Alex Fine

After 7 years quietly building, it was fun to share some of my mistakes & learnings @MIT this week.

@fun's last financing was done at a $425M valuation, and we’ve now grown 9x since our $72M Series A. Our 30d volume just crossed >$100B annualized.

A few of my mistakes along the journey 🧵

My unfiltered takes dropping soon on @stabledash

only the paranoid survive

equally said – the paranoid are right

Very grateful to support the best builders

Crypto vendors hide fees in creative ways. Let's break down how these flows actually work.

First, to be clear, Polymarket has:

1. The lowest fees.
2. The fastest settlement.
3. The highest limits.

Here's how.

First, how do most crypto orchestrators work today?

A user attempts a trade. The orchestrator scans the market for the "best rate," picks a route on the user's behalf, and executes.

Sounds great. So what's the problem?

Issue 1) Markets are dynamic, and pricing can be poor

If you rely on orchestration alone, you will give users bad pricing at some point - markets aren't always efficient.

USDT to USDC trades at a floating market price, not a fixed one. As of writing, depositing USDT on Hyperliquid carries an 8.7bp markup & Relay costs 15.5bp. The same transaction on Polymarket costs 5.5bp.

The bigger the transaction, the worse pricing gets. A $10M USDT deposit through an orchestrator means brutal slippage, slow settlement, or no quote at all - funds stuck.

And it is worst exactly when it matters most. Markets dislocate hardest during high volatility, so users get their worst pricing during a big airdrop, election, IPO, or sports game.

To give users a good experience, someone has to eat the cost when markets aren't efficient.

Apps will say "we have zero fees" - but when markets dislocate, retail gets their face ripped off.

Btw – it is not possible to offer 1:1 USDT to USDC for deposits & withdrawals in any size without be arbitraged.

Issue 2) Most bridges can't guarantee SLAs

Say a user is depositing $10M into Polymarket. Try any leading router rn and you'll struggle to find a quote for a $10M USDT transaction. Should those transactions just get stuck?

What happens during the Super Bowl, when everyone bridges at once? We've seen liquidity on major bridges drop below $50k. Should users just not be able to deposit?

What happens when rebalancing goes down on a major chain? It happened to BNB recently - funds couldn't get off the chain. Should every BNB user eat 1h+ delays?

Every cross-application bridge (i.e., ALL bridges today) is exposed to these failures.

To give users a good experience, you need an app-specific bridge with real capital behind it - in Polymarket's case, >$10M dedicated to instant transactions for its users alone.

Apps will say "we can bridge from any chain" - then fall over when markets dislocate.

How does Fun + Polymarket solve this?

Fun + Polymarket routing combines three capabilities:

1. An OTC desk.
2. A world-class aggregator.
3. An app-specific bridge.

The result:

1. Best pricing from Relay, Across, LayerZero, CCTP, and others - at all times. Thanks to our relationships with Polymarket, Aave, Lighter & others Fun is one of the largest transaction originators in crypto. This size enables highly preferential pricing.
2. Fun steps in and takes the loss if routers can't deliver SLA-level pricing.
3. Fun holds $10M to fill the transaction if routers can't deliver the size.

This thread doesn't even mentioned the other infra to make crypto tx's great: transaction retries, gas retries, slippage monitoring, broad token & chain support, fast pickup times, RPC fallbacks, streaming fallbacks, and more.

@Polymarket and @Fun have iterated on this tech for two years. I'm proud of the work, and grateful for the partners we've built alongside.

I admire what the Kalshi team has built - this isn't a hit piece, and John is well-intentioned. But it is simply inaccurate to imply Polymarket has inferior crypto rails.

Polymarket supports larger orders, faster, with tighter settlement and greater uptime than any other crypto platform today.