
Manifold Founder Shares: How to Avoid Giants Like Jump and Capture CEX-DEX Spread Arbitrage?
Author: Jae Chung
Compiled by: TechFlow
TechFlow Summary: Manifold’s early edge came from arbitraging price discrepancies between centralized and decentralized exchanges. At the time, giants like Jump and Tower were entrenched in CEX infrastructure, giving small teams a window to carve out a slice from on-chain inefficiencies. This founder’s narrative breaks down how a zero-experience team selected the right tables, found their edges, and iterated rapidly—a highly valuable reference for practitioners looking to enter quantitative or on-chain trading.
Why Start a Quantitative Trading Firm?
In 2021, I barely slept. Naturally, I wanted to find a way to make money while I slept.
I dreamed of having bots print money for me while I lay comfortably in bed.
As a naive 21-year-old, I thought: How hard could it be?
All my advisors told me it was a stupid idea. I had absolutely no clue who my competitors were (Jump, Tower, Jane Street), and I had zero experience in the field.
They were right.
Before founding Manifold:
I entered the crypto space in 2016/17 as a white-hat hacker
I ran validator nodes for a few years
I made some lucky trades and investments during DeFi Summer and the NFT frenzy
None of this had anything to do with quantitative trading. In fact, I didn’t even know what “alpha” was!
But as usual, I didn’t listen to my advisors. Five years later, I’ve managed to get decent at this (right shoku?), but the 21-year-old me was incredibly stubborn and preferred learning the hard way.
The Early Days
Many people think quantitative trading is just:
Find a pattern
Make money
I used to be one of them. But it’s far from that simple.
Some of our early iterations on trading strategies were quite embarrassing.
We tried everything. Arbitrage, spread capture, basis trading, statistical arbitrage, yield farming, even deep learning!
At first, none of them worked. Trading is a zero-sum game. If enough people play the same game better than you, all the meat off the bones will be gnawed away.
We had to find our own edge. But what would it be?
Picking the Right Table
Most top-tier firms typically excel in one of the following areas:
Speed/Latency (trading infrastructure)
Proprietary data/order flow
Research/alpha generation
On centralized exchanges, these areas have long been completely dominated by firms like Jump and Tower.
They’ve deployed state-of-the-art low-latency infrastructure across all major exchanges (Binance, OKX, Bybit, Coinbase, etc.). They hold fee tiers we can’t reach since we don’t command the same trading volume. Their clients provide them with lucrative order flow and data.
Our trading and research infrastructure was merely entry-level, and we didn’t have an army of engineers and researchers to fight on their turf.
We needed to find a different game first and build our advantage there.
I realized our edge had to lie in DeFi, specifically on decentralized exchanges.
At the time, most blockchain foundations were pumping massive capital to incentivize users to provide passive liquidity to DEXs. Most DEXs lacked order books, instead relying on AMM (Automated Market Maker) models where trades execute along pricing curves.
Simply put, DEXs require a unique trading system capable of handling both market data (prices, liquidity ranges/slippage) and order submission (trade execution/confirmation).
Top firms hadn’t gone too crazy in this space yet, likely because: a) it deviates from their core infrastructure strengths; b) the market is smaller compared to classic CEXs; c) regulatory uncertainty surrounds DEXs; d) they’re already making plenty of money elsewhere.
The talent requirements also suited our team better. We had a unique mix: crypto-native engineers familiar with smart contracts (essential for on-chain exchange execution) working alongside quant researchers from more traditional backgrounds (Citadel, Tower, etc.).
For example, one of our best hires later was formerly a data engineer at an insurance company. Privately, he was also a solo MEV searcher, running his own atomic arbitrage strategies at scale on the Polygon blockchain. Before he joined, I communicated with him via his anonymous email, 0xaddress@protonmail, and saw potential just by scrolling through his bot’s transaction history on a block explorer.
Before shifting the whole team’s focus to DEXs, I validated the concept myself: I wrote a simple on-chain trading system entirely in TypeScript to run an arbitrage strategy between CEXs and DEXs (comparing Binance/FTX against the top ten blockchains by liquidity). We called this CEX-DEX arbitrage.

This opportunity existed because blockchains “block” transactions over a certain block time (ranging from 200ms to 1-2 seconds, or even several seconds). This is inherently slower than every single quote on a CEX, creating a consistent lead-lag relationship. Most liquidity on DEXs is passive/outdated and not actively managed by bots, leading to price discrepancies across venues since price discovery mainly happens off-chain.
Even with such a rough prototype that wasn’t latency-sensitive, the strategy still captured real arbitrage opportunities and was profitable after fees!
After months of research hitting dead ends, I finally saw a glimmer of potential.
Finding Our Edge
What was exciting about this strategy was that even in its prototype stage, it actually worked.
For arbitrage strategies, it’s rare for a prototype to work out of the gate. By definition, arbitrage is risk-free profit left on the table. Usually, someone else can grab it before you (faster than you), causing the opportunity to vanish or blur.
We already knew several ways to improve the strategy:
Rewrite the TypeScript system in another language (for speed)
Add more DEXs per chain and add more chains (more arbitrage opportunities)
More precise mathematical modeling (position sizing, slippage)
Fee optimization (improve profit margins per trade)
Our CEXs were nowhere near top-tier fee tiers, and fees would drop further as volume scaled
Implement creative methods to lower Gas fees per on-chain trade (blockchain network fees)
Improved on-chain execution techniques (increase fill rates)
Inventory optimization (boost returns, capital efficiency, and uptime)
From here on, the key was purely execution.
One benefit of high-frequency trading is the immediate feedback from the market. When you ship something and run the strategy, you almost instantly get a binary, black-and-white result: does it increase profitability? This gave us an extremely fast feedback loop between “engineering implementation” and “becoming dollars.”
Quick anecdote:
I remember December 24, 2021, when Sid (my co-founder) and I sat in our apartment office coding all night. We had no weekends and no holidays. Ironically, founding Manifold forced me to work more and sleep less.
Amidst the constant clacking of keyboards, neither of us noticed Christmas had arrived. It wasn’t until around 1 AM that I checked the time and said, “Sid, it’s Christmas!”
Thirty seconds later (I thought he completely ignored me, but he’d actually just forgotten about it and kept working) he replied, “Oh. Merry Christmas.”
And then we got back to coding.
Looking back, this was probably my favorite office, and everything started here.

Scaling the Edge and Monetizing It
This article is longer than I intended. I need to practice writing concisely.
To save time, I’ll only dive into a few specific improvements we made to CEX-DEX arbitrage that boosted profitability.
If you’re not interested in trading minutiae, feel free to skip this section. But maybe some of these tactics will inspire you to find and refine your own edge in another market—or even an entirely different industry. I’ll try to explain using plain English rather than math formulas and code.
While implementing those general improvements, CEX-DEX arbitrage began generating over $10k daily with only a few million in deployed capital. Our initial focus was getting the new system (written in Go + Solidity) robust enough to scale across multiple chains, targeting 10-20 DEXs per chain. We also began integrating newer DEXs that utilized the UniV3 tick pricing model. These DEXs offered higher efficiency due to concentrated liquidity and tighter price slippage. As we scaled, our CEX trading fees dropped because our volume tier increased. With growing capital, we added more trading pairs to our pools.
The Evolving Game
As we picked the low-hanging fruit, we noticed competition ramping up quickly. On some chains, we could no longer trigger arbitrage with a 10-basis-point spread because other bots were willing to execute at lower thresholds. The game evolved from simple “spot and execute” bots to ones requiring slightly higher technical sophistication. These automated bots had already displaced the original manual clickers.
Quick example:
ETH is $2,000 on Binance.
ETH on Quickswap (Polygon) is $2,001.

The spread is 2,001/2,000 = 0.05% = 5 basis points.
A bot (assuming it accounts for fees and slippage) is programmed to trigger on any 5-basis-point arbitrage opportunity. It will sell ETH on Polygon and buy ETH on Binance, pushing prices back toward parity.
This means our bot never sees the arbitrage opportunity because another bot closes the spread before ETH on Polygon hits $2,002—that’s the 10-basis-point spread we were originally after.
Of course, everyone’s cost per arbitrage trade differs. It depends on several factors: a) their CEX fee tiers; b) whether they act as takers or makers on the CEX side (maker orders are usually cheaper and sometimes come with rebates); c) their Gas fees (blockchain costs to process trades, detailed shortly).
We started encountering firms like Wintermute. We had to deploy Priority Gas Auctions (PGA) across a set of bot addresses to bid higher on specific transactions. This is a technique borrowed from the MEV (Maximal Extractable Value) space. You repeatedly resubmit a specific transaction with a higher priority gas fee to increase its inclusion priority within a given block.
To participate in these auctions, you must:
Accurately estimate how much you’ll earn from a specific trade to determine the exact fee you’re willing to pay
Possess reliable infrastructure that allows you to land the trade within a single block after spotting the opportunity
Know which addresses you’re bidding against and quickly realize whether you won (or lost) the auction
Different blockchains employ similar or adjacent techniques depending on their transaction ordering and submission architectures. Blockchains constantly alter rules and mechanisms, attempting to direct more value accumulation toward users, protocols, or the chains themselves. Arbitrageurs must continuously adapt to these shifts.
Gas Optimization
The total Gas cost to submit an on-chain transaction is roughly a function of the gas units required multiplied by the current gas price. Gas price fluctuates based on real-time network congestion.
Gas price/priority fees come into play during PGA-style auctions. As long as you can land the trade within a single block, there’s limited room for further optimization here.
However, significant reductions in gas consumption can be achieved through code-level optimization. This linearly decreases the on-chain costs required to execute trades.
Among many techniques, one fascinating approach we employed is efficient bit packing. The Ethereum Virtual Machine (EVM) stores data in 32-byte (or 256-bit) slots. By placing smaller data types adjacently, we can pack them into a single slot via Solidity. This dramatically reduces the execution cost of read/write operations (SSTORE and SLOAD).


As you can see, our trades feature significantly less zero-padding. This reduces the overall size we must pay for, thereby lowering the total cost of our transactions.
Now, we can exploit opportunities that remain “invisible” to those paying higher Gas fees, as these optimizations render them profitable. For instance, an arbitrage trade might net us $1 after fees, but leave unoptimized bots with $0. We close the spread before other bots even register the opportunity.
This also allows us to bid higher on trades to “win” the arbitrage while maintaining profitability. Over time, winning more arbitrages means not just higher profits, but also greater volume. This compounds into more advantages: better CEX fee tiers and the ability to capture larger spreads per trade.
Capital Efficiency
With continuous optimizations and improvements, we went all-in on scaling this strategy. We expanded our footprint by integrating more chains, DEXs, and trading pairs. The broader the trading scope, the higher the capacity. Allocating the majority of our capital to CEX-DEX arbitrage was sound. The strategy boasted an annualized return exceeding 100% with zero drawdown days.
But even though our capital grew rapidly alongside profits, we recognized the strategy had more headroom. How could we squeeze more out of our existing funds?
The issue at the time was that CEX-DEX arbitrage required capital distributed across various venues. It also demanded inventory preparation for several different trading pairs.
Take a simple example. If you’re arbitraging ETH/USDT between Binance and Arbitrum, you need both ETH and USDT ready at both venues. Say you initially deploy $500k worth of each asset at both locations, and Binance carries a premium. You’ll ultimately sell ETH on Binance while buying ETH on Arbitrum. If the premium persists, you might quickly sell out your ETH on Binance. Simultaneously, your USDT on Arbitrum gets depleted from buying ETH.
CEXs offer margin services, allowing you to leverage positions and continue trading when inventory becomes unbalanced. But on these chains, if inventory runs dry, the bot halts arbitrage. To keep it running, we’d have to rebalance: transfer excess USDT from Binance to Arbitrum, and move surplus ETH from Arbitrum back to Binance.
But even if we could rebalance 24/7 instantly, we’d still miss substantial profits. Arbitrage gains heavily concentrate during periods of high market volatility, as spreads widen. During these moments, a $500k inventory could vanish in seconds. Furthermore, withdrawal times from exchanges stretch beyond the already-slow ~5 minutes during high volatility. This meant our bots couldn’t arbitrage the most profitable windows.
Our first idea was to build an automation system named Hydra, designed to continuously shuffle capital using cross-chain bridges and CEX withdrawal APIs. We hoped it would detect surplus assets at certain venues and replenish shortages elsewhere. But it failed spectacularly. Cross-chain bridges were highly unreliable. Assets were lost or vanished, forcing us to manually track which bridge failed and when. Bridging often took too long anyway, causing us to miss volatile moves regardless.
Enter Hydra v2. Instead of relying on bridging, we built our own on-chain spot margin system by integrating Aave (and similar protocols where available). Aave is a lending protocol natively deployed across most major chains.
The concept was straightforward. In the previous example, if we buy ETH on Arbitrum, we end up holding too much ETH while USDT depletes. The system identifies the surplus ETH, lends a portion out on Aave, and borrows USDT against that collateral.

Though simple to implement, Hydra v2 had one of the biggest impacts on our PnL at the time.
Looking back, we couldn’t possibly have devised all these techniques upfront, despite none being rocket science. Had the initial prototype crashed and burned, we probably wouldn’t have kept innovating. But once we had a live strategy, we iterated relentlessly. We monitored fill rates and profitability, leveraging rapid market feedback to pinpoint areas for improvement or optimization. This feedback loop accelerated our growth.
My takeaway is that it’s nearly impossible to crack a game when facing a mountain of opaque barriers at the start. Once you’re among the top players (cutting-edge) and have mastered the fundamentals, staying ahead becomes much easier. Especially in markets, picking the right table/game is just as crucial as execution.
The Lifecycle of Alpha
CEX-DEX arbitrage served as our primary profit engine for a long time. During peak periods of high market volatility, it generated six-figure+ daily returns. We engineered our infrastructure to be sufficiently generic, enabling us to onboard new chains within an hour. This frequently positioned us as the earliest arbitrageurs on newly launched chains, harvesting massive spreads at their most inefficient stages.
But like almost all alphas, it decays over time. Particularly in arbitrage, competition compresses spreads until profit margins shrink drastically. Imagine two firms operating with identical Gas optimization and CEX fee tiers, as in our earlier PGA example. Since trade costs are equal, they inevitably drive PGA bids up to their minimum viable profit threshold. By the end of 2025, CEX-DEX arbitrage was no longer the wildly high-yield strategy it once was.
Fortunately, steady, high-yield strategies were running reliably in the background. Over time, we were able to reallocate our time and capital into other trading strategies with superior profit potential.
I could write endlessly about the other wild ideas we conceived in subsequent years, but that’s a separate story. CEX-DEX arbitrage holds a special place in my heart because it gave us a legitimate starting point and an anchor to establish our footing.














