dyno itself
THE EXECUTOR
Bitget testnet rails — real demo fills · Kelly-capped tickets
THE CONTRACTS
ERC-8004 identity · TradingVault · JobManager — deployed on Mantle
THE RECORDER
every decision sealed: Supabase public table + Mantle anchor
THE FEEDERS
local telemetry rigs push positioning, carry & whale data hourly