How does Jupiter find the best swap across all of Solana's liquidity? @melvinzzy from the Metis team breaks down the routing infrastructure behind every swap, how it evolved, why it's harder than it looks and how unexpected ideas end up shaping a swap engine. The layer most people never think about, explained by the people building it. Just Build With Jupiter. 💻

This is the engineering story behind how we built Limit Order v2 from the ground up. For a while, people wanted two things: more order types, and to think in dollars instead of doing token-for-token math. But every time we looked at building them, we hit the same wall. The v1 onchain program and the ratio model couldn't express what people wanted. So we bit the bullet and rebuilt it with a different approach: an order is a price, not a barter ratio. It surfaced problems the v1 model had been hiding: pricing any token in USD, monitoring every order at scale, and executing safely on a ledger that never forgets. @arzine_ documented the whole thing in 3 parts 👇 Part 1 - The any-any problem: Part 2 - Correctness under failure: Part 3 - From barter to money:

Money isn’t the thing you transact in. It’s the thing you think in. We rebuilt Limit Orders around that: orders are now prices, not ratios. We know the switch broke some muscle memory, and many of you have asked why. @arzine_ lays out the full reasoning:

Complaining created the best QC team ever. @YodaHeld wasn’t hired for QC. He just complained about bugs too accurately, and Jupiter said: “congratulations, you complain so well, you’re the team now.” No vote. No interview. 😭 Now he joins @MessiasMorten & @jup_burst as our in-house Karens. Just Build With Jupiter.

How @jup_mobile devs are using AI to ship across iOS and Android faster. Write the logic > ask Claude to read it > iOS / Android implementation gets easier. Maybe we don't need two teams anymore 👀 (Relax. We're joking. Mostly.) But Jupiter's mobile team is using AI to make work across platforms more efficient. Jupiter devs are cracked. Just Build With Jupiter. 💻

Limit Order v2, Part 2. Moving order state off-chain bought us privacy, but it meant rebuilding every correctness guarantee the blockchain used to hand us for free. @arzine_ breaks down how we did it:

Off-chain routing has a ceiling. No matter how fast the algo or how fresh the quote, pool states shift between route computation and transaction landing. The route plan you were quoted becomes stale by the time it executes. JIT Swap lets our program re-decide on-chain, at execution. @gn_dnomsed breaks it all down:

Introducing: Jupiter Forecast for MMs, Solana’s first fully native prediction market. For users, Forecast is built into @jup_predict, but provides an additional liquidity model that can deliver better prices and better execution. Instead of trading against a single pool of liquidity, users will be able to buy from multiple competing market makers. Any Prop AMM can start quoting into markets they are interested in, and users will automatically be able to buy from whichever market maker offers the best price. To be clear - we’ll continue working very closely with Polymarket and supporting their markets. Forecast simply offers an additional type of market liquidity that users can take advantage of. But this isn’t just better for users, it’s better for Solana as a whole. With Forecast, markets will each have their own native tokens. Integration for partners becomes seamless and composability increases across the entire ecosystem. PropAMMs changed spot trading on @solana. Now, we’re going to see if they can have the same impact on Prediction Markets. We’ll be starting with 15 minute crypto markets and expanding from there - if you’re a market maker who wants to get involved, reach out!

Back when Jupiter Limit Orders v1 launched, it did something most platforms still won't let you do: trade any → any token, not a fixed set of pairs. With that came a hard pricing problem. Every new token adds a row and a column to the matrix of pairs, so the number you have to price grows quadratically. In Part 1 of our LOv2 series, @arzine_ breaks down the challenge and how we're handling it in v2:

We’re partnering with @JupiterExchange to bring smarter swaps to @solana! @demetherdefi leverages Jupiter in the backend to power optimized swaps and routing, providing users with efficient trading and liquidity access across Solana. Try the app at and let us know what you think!

Route A quotes 100 tokens, executes 95 Route B quotes 150, executes 50 Metis v8 picks Route A. It routes on what each path will actually execute at, not what it quotes. The best quote isn't the best swap. Read the full article by @melvinzzy here:

What you see is NOT what you get when you swap onchain. The app, wallet simulation, and actual execution all diverge. We designed Metis v8 to minimize drift through faster quoting, slippage-aware market selection, onchain split optimization, and next-slot transaction landing.

Metis V8: Solving The QE Drift Metis, the top router in DeFi with more than $2T+ in lifetime volume, has always been focused on the hardest problems in routing. And these days, it is not just about finding the best quote, it is about making sure what lands is as close to the price you were quoted as possible, aka The Quotation-Execution (QE) Drift. V8 is focused on solving that holistically from multiple dimensions in a systemic way: 1. Incentives: Slippage Penalties 2. Design: JIT Onchain Finalization 3. Infra: Sub-2 Slot Latency 4. Option: Rapid Quotation Mode Let us explain:

for as long as i can remember, vrfd is that one feature i've only heard good things about also available as an API for verified tokens, token info + news, used by some of the biggest exchanges and teams on solana