Erik Voorhees wanted a cross-chain exchange for THORChain that ran straight off a KeepKey hardware wallet. Ten weeks, three milestones, a deployed prototype dated December 18, 2020.
The contract was signed October 11, 2020, and the code handed over in April 2021. The client was Erik personally. The rest of this page hangs off those two documents.
Years after ShapeShift launched out of our office, the same founder called again, this time personally rather than through the company. The brief was a hardware-based cross-chain DEX for THORChain, run from a KeepKey wallet. In the brand questionnaire he filled out himself, he called it niche and community owned: his own prototype, not an official ShapeShift product.
“the only game in town with a good UX”
Erik Voorhees · client, describing what Valhalla had to be
The rule from the ShapeShift years never moved: we never held the keys. Valhalla is that rule drawn as a product. The keys stay on the device in your hand and the exchange comes to them, which is also the reason this build was hard.
Four weeks of discovery before any screens.
Business goals, user interviews, user goals, competitive research, a discovery workshop. Then the feature list, the information architecture, the user journey, flows and wireframes, and wireframe iteration against user feedback. That scope was written in October 2020, before the Innovation Theory name existed, and it’s close to how we run discovery now.
The brand was milestone two, not milestone one.
Valhalla got its name, its look, and its interface after the technical scope was settled, not before. On a build whose whole premise was a hardware handshake, the order matters: find out what the device will actually do, then design the thing a person touches.
Built to closed-beta ready.
We did the brand, the discovery, the UI, and built the web frontend and the Electron desktop app. The final code archives handed over are dated April 20, 2021: frontend, Electron, and bitcoin.
The device was the whole risk.
Prototype requirements and KeepKey technical exploration notes came first. Wiring a hardware wallet into cross-chain swaps isn’t a layout problem, and pretending otherwise is how prototypes get expensive. We went and found out early. We still didn’t fully win it.
The contract set the shape and the order: technical scope first, brand and design second, deployed prototype third, dated December 18, 2020. Nine weekly status updates went out between October 2020 and March 2021, and they’re the better record of what actually happened.
Contracted October 11, 2020 · prototype milestone December 18, 2020 · code handed off April 2021
A prototype’s job is to tell you which assumptions survived. Here’s the state ours were in on the day the code changed hands.
It worked.
The piece most likely to get blamed in a post-mortem behaved, which narrowed the problem to one thing instead of a maybe on every layer of the stack.
Still fighting us at handoff.
An exchange that runs off the device in your hand only exists if the device shakes hands cleanly. Ours didn’t, all the way through April 2021. That’s why it’s in the headline.
Ready is not launched.
The web frontend and the Electron app were built to closed-beta ready and a beta was on the calendar for April 12. Whether it ran, the archive doesn’t say.
We handed off as the chain itself went live.
The code changed hands in April 2021, right as THORChain went live, so the target kept moving while we built against it.
It was a prototype, and we’re calling it that. The point of ten weeks is to find out what the hardware will really do, and we found out, including the part we hadn’t beaten yet.
Ten weeks against a hardware target nobody had hit, and at the end of it a deployed prototype, a brand, a web app and a desktop client, with WalletConnect working end to end. The KeepKey path was still being fought when we handed over, and that is the answer the ten weeks were bought to get: what the device will really do, found out early and in the open rather than late and expensively.
The code was handed over complete, and the chain it was built for went live the same season.
Everything above is what a prototype is supposed to produce — a working shape, a known unknown, and a client who can decide what to spend next with real information instead of a guess.
Valhalla was executed on October 11, 2020 as Ideas By Nature LLC. In November and December of that same quarter, the Innovation Theory logo was being drawn. The two things sit in the same folders, weeks apart. Most shops would leave a mid-engagement name change out of the story. We kept it in because it pins the transition to the exact quarter.
For over ten years, they’ve been part of the ShapeShift journey. They’ve helped shape the brand, the product, and the tech behind it. You won’t find better.
Erik Voorhees · Founder, ShapeShift
The new frontier of crypto in the OG days is very similar to the constant change and upheaval with AI now. UIs are almost temporary as things change so fast. That’s the level of dev you need now with AI.
Phil · Innovation Theory
Repeat work is the only metric on the record we actually optimize for. Valhalla is what that looks like from the inside: a founder who’d watched us build for years handed us his hardest unproven idea, personally, and got a straight answer in ten weeks.
The evidence base is a contract signed October 11, 2020, a client brand questionnaire, prototype requirements and KeepKey exploration notes, a discovery scope, a roadmap, nine weekly updates, a demo video and three code archives dated April 20, 2021. That is all of it, and it stops at handoff.
What is not here. No launch metrics, no user numbers, no revenue and no outcome claim. A closed beta was scheduled for April 12, 2021; whether it ran is not in the archive.
On the December date. The contract sets the third milestone as a deployed prototype dated December 18, 2020, and a demo video from that month survives. The video does not prove the deployment date — the date is what the contract specifies, not something the folder independently confirms.
Everything above this line is in a file somebody can open. Nothing below it is. This is how we would run the same engagement today, with AI in the work — written as an opinion, because that is what it is.
Ten weeks to a closed-beta prototype across brand, web and desktop is the shape of engagement AI compresses hardest. The exchange screens, the wallet flows, the desktop client scaffolding and the four weeks of discovery documentation would all move faster today, and WalletConnect — the half that worked — is well-trodden enough that a model would get you there in days.
The device would still have said no.
The KeepKey integration was the whole risk and it was still fighting us at handoff, and that is not a code-generation problem. Hardware does not care how fluent the thing writing the firmware call is; it either signs the transaction or it does not. The point of a ten-week prototype is to find out what the hardware will really do, and finding out includes the part you have not beaten yet. AI makes the answer arrive sooner. It does not make the answer yes.
The question was whether a hardware wallet could carry an exchange. Ten weeks later there was a prototype and a real answer, including the part still fighting us. An assessment finds out what holds before you fund the whole thing. Two days, no charge, answered in writing.
Let’s find outNot ready? Take the Guide: the six rules with checklists you can run yourself. Free. No email required.