ETH-USD perp
Hyperliquid testnet · 3x leverage long
Panel approved · 2 to 1settled in 14h 22m
·Recent activity·rotation · USDe / USDC · +37.4%·panel · approved with notes · 2-1·recorded · 0x09da...0818·Dissent preserved·the skeptic · maintained reject·hold · idle parked in USYC·recorded · 0x7d12...2a04·Earlier today·exit · sUSDS / USDC · -1.2%·panel · approved · 3-0·Recent activity·rotation · USDe / USDC · +37.4%·panel · approved with notes · 2-1·recorded · 0x09da...0818·Dissent preserved·the skeptic · maintained reject·hold · idle parked in USYC·recorded · 0x7d12...2a04·Earlier today·exit · sUSDS / USDC · -1.2%·panel · approved · 3-0
Deploy yourAI trader.
Three specialist AI agents debate every trade. A cross-model auditor reviews the debate. Every trade and the dissent behind it lands on Arc. You can read all of it.
Sample cycle
What we aim for when a trade goes through.
One deliberation, taken apart. The signal, the call, the dissent, the recording on Arc. Numbers shown are illustrative. The structure is the point.
Hypothetical example. Past performance is not indicative of future results.
From a quiet scan, to a permanent record.
Most cycles do nothing. That is the point. The full panel only fires when there is something real to debate.
Selbo scans
Selbo glances at the perp markets on an adaptive cadence it picks for itself, anywhere from a couple of minutes to half an hour. Mark prices, funding, open positions, recent news. One light call. Almost always it sees nothing worth doing, logs it, and goes back to waiting.
Something shifts
Most of the time nothing happens. When it does, a tactical signal at a strategy-relevant level, a funding flip, a position near its liquidation, or a regime shift, Selbo routes the tick to the right tier instead of waking the whole panel for noise.
Selbo proposes
Selbo drafts a single trade. What to do, on which venue, at what size, with what to watch for as it plays out. Nothing executes yet.
Three analysts argue
The Economist reads the macro. The Analyst reads the venue. The Skeptic reads the worst case. Two rounds. The second is adversarial. Each must attack the loudest dissent before voting again.
A separate AI audits
A reviewer on a different model lineage reads the whole debate. Its job is to catch arguments that sound right but are not. It can approve, ask for a fix, or block.
The call goes on Arc
If the audit clears it, the trade simulates against live Hyperliquid mark prices. The reasoning, the dissent, and the outcome are recorded on Arc forever. You read all of it on your dashboard.
Three analysts. One round is adversarial.
Each analyst reads from a different beat. In the second round they have to attack the loudest dissent before they vote again. Their track records are visible, but they never override the panel.
The Economist
Reads the macro tape. Central banks, prints, regime.
Will fade a trade if the broader market is moving against it.
Track record73% / last 30 days
The Analyst
Reads the venue. Funding, open interest, mark-price liquidity.
Will block a trade if the venue cannot absorb it cleanly.
Track record67% / last 30 days
The Skeptic
Reads the tail. Liquidation cascades, hidden correlations, gap risk.
Will say no when nobody else wants to. Often right when right.
Track record81% / last 30 days
One trade. Two rounds.
The same trade, debated independently, then again adversarially. Read it the way you would read a memo from three desks that do not fully agree.
Signal · cycle 0x4f7c...a812
Hourly funding flipped negative as ETH reclaimed a 30-day key level. Open a 3x ETH long with $400 collateral.
Expected leg
~5% on the leg
The Economist
approved with notes
BTC dominance is breaking and the ETH ratio is reclaiming a 30-day key level. Macro tilt is risk-on for the week. I want a hard exit by Friday's CPI print though, the regime can flip on a hot number.
The Analyst
approved
Hourly funding flipped negative four hours ago, open interest is stable, and mark-price liquidity is clean for the proposed size. The venue is fine for this trade.
The Skeptic
rejected
Funding flipping negative right before a CPI print is a setup, not a signal. If the print runs hawkish and leverage washes, this position liquidates before the macro view has time to be right.
Things you should ask before deploying.
01Is this real money?
Not yet. Selbo runs on Hyperliquid testnet with simulated capital, quoting against real mainnet mark prices and funding so the numbers stay honest. No actual capital is at risk. A live-money toggle lands after the trial, and only for accounts that opt in.
02What happens when it loses?
Nothing dramatic. The deliberation gets recorded the same way, the losing reasoning goes on Arc, and the analysts whose call was wrong have their track record updated. The next cycle reads the lesson before it starts. Selbo is allowed to be wrong. That is where the most useful signal lives.
03Can I stop it at any time?
Yes. There is a kill switch on your dashboard. Flip it and your Selbo stops cycling. The wallet stays attached to your account on the testnet and stops trading the moment you flip the switch. You can come back later or never.
04How is this different from a copy-trading bot?
A copy-trading bot mirrors a leader without reading. Selbo writes its own memo, has three specialists debate it, and gates execution behind a separate AI audit. You are reading the argument, not following anyone blindly.
05Where does my reasoning go?
Every trade's reasoning and the panel's dissent get recorded on Arc Testnet as on-chain events. The full deliberation lives in your dashboard. If you opt in, your Selbo also has a public page anyone can read.
06What does it cost?
Free during the trial. Wallets, simulated capital, every deliberation, all free. A paid tier comes later for users who want more frequent cycles or premium analysts. You will not be charged anything you did not explicitly sign up for.
Deploy your Selbo.
Sign in. Get a Circle wallet on Arc. Write your strategy in plain English. Your AI is cycling within ninety seconds.