Ten steps became one screen.
Then it became a decade.

Nobody signs
a decade.

You get re-hired, one surface at a time, by people who watched the last one hold. He called again years later for the Valhalla project on THORChain.

One deliverable from that decade is not a screen. The 2018 brand manual is a full identity book — primary and secondary lockups, a single-colour exception for embroidery and etching, three blues, Nexa as the brand face with a custom i and f, and a rule for the fox's tail: blue shadow on white, all white on dark. A rule that specific only gets written by whoever has to answer for the artwork.

We never held the keys.

That held for the whole decade: the company never touched customer key custody. Key ownership would all have fallen on us, so we stayed a service company, not a product owner. Knowing what not to own is part of the job.

The problem was never crypto.
It was the ritual.

Legacy exchanges took about ten steps to complete a transaction: an account you did not want, a screen of numbers you could not read, and sometimes an order that sat unfilled for hours. The research said people were tired of all of it.

More than a million transactions a day, at peak.

The flow was the visible half.
Coin support was the other one.

Everything above this line is design. It is not the whole engagement, and the reason it looked like the whole engagement is that design files survive and code does not. Our own published account of this work names the rest, and it is worth quoting rather than paraphrasing.

Widening what an exchange can hold is plumbing, and the plumbing is the product.

Every asset a user can swap is an integration somebody had to write, test and keep working while the asset’s own network changed underneath it. Nobody markets that. It is the difference between an exchange that supports six things and one that supports the thing you actually own, and it is the same job we were doing on the other side of the building — the KeepKey indexing services that widened what the wallet could hold. One family of work, two products.

The published scope also lists mobile and dApp development, a token launch, and business development alongside the product and design work. The FOX membership build further down this page is the design half of that token; the launch itself sat in the same engagement.

This section is sourced from our own published portfolio of the engagement.

Twenty screens, delivered
2017-09-19.

The Partner Portal moved the same discipline from the consumer flow to the business one. Two partner types, Referral Link and Partner API, sorted by a three question partnership questionnaire that routes a prospect to the right tier. Then account creation with mandatory 2FA, organization verification with document upload, an approval hold, API key generation, a dashboard with per asset volume and a live transaction ticker, filterable reports, user roles with an audit log, and a payout flow.

We hit the date.
Twenty-four days, brief to live.

2018 was the year the FOX token became a product surface. The mobile membership integration was briefed 2018-09-07 with a hard October 1 deadline: OAuth through the membership portal, 2FA before entry, shifting blocked until verification is complete, FOX Back status on every shift, and the old app switched off. Copy was specified down to the line, including “Congratulations! You’ll be earning FOX back on this shift.” The apps went live on time. Then we wrote down what it cost.

We wrote the fix into the process.

The retro produced named stage gates that still run here: brief, scoping, charter, UX and UI approval, tech requirements meeting, testing and QA, handoff. Plus one rule with teeth, quoted from the document: “We need to ensure that any stakeholder who might dictate changes sign off on the UX prior to moving forward with the project/effort.” The FOX membership dashboard that followed, levels 2 through 5, unlocked and pending FOX balances accruing over a 30 day period, ran to that process from 2018-11-02 to 2018-11-29.

The same six sentences,
on two very different screens.

KeepKey work ran alongside ShapeShift: a 2017 brand system locking the logo lockups, a four color palette anchored on Pantone 382 C, Raleway, and hard minimum sizes so the mark never gets crushed. Then a white label spec that reads like manufacturing, because it is one: the device back at 38.0mm by 93.5mm, 79.5mm of width for a client logo, and the KeepKey mark at 8.5mm by 10.51mm sitting 3.5mm off the bottom right corner. That is the difference between a brand guideline and a brand that survives a factory.

If you lose your key you’ve lost everything.

Phil’s own line from the writeup, and the reason none of this was decoration: “If you lose your key you’ve lost everything, its very important to understand that. Building trust.” Simplicity was the security feature.

Ten years is a long time
to keep a vendor.

Years after this, Erik called again for the Valhalla project on THORChain.

One client.
Four products.

The exchange is the story on this page. ShapeShift also owned a hardware wallet, a portfolio market and a market-data platform — and we worked on all three.

The card sort ran in our own office in April 2018, on people already using Prism, and asked one question: where do you expect each of these to live? KeepKey has its own page.

What the files hold, and what they do not.

AI would not have
refused to hold the keys.

Ask a model to build a crypto exchange and it hands you custody without blinking, because nothing in the prompt priced the liability. We refused in 2013, it cost us product revenue for a decade, and it is the reason there is a decade here to write about. That refusal is what this page is evidence for, and it is the same call we are hired to make now, with AI in the builder's seat instead of a contractor.

The rest of it does go faster now — twenty versions of the ten-steps-into-one-screen flow in an afternoon rather than three weeks. That is only worth having if somebody in the room already knows which one to keep, and which one to refuse to build at all.