Operator-trained AI for crypto rails

An expert you cannot hire. On call, for on-ramps, stablecoin and remittance.

An AI trained on the work of operators who have actually run this. Not a consultancy, not a hire, not a search — it decides the way they would, and hands you one of them when only a person will do.

Whale merchant pricingReturning-user pre-selectionPerceived-value pricingFX and currency captureSpeed pricingCorridor mixShare of checkoutTier gating at checkoutWhale merchant pricingReturning-user pre-selectionPerceived-value pricingFX and currency captureSpeed pricingCorridor mixShare of checkoutTier gating at checkout
Wallet listing termsMessenger-scale merchant dealsRecovery flowsQuote segmentationCheckout step orderVolume-based economicsGas selectorNever lose a transactionWallet listing termsMessenger-scale merchant dealsRecovery flowsQuote segmentationCheckout step orderVolume-based economicsGas selectorNever lose a transaction
It has priced this corridor before.

Fifteen years on the record: every optimisation he shipped, why it was tried, what it cost, what moved.

Team spotlight

Our operators are highly experienced.

Omar Ben Hachme is a senior product and growth operator of fifteen years, eight of them in payments. He took a top-three European on-ramp from roughly $180M to about $1B annualized and built the value-based pricing this category now argues about.

The answers come from the chat — an AI trained and fine-tuned on our operators’ own practice, so it argues the way they would, at any hour and as often as you like. It rarely needs them: escalation to the operator is the exception, not the service.

Omar Ben Hachme
Trained on the practice of
Omar Ben Hachme
On-ramps
Track record

What the practice was built out of.

Every desk is built from operators who have run it, their written practice, and an AI trained on that. This is the work those practices came out of — held, not observed.

Omar Ben Hachme · Mercuryo
A top-three European on-ramp taken from roughly $180M to about $1B annualized
Rasmus Fahlander · Klarna
Led the checkout carrying $20B+ a year — more than one in five Nordic e-commerce checkouts
Daniel Sneijers · Uber
Decides which payment methods Uber's riders are offered, and in what order
Rasmus Fahlander · Kustom
Klarna Checkout carved out into a standalone company, now serving 20,000+ merchants
Daniel Sneijers · Adyen
Knows how acceptance rates are won, from inside one of the largest acquirers in the world
Omar Ben Hachme · Mercuryo
Signed the wallets and messaging apps that put a ramp in front of millions of users
Working with it

You work with it the way you would work with a real expert. Bring it the thing you are stuck on.

It is one head, not a menu. Flows, pricing, merchants, and which part of your integration to fix first.

01

Walk it through your flow

Give it your onboarding, your checkout, a screen recording. It reads what happens in the flow, against your own numbers.

02

Take the pricing apart with it

Pricing optimisation, share of checkout, recovery, FX capture, speed. It argues what a fee can carry, and holds its position.

03

Rehearse the merchant pitch

Before the call: how the deal gets framed, what the counterparty will accept, and where you are about to overreach.

04

Hand it back to him

Where the practice stops, it says so and puts you in front of the operator.

Against a general model

A general model has read everything. It has done none of it.

Same interface, entirely different thing underneath.

Trained onThe public internetOne operator's eight years in payments, on the record
Knows your businessWhat you paste inA file built from real calls with you
When it needs a numberInvents a plausible oneYours, the practice's, or it asks
Where it stopsIt doesn'tSays so, and hands you the operator
Quantitative workEstimated in proseComputed by a deterministic model
Kept currentFrozen at its training dateHe is still in this market. What he learns lands here

Put the same situation to both, side by side.

Run a live decision through a frontier model and through this, in the same hour. We would rather you ran that than took our word.

Where it is going
100+
chapters of the practice
50+
playbooks, ready to run
Per desk
specialised models, one per operator
See it work

What it knows that a model does not.

One workstream, end to end — the mechanisms, the ratios that survived negotiation, the exclusions that stop a good idea costing you money, and the one call it will not make for you.

Ten minutes, one real workstream.

In production
Get started

Sign up with one live decision.

Bring something you are actually deciding this week. You will see how it argues before anything is agreed.

  • Thirty NDA'd minutes with the operator each month.
  • A quarterly commitment, so nothing has to be decided fast.
  • Nothing is billed until the first recommended fix has shipped.

A call, not a checkout.

Thirty minutes, nothing to sign, and you leave with the operator’s read on it whether or not you take a desk.

Sign up