Trading is painful

Trading ought to be a great opportunity: Put capital where it's useful and earn from the risk.

But trading doesn't feel this glorious. Instead, you feel like someone is always out to get you: You don't settle for what you were quoted, fees are higher than advertised and, when it counts, the trade reverts.

Apps ask for blind trust without any skin in the game. It's clearly not a good idea, but you have no alternative.

Not even tradfi traders open this many tabs. We don't trust the infra that handles our trades.

Aggregators overquote 0.49% to 13.18% of the time and revert 3.20% to 23.22% of the time (0x). Aggregators now quote nicer on dashboards, but still overquote their clients in private (by changing their quotes depending on API Key and IP address).

Where does the pain come from?

Intermediaries limit trading

Trading, outside of defi, is quite standardized: Someone hosts an order book, someone quotes prices, and you take offers from the book when you want to buy or sell.

The meme is that trading through an exchange, with a maker, at the price the maker decides, is the most efficient and only way to trade.

It is not. Traders can just trade with each other. No exchange, no maker. But you need to trust each other, and in a global market with strangers that doesn't scale.

The need for trust at scale made middlemen useful.

Middlemen have more information and can extract from traders who know less.

NASDAQ, CME and even much of defi infrastructure survive on fees that traders pay to makers and HFT firms, which in turn pay exchanges for the right to extract from the trader.

Exchanges are very aware of this. They compete on incremental improvements (deeper books, better data, some variation on order types and faster execution), but not on the structure of the market.

One exchange came close. The Chicago Stock Exchange's 2017 asymmetric speed bump tried to directly address latency arbitrage: By delaying liquidity-taking orders by 350 microseconds, it gave liquidity providers time to update stale quotes. Incumbents vehemently opposed the proposal. Then ICE, the parent of NYSE, acquired CHX on July 18, 2018, and CHX withdrew the proposal seven days later. The exchange became NYSE Chicago, but the alternative market design disappeared.

The need for scalable trust and an oligopoly protecting rent define traditional market structure.

For traders, the tightly bounded competition in traditional markets is performance theater to stoke the meme of efficient markets.

Better designs are possible

Designs that break out of tradfi orthodoxy create significant surplus:

  • AMMs: Anyone can be a maker for spot.
  • Hyperliquid: Anyone can be a maker for perps (HLP).
  • Fomo: Add a social signal to a trading terminal.
  • propAMMs: Make RFQs path-dependent, permissionless and last look.

These examples are only inches from traditional market design and do not yet fully embrace the leverage of Ethereum.

The path to more useful markets

Trading should allocate resources where they are most useful, at minimal cost. This clarifies how to build better markets:

1. Make more things tradable: Traders can only allocate to what is on offer. Better markets make more things tradable. (Examples: Trade outcomes on Polymarket, pre-IPO perps on Trade.xyz and turn RWAs into legos for lending, leverage and hedging).

2. Remove friction: Replace infrastructure choke points that limit developers and traders and extract from them. For example, remove the risk makers charge by letting holders provide liquidity (AMMs), make fed rates accessible (by matching institutions with retail in defi lending markets), and make quoters compete (open auctions).

3. Improve decisions: Give traders relevant signals they can trust and help them execute competitively with less effort. Better decisions direct more capital to where it's useful.

Developers are very unhappy not having control over their trades

Trusted offchain middlemen make defi trading:

  • unreliable (APIs change, rate limit or go offline),
  • expensive (middleman cost),
  • adversarial (overquoting and kickbacks), and
  • hard to build on (closed source).

Developers (ask any) are very unhappy with the current options.

Founder of a team you know.

The Open Trading Stack

The open trading stack brings liquidity directly to traders, making trading as reliable, accessible and composable as Ethereum itself.

So far, we built three pieces.

Tycho indexes onchain liquidity: more than 180,000 pools and 95,000 tokens across more than 40 protocol deployments on Ethereum, Base, BSC, Polygon, Arbitrum, Unichain and Robinhood, including AMMs, RFQs, propAMMs, vaults and RWA mints.

Most solvers and their indexers rely on pulling all DEX state periodically from RPCs. Tycho is RPC-less; it streams only state deltas from instrumented nodes to the solver and can therefore handle all pools and protocols (even on fast chains).

Most indexers do the equivalent of "SELECT *" every few blocks. Tycho behaves like a write-ahead log, with only a single, small delta message per block that takes less than 1 ms to add to the full DEX state locally.

Six of CoW Protocol's ten active solvers use Tycho for indexing, simulation and execution. Together, Tycho-powered solvers settle about 45% of CoW volume and ~$2B per month on Ethereum.

Tycho cuts the time to build your own defi router from 18 months for a five-person engineering team to one month for a single engineer.

Fynd is a production-grade DEX aggregator built on Tycho. It has routing algorithms that optimize for different goals (speed, depth, low reverts), auto-scales over as many CPUs as you have, and lets you configure worker pools across which you can load-balance thousands of requests per second. It also comes with benchmarking and per-transaction fees you can charge, and it is simple to deploy.

Fynd is about 20 times faster and can handle 100 times as many requests per second as hosted routers. A 32-vCPU AWS instance can solve 1,220 requests per second at 38 ms median latency and 49 ms p99 latency. By comparison, the 2026 aggregator benchmark measured median response times from 529 ms to 2,270 ms.

Because Fynd is open source, you can choose which pools and tokens to route through, whether to wait longer for a better trade or to favor routes that are less likely to revert.

Since Fynd launched on April 26, 2026, teams built oracles to price their inventory, liquidity monitors to keep their vaults healthy, in-house arbitrage bots to balance their pools, competitive solvers on CoW Swap and fast in-browser quoters for better UX.

Rate limits, slow responses and no option to customize anything in the router make it impossible to build the same things on closed-source APIs.

Hosted API: Fynd now also has a hosted API. Same interface as closed routers but trustless. You can decode and verify every quote locally; overquoting and hidden fees are impossible. Within days, the teams that route the largest volume in defi started to execute with Fynd.

One team reported that Fynd wins 83% of their quotes overall, roughly 60% on Ethereum and 99% on BNB Chain.

Turbine is the first market that can settle large trades on Ethereum at a lower spread than centralized exchanges. Turbine uses Tycho and Fynd to bring onchain liquidity into a trustless batching solver. It hosts a private order book inside a TEE.

Turbine orders follow the market instead of getting stuck like limit orders. A speed bump protects your order from HFT. Turbine optimizes for low spreads and peer-to-peer trading.

In Turbine, you can trade time for better execution. Your order follows the global CEX mid-price and stays fresh for minutes, hours or days. This lets you match peer to peer at zero spread, beating Binance VIP 9 or propAMMs.

We are not done here.

Network trading

From fragmented local markets, through centralized hubs, to an open trading network.

The open stack makes trading accessible locally. Anyone can run a router, just as anyone can run a light client. Trading is aligned with the trader.

Through the open stack, you can put up liquidity for any other trader.

This removes the third-party router from trade execution. Traders can network their intents.

A network of traders has less friction, makes more things tradable and surfaces more information in the form you choose.