Market-neutral crypto trading, supervised by an AI risk copilot — 24/7.
Fully automated volatility-harvesting strategies on major exchange perpetuals, hedged on both sides of the market, with an institutional-grade AI agent watching every account around the clock.
01How the strategy works
The engine earns from volatility, not direction. It runs paired long and short positions on the same assets across two quote markets, continuously buying dips and selling rips inside a dense order ladder — thousands of small, disciplined trades instead of a few large bets.
◆Both sides, always
Every asset trades with a long leg and a short leg held in balance. When the market moves, one side realizes profit while the other accumulates inventory the ladder later unwinds — net market exposure stays close to zero.
◆Volatility is the yield
Crypto ranges far more than it trends. Each oscillation through the order ladder converts price noise into realized profit, on liquid perpetual markets with deep order books.
◆Engineered sizing
Per-leg order sizing, laddered averaging with hard caps, automatic take-profit management, and leverage guards — all parameters battle-tested on multi-year historical data before touching live capital.
02The AI copilot — a risk manager that never sleeps
Every account is supervised by an autonomous AI agent built on frontier large-language models. It does not place random trades — it works like a professional risk desk with hard-coded limits it can never exceed.
Observe
Every few minutes: positions, hedge balance, margin utilization, liquidation distance, realized PnL, live order ladders, funding costs, and a proprietary market-intelligence feed (volatility, trend strength, liquidity, cross-market dislocations, open-interest flows, positioning crowding) — computed once and shared across all accounts, hedge-fund style.
Assess & critique
An AI analyst reviews the account against your risk targets; an independent AI critic then tries to reject every proposed action. Only actions that survive both — plus deterministic code-side validation — go forward.
Guardrails decide
Hard limits enforced in code, not prompts: per-action dollar caps, daily action caps, per-asset cooldowns, loss-adaptive pauses, fleet-level coordination so accounts never compete with each other, and platform ceilings no setting can exceed. Anything that would realize a meaningful loss always requires the human’s explicit confirmation.
Act — then measure itself
Approved actions execute automatically (or arrive as one-tap Confirm buttons in Telegram — your choice). Every action is then tracked: its own volume is followed from open to close and marked out at 1 hour and 24 hours, so the copilot learns from its real track record — and you see it.
03Your personal Telegram command center
Each client gets a private Telegram bot. Choose your notification style — from full silence with a dashboard audit trail, to every action as a Confirm/Cancel button.
◆Ask anything
The same bot is a full AI assistant over your account: “how did I do this week?”, “why is margin up?”, “show my exposure” — answered from live data, with charts.
◆Dashboard included
Live portfolio, per-leg PnL decomposition (price / fees / funding), risk timeline, market analytics, the copilot’s full action history with its reasoning, and per-leg sizing controls that apply live — no restarts.
◆Modes
Auto — the copilot acts silently inside guardrails. Propose — every action waits for your tap. Switch anytime; loss-realizing actions always ask first.
04Real performance — real accounts
Live client accounts, anonymized. Realized trading PnL net of all fees and funding; deposits and withdrawals are excluded by transfer detection. This is money actually settled — not paper marks.
Cumulative realized PnL — % of starting equity
Daily realized PnL — Account A
Margin utilization — capital stays safe
05Risks — stated plainly
No strategy is risk-free, and past performance does not guarantee future results. Here is exactly what can hurt, and what stands between each risk and your capital.
Sustained one-way trends
Range-harvesting strategies accumulate inventory against strong trends. Drawdown days happen — the worst single day across the accounts above was about 3.5% of that account’s equity, and deeper episodes are possible.
Leverage & liquidation
Positions use exchange leverage. Utilization is deliberately kept near ~7% on average with hard ceilings, liquidation-distance monitoring and automatic de-risking — but extreme gap moves are a real tail risk.
Exchange & infrastructure
Exchange outages, API limits, or liquidity evaporation on thin assets can delay position management. The platform runs redundancy, rate-limit budgeting and stale-data protection, and the copilot pauses itself when its data degrades.
Funding & fees
Perpetual funding and taker fees are a constant drag the strategy must out-earn. Both are measured per leg and fed to the copilot — a structurally bleeding leg gets slowed or cut.
◆The safety stack
- Daily loss halt: past a realized-loss threshold you set, the copilot stops proposing changes for the day.
- Liquidation-proximity and drawdown auto-pause; margin-utilization target band actively managed.
- Loss-realizing actions never execute autonomously — they always require your confirmation.
- Per-action and per-day caps, per-asset cooldowns, platform-level hard ceilings above all user settings.
- Fleet coordination: accounts never crowd the same thin asset, capped against total open interest.
- Full audit trail: every review, every considered-but-rejected action, every execution — visible to you.