AMMs go mainstream

From order books to curves

Uniswap’s x*y=k unlock

In 2018, Uniswap v1 launched with a tiny grant and a simple rule: the product of reserves stays constant (x*y=k). No order book, no market makers to negotiate with. Anyone could deposit both sides of a pair, get LP tokens, and earn a slice of the 0.3% fee. Pricing was automatic: trades slid along the curve. The original Uniswap v1 post ↗ captures the simplicity.

This model prioritized simplicity over capital efficiency. Slippage was high for big trades, but onboarding was zero-friction. For the first time, liquidity provision was a public good—no gatekeepers, no listings to beg for.

Specialization: Curve and Balancer

Curve optimized for stable/stable swaps with low slippage via its “stableswap” curve, making USDC/DAI/USDT trades far cheaper than Uniswap’s constant product. Balancer generalized pools to multiple tokens and custom weights, letting projects create index-like baskets with fee revenue. Curve’s stableswap paper ↗ and Balancer’s whitepaper ↗ showed AMMs were a design space.

These designs showed that AMMs were a design space, not a monolith. Different curves for different assets—and different trade-offs for LPs and traders.

Impermanent loss as the LP tax

LPs learned about impermanent loss: if prices move, your pooled assets rebalance against you compared to just holding. Fees can offset it, but volatile pairs punish passive LPs. AMM UX rarely explained this; many newcomers learned the hard way. Dynamic fee tiers (later Uniswap v3) were a response: price risk should pay more.

“AMMs turn liquidity into a public utility.” — Paraphrasing early Uniswap community mantra

Liquidity mining and vampire bites

Yield farming ignites

Mid-2020, protocols began paying LPs with native tokens (“liquidity mining”). Compound’s COMP distribution kicked off “DeFi summer,” and AMM LP positions became yield-bearing assets. Total value locked exploded as farmers chased APRs, often ignoring smart contract and market risks. The COMP launch post ↗ lit the fuse.

Sushi’s vampire attack

SushiSwap forked Uniswap v2 and launched a token with high rewards for migrating LP tokens. It worked: billions in liquidity moved, proving that incentives could uproot even a beloved protocol. Uniswap later responded with UNI and a v3 redesign; the episode cemented “liquidity is mercenary” as a rule. The original Sushi proposal ↗ reads like a daring heist plan.

Oracle games and flash loans

AMM prices feed on-chain oracles. Attackers used flash loans to swing AMM prices, trigger liquidations elsewhere, and profit. Protocols tightened oracle sources (TWAPs, Chainlink feeds) and added safeguards (price caps, circuit breakers). AMMs inadvertently became oracle infrastructure—good and bad. Chainlink’s oracle primer ↗ became default reading for DeFi teams.

What AMMs changed for everyone

Composability and routers

Smart order routers (1inch, Matcha) split trades across multiple AMMs to reduce slippage. Contracts composed AMMs as building blocks: lending protocols accepted LP tokens as collateral; yield optimizers auto-compounded fees. “Money Legos” became literal—protocols calling other protocols by default.

Access and listings flipped

Tokens no longer needed exchange listings; a pool could be created by anyone. This democratized launches but also enabled rug pulls and spam tokens. Users learned to check contract addresses and pool depths before trusting a trade.

Capital efficiency race

AMMs paved the way for concentrated liquidity (Uniswap v3), hybrid books/AMMs (dYdX, CEX-DEX blends), and intent-based trading. The constant product was the spark; the race now is to minimize slippage and risk while keeping permissionless access.

Legacy

AMMs turned passive capital into programmable market making, opened liquidity to the crowd, and made DeFi usable without order books. They also introduced new risks—impermanent loss, oracle swings, mercenary liquidity—that pushed the space toward better risk disclosures and designs.