TL;DR

A two-person arbitrage team built a Delta arbitrage bot between Hyperliquid’s HIP-3 stock perpetual markets and the traditional brokerage IBKR. Over 10 months, it processed roughly $32 billion in trading volume and earned about $10 million in profit, with an annualized return on capital of 35%-45%.

The original piece was written by Twitter user CBB (@Cbb0fe), compiled into Chinese by Odaily (translator: Azuma), and published on September 3, 2026. The image-text notes circulating on Xiaohongshu are screenshot highlights of the same article.

Core Strategy: Cross-Market Delta Arbitrage

The logic is surprisingly simple, but the execution is extremely precise:

Pricing benchmark: Treat IBKR (Interactive Brokers) real-time quotes as the “true price” and continuously scan for deviations on Hyperliquid HIP-3.

Arbitrage directions:

• HIP-3 price at a discount to IBKR → go long on HIP-3; once the order fills, open a short hedge on IBKR
• HIP-3 price at a premium to IBKR → go short on HIP-3; once it fills, open a long hedge on IBKR

This is a classic delta-neutral arbitrage — the two legs offset directional risk, and profit comes from spread convergence and the funding rate. The key detail: only after the Hyperliquid-side order fills does the bot build the hedge position on IBKR, which avoids the risk of “one side filled, the other didn’t.”

How Fine-Grained the Parameters Get

The original post disclosed concrete strategy parameter configurations, using NVDA as an example:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

// IBKR side
["NVDA", 55, 400, { maxDelta: 800, sliptage: 0.1 }]

// HIP-3 side
NVDA: pair("NVDA", "xyz:NVDA", {
  makerSize: 400,
  makerOffsetBuy: 0.12,
  makerOffsetSell: 0.12,
  cancelDelta: 0.02,
  takerRatioBuy: 0.05,
  takerRatioSell: 0.1,
  takerMin: 1,
  takerMax: 2000,
  limit: 110000,
  makerEnabled: true,
  preMarketOffset: 0.04
})

These parameters control maker quote offsets, taker ratios, cancellation thresholds, per-order limits, and pre-market offsets — each underlying asset needs to be tuned individually. This is not a “set it and forget it” strategy; it’s a precision machine that requires constant maintenance.

Timeline and Profit Cadence

Key observation: Profitability is heavily driven by macro events — the precious metals rally, geopolitical conflict (Iran), and the semiconductor frenzy each created enormous spreads between Hyperliquid and IBKR. The team themselves admit “there was a lot of luck involved — we happened to be in the right place at the right time.”

The $1.1M Lesson: When Technical Risk Strikes

The most valuable part of this article isn’t the profit bragging — it’s the honest post-mortem of a catastrophic failure:

Root cause: The IBKR API’s data refresh lagged or glitched. The bot incorrectly judged that a Delta imbalance existed between its Hyperliquid and IBKR positions, and kept shorting gold on IBKR to “fix” an exposure that didn’t actually exist.

Consequence: It accumulated a net short position worth $120 million in gold futures — while gold was in a violent rally. After manually closing out, the team was down $1.1 million.

Lessons:

1. Data source reliability is existential — a single data refresh failure can turn “hedged” into “naked one-sided gamble”
2. Scale exceeded design limits — the bot was never designed for that volume and that density of opportunities
3. Missing layered risk controls — there was no hard brake for “position size growing abnormally”

After this incident, the team switched to a direct Nasdaq data feed via Databento to replace IBKR quotes, and designed a dynamic liquidity management system: when IBKR liquidity is low, the bot proactively closes positions to free up capital and demands a wider spread before opening new ones; when liquidity is high, it deploys more aggressively.

Claude AI’s Role: From Learning to Optimization

The article reveals an interesting detail: the two-person team has no formal finance background — “before this, we had never traded a single stock in our lives, and we weren’t even really sure what futures were.”

Learning phase: The author screenshotted everything in the IBKR interface and sent it to Claude, asking “what is this?”, “how do I operate this?”, “how should we hedge XYZ100?” — using AI to learn traditional finance from scratch.

Optimization phase: The team fed all their Hyperliquid and IBKR trading data to Claude for analysis — “where are we losing the most money? What went wrong? What can we improve?” — it was their first time using AI to analyze trading, and the author says it “made a pretty significant difference.”

This echoes a broader trend: AI is lowering the barrier to cross-domain expertise, letting crypto-native players quickly break into traditional finance.

The Team’s Edge: Speed as a Moat

The duo’s core advantage isn’t technology or capital — it’s speed:

They can’t deploy a strategy within 48 hours. They have regulatory constraints, internal processes, approval procedures, and we don’t have any of that.

This is a remarkably honest self-assessment. The institution’s weakness is precisely the retail arbitrageur’s window of opportunity:

• No compliance approvals → rapid iteration
• No internal processes → test directly in production
• No compliance team review → free to operate in gray zones institutions won’t touch

But that window is closing — institutions like Ethena have announced their entry into equity basis trading, which means arbitrage spreads will compress fast.

Cross-Verification

I independently verified the key claims in the article:

✅ Hyperliquid airdrop scale: Decrypt’s reporting confirms Hyperliquid ran a $1.6 billion airdrop, one of the largest in crypto history. The team’s motivation of “waiting for the Hyperliquid Season 3 airdrop” checks out — farming volume for airdrops is a well-known play in the Hyperliquid ecosystem.

✅ HIP-3 is a Hyperliquid product: HIP-3 (Hyperliquid Improvement Proposal 3) is the proposal that introduced traditional financial asset perpetuals into the Hyperliquid ecosystem. TradeXYZ is a stock perpetuals market built on HIP-3, and XYZ100 is its first product.

✅ IBKR’s competitiveness: Interactive Brokers (IBKR) is indeed one of the most mature retail-accessible traditional brokerages globally, with a full-featured API and extremely low fees — a sensible choice for connecting to TradFi.

✅ Databento data source: Databento is a real institutional-grade market data provider offering direct feeds from exchanges like Nasdaq, known for low latency — a reasonable upgrade path away from broker-mediated quotes.

✅ EtherFi’s role as a capital rail: EtherFi, as a liquid restaking protocol, can indeed be used for rapid deployment of large amounts of capital; the article calling it the “MVP” is consistent with reality.

⚠️ What couldn’t be independently verified:

• The specific figures of $32B total volume and $10M profit (self-reported by the author; no publicly auditable on-chain P&L)
• The claim of “contributing 1.5% of trade.xyz’s total volume” (public volume data for trade.xyz is limited)
• The details of the $1.1M loss incident (self-reported by the author)

Given that @Cbb0fe has some reputation in the crypto arbitrage community and Odaily, a mainstream crypto outlet, backed the piece with a compiled translation, overall credibility is fairly high. But the self-reported nature of the data doesn’t change — readers should treat this as a “plausible case study” rather than “audited fact.”

Replicability Assessment

Conclusion: This is a fascinating but not simply replicable case. The real value isn’t “copy this and you’ll make $10M too” — it’s understanding three deeper dynamics:

1. First-mover dividends in new markets: In HIP-3’s early days, institutions hadn’t arrived yet, giving retail arbitrageurs a multi-month window
2. Macro volatility is arbitrage fuel: Without the gold surge, the Iran conflict, and the semiconductor frenzy, there would be no excess profits
3. AI as a cross-domain accelerator: Using Claude to learn TradFi from scratch + using Claude to analyze trading data compressed the “learning curve” from months to days

Risk Disclosure

This article is an analysis and cross-verification of a publicly published arbitrage case. It does not constitute investment advice of any kind. The strategies discussed carry the following risks:

• Technical risk: API failures and data latency can cause catastrophic losses (the article itself documents a $1.1M loss)
• Capital risk: Cross-market arbitrage involves moving large sums between banks, crypto wallets, and brokerages
• Regulatory risk: The legal status of stock perpetual contracts is uncertain across jurisdictions
• Non-replicable returns: Past performance does not guarantee future results, and the author’s figures have not been independently audited
• Market risk: As institutions enter, the arbitrage window is closing

This article is based on the Odaily compiled piece (original author @Cbb0fe, translator Azuma, published 2026-09-03) and publicly available information. Key facts were cross-verified via the Tavily search engine. Cover image generated by LynxCard.