How SyntheticSwap's AI-powered swap engine works: quote, route, confirm, pay out

What the engine actually does in the seconds between your amount and your estimate: collecting quotes, normalising them to an all-in output, choosing a route by pair and size, then watching the blockchain for your deposit and paying out. Plus what it deliberately does not do — profile you.

The problem the engine solves

An instant swap looks like one number — "you will receive X" — but producing that number well is a routing problem. For any pair there are several places the trade could be done, each with its own price, its own depth, its own minimum, and its own cost to pay you out. The best one changes by the minute and by the amount. A service that always asks the same source gets a good price on deep pairs and a poor one on thin pairs; a service that asks all of them and picks well gets a good price on both.

SyntheticSwap's engine does the second thing, automatically, for every quote. This page describes what it does at each stage, in plain terms, with no claims beyond what the product actually does.

Stage 1 — Quote collection

When you enter an amount on a pair page such as XMR → USDT (TRC-20), the engine requests quotes from the liquidity sources that can serve that pair at that size. A quote is more than a rate: it carries the source's minimum and maximum for the pair, the payout network fee it will charge, and whether the pair is currently enabled at all (some sources pause thin pairs during volatility).

Stage 2 — Normalisation

Quotes are not directly comparable. One source may show a better rate but charge a higher payout fee; another may quote a rate that only holds up to a certain amount. The engine reduces each quote to a single figure — the amount you would receive, net of everything — and evaluates that. This is the number the form shows you, and it is the reason the estimate is a real output rather than a rate to which fees are later applied.

Stage 3 — Route selection

The engine picks the route with the best net output for your amount. Two things make this more than "sort by rate":

Depth. The best source for $200 is not always the best for $20,000. Larger amounts consume liquidity, and the engine evaluates the quote at the requested size, not at a nominal one.
Pair-specific behaviour. Sources have strengths: one prices majors tightly, another has real Monero or Zcash liquidity, another pays out on TRON cheaply. The routing adapts per pair rather than applying one ranking to all.

This selection is what we call AI-powered routing: the choice is made per request from live data, and it continually adjusts to which source is delivering the best net result for that pair and size. It is not a marketing layer over a single provider.

Stage 4 — Order and deposit watching

Once you click Swap now, the engine creates the order: a one-time deposit address, your receiving address, an optional refund address, and the chosen route. From then on it watches the deposit chain for your transaction and counts confirmations — 2 for Bitcoin, 15 for Ethereum, 1 for USDT on TRON, 5 for Monero. The status page reflects each change.

Stage 5 — Execution and payout

At the required confirmation count the floating rate is fixed for the amount that arrived, the trade is executed with the selected source, and the payout is sent to your address. If the amount differs from the order (a wallet deducted its fee), the engine recalculates rather than failing; if it is below the pair's minimum, it is returned to the refund address minus the network fee.

What the engine does not do

Being precise about the limits is part of describing a system honestly:

It does not profile users. There is no account, so there is nothing to learn a "preference" from, and no history to personalise. Every quote is computed from the pair and amount alone.
It does not predict the market. The rate is floating and fixed at confirmation. The engine optimises the route at quote time; it does not try to time your deposit.
It does not execute on-chain through smart contracts. Cross-chain swaps involving Bitcoin, Monero or TRON cannot be done inside one contract. The trade is executed with a liquidity source off-chain, and only the two transfers — your deposit, your payout — are on-chain.
It does not override the blockchain. If your Bitcoin deposit takes three hours because of a low fee, the engine waits three hours. Confirmation counts are not negotiable.

Seeing the engine's result before you commit

Each pair page shows the current all-in output for a sample amount against the CoinGecko mid-market rate for the same pair, as a single percentage. That number is the routing's result made visible: on a deep pair like BTC → USDT (TRC-20) it is small; on a thin pair it is larger, and the routing is what keeps it as small as it can be. Comparing that percentage across pairs tells you more about real cost than any fee table.

Frequently asked questions

Does the engine split an order across sources? It selects the best route for the full amount. If the amount exceeds every source's maximum, the pair page shows a lower maximum, and the practical answer is two swaps.

Why is the estimate refreshed every minute? Because quotes expire. A source's price and depth change continuously; a minute is short enough that the estimate is realistic and long enough not to flicker.

Can I choose the source myself? No — the point of the routing is that you do not have to. The output amount is the only thing that matters to you, and it is shown before you send.

Does "AI-powered" mean my data is used to train something? No. There is no user data. The routing learns from quotes and execution outcomes across the system, not from anything about you.

Where can I see it in action? Any core pair — the pairs page links to all of them with live limits.