The roadmap

Running in production against chain 4663 today, MIT licensed, with every figure read from chain state at request time.

• Route engine across six venuesUniswap V2 direct, V2 via WETH, all four V3 fee tiers, and V4.
• Price impact on every quoteMeasured against a probe trade 1000x smaller, warning above 5%.
• Launch discoveryCurves resolved from the launch mint, so it does not depend on one launchpad's ABI.
• Clone detection and safe modeDuplicate name and symbol pairs flagged, explicit names blurred by default.
• Portfolio without a token listHoldings found from your wallet's own transfer history.
• Cashback accrual0.12% of routed volume, computed from your own swaps. Not claimable yet.

No new capability. This is the work that makes what already exists trustworthy enough to send real traffic at.

• Dedicated RPC endpointDone. A dedicated endpoint serves the app through a proxy on this domain, so its key is never shipped to the browser, and the public node sits behind it if that one stops answering.
• Verify the sell path end to endBuying is proven against live pool math. Selling needs a Permit2 approval and wants a real trade behind it.
• Resolve curve ETH reportingThis launchpad does not hold sale proceeds in the curve, so the field reads zero. Find where they sit or drop the metric.
• Widen historical coveragePartly done. V4 pool discovery used a 400k block window, which is eleven hours on a chain with tenth-of-a-second blocks, so tokens older than that had no route at all. It now walks backwards until it finds them. Deposit scans still start from the pool's own deploy block.

Today no fee is taken at all: swaps route straight to the venue and output goes directly to you. This phase changes that, and it is what funds everything below it.

• A share of the routing surplus, not of your tradeThe surplus is what we found above the venue you would have used anyway. Capped so a fill can never land below that baseline.
• Public fee accountingHalf done. Every figure is on-chain and readable now, fees taken, surplus found and volume per wallet, and the Routed event publishes the baseline alongside the output so the fee can be checked against it. The page that presents all of that is not built yet.

A shielded pool on chain 4663. Deposit any amount, spend any part of it to an address that was never linked to you, and the remainder comes back as a hidden note only you can open. Non-custodial: there is no operator, no admin key and no server that knows your balance.

• Hidden amountsEach note carries its own value, so one pool holds every size. Fixed denominations would split users into a smaller crowd per amount, which is the opposite of what an anonymity set needs.
• Partial spendsSpend part of a note and the change returns encrypted to you. Only what you withdraw is ever published.
• Balances rebuilt in the browserNotes travel encrypted beside their commitment and your wallet trial-decrypts the stream locally. Asking a server which notes are yours would hand it your whole position.
• Multi-party trusted setupNot done. Phase one is currently a locally generated file, which means whoever generated it could forge a proof. Real funds should wait for this.

The pool itself trades through the REFRACT router, so what lands on chain is that the pool swapped, never that your wallet did. The routing engine and the shielded pool are the two halves of this product and this is where they meet.

• Pool-executed swapsA spend sends value to the router instead of to a recipient, and the proceeds go to an address with no history. The proof is the same one a withdrawal uses, so this needs no new circuit and no new ceremony.
• Router fixed at deployThe pool may call exactly one address. An arbitrary call target would let anyone drain it.
• Shielded proceedsNot in the first version. Notes are denominated in ETH, so a token bought through the pool leaves it. Holding token notes needs the asset in the commitment and a new circuit.
• A real swap with real moneyProven against the live Universal Router on a fork of 4663, and not yet run on the chain itself. The V4 route executed there; the V2 and V3 encodings have still never been used in anger.

The accrual maths already exists and is already auditable. What is missing is the contract that holds collected fees and lets a wallet claim its share.

• Claiming, with nobody to askThe first design paid from a Merkle root published by a trusted account, and its own test proved that whoever published could name themselves and take the pot. It was deleted. Entitlement is now computed by the contract from volume it recorded itself, and a claim takes no account argument, so it can only ever pay its own caller.
• Honour everything accrued so farNot done, and worth being plain about. Cashback accrues only on trades routed through the fee router, which went live today, so volume from before then earns nothing. Paying it would mean funding historical accrual out of future fees, which is a real cost and a decision rather than an oversight.
• The Rewards page stops needing a disclaimer

Only worth doing once fees flow. A boosted rate funded by nothing is just an emission, and everyone can tell the difference.

• Boosted rate for holders0.30% of routed volume instead of 0.12%. The token pays you more for something you already do.
• Buyback from collected feesEvery transaction published, so the claim stays checkable.

Each of these is only worth building on top of a working loop, so they are deliberately last.

• More venues as they deployThe engine's value scales with how many places it can look.
• Limit orders
• Multi-hop routingBetter prices on thin pairs, which is where the surplus is largest.
• Launch and clone alertsBoth already run server-side; this is a feed on top.
• Other chainsLast on purpose. Expanding early only spreads thin liquidity across more surfaces.