Four hundred prototypes and the one that worked

Placeholder copy from the design prototype — not approved. See docs/design/07-copy-and-voice.md.

The first version of the Twist was a bolt set in resin poured into a silicone ice-cube mould. It weighed too much on one side, it made a dry scraping noise, and it was the single most satisfying object any of us had held. That combination — clearly wrong, obviously worth continuing — is what the next two years were about.

We kept every prototype. There are four hundred and eleven of them in a drawer under the bench, in rough chronological order, and the drawer is the most useful document this company has.

The noise problem was never the mechanism

For the first six months we assumed a quieter Twist meant a better bearing. It did not. A sealed bearing is almost silent on its own; what makes noise is everything around it — the shell flexing, the disc bottoming out, a fractionally loose core knocking as it turns. We chased the wrong component for half a year because it was the only part we had bought rather than made.

The fix, when it came, was geometry. Give the disc no hard stop to hit and there is nothing to click against. That single change removed more sound than every bearing we tested combined.

If it does not survive a year in a pocket, it does not get a number.

Three moulds, thrown away

Resin shrinks as it cures. Not much — under a percent — but a mould cut for the size you want gives you a piece slightly smaller than that, and the error compounds across a batch of forty. We learned this three times, each time by cutting a new mould and discovering the same thing in a slightly different way.

The third one taught us to stop guessing and start measuring every batch cold, twelve hours after the pour, before anything is assembled. It sounds obvious written down. It cost about eight months.

The second mould, kept for scale reference. The third is somewhere in the drawer.

The supplier who stopped replying

Halfway through year two, the company making our bearings stopped answering email. No warning, no notice, just silence and then a bounced address. We had six weeks of stock and no second source, which is a mistake you only make once.

The replacement is a smaller shop two hours from here. Their bearings cost more, arrive slower, and have never once failed a batch test. We now hold three months of them and buy from two suppliers, which is the least interesting lesson in this piece and probably the most valuable.

What the drawer is for

People assume the point of keeping prototypes is nostalgia. It is not. When we start a new number, the first thing we do is pull the drawer open and find the point where an earlier idea failed, because it usually failed for a reason that applies again. The Roller Bar exists because prototype 217 was a bad Twist and a very good something else.

Four hundred and eleven objects, one that ships. That ratio has not improved and we do not expect it to.