Lot Size One

Fahz 2026

Why the printer was never the hard part.

The claim is simple and it is not really about 3D printing. It is that when you decide every object will be different, you do not just discard the conventional way of designing and building things. You discard a hundred years of accumulated logistics: warehousing, forecasting, batching, palletizing, sampling-based quality control, restocking, and a hundred other quiet optimizations that all rest on one assumption, which is that the next thing coming down the line looks exactly like the last one.

The printer solved the making. Nobody solved the rest of it, and the rest of it is most of it.

What 2016 actually proved

It is easy to remember the first Fahz as a thing that failed. It is more accurate to say it succeeded at the hard technical problem and then got crushed by an easy one.

Two thousand individuated objects, each one generated from a stranger’s photograph, each one different from every other, all of them made and shipped. As a proof that mass customization is physically possible, that is not ambiguous. It is done. I did it on a kitchen-table budget with software I wrote at night.

And the printing was the part that worked. Batches of eight ran overnight and were ready in the morning. Bad builds almost never happened, because the geometry followed a fabrication logic. What ended it was labels, tubes, padding, and the drive to the post office. It was the twenty minutes per order that no amount of clever geometry could remove, multiplied by every order, forever, against a life that already had a full-time job and two children in it.

So the question was never can unique objects be made. I answered that in 2016. The question still open, ten years later, is whether they can be made without consuming a person.

What you throw away

It is worth being specific about what breaks, because the vagueness is what lets people wave the whole problem away.

Inventory stops existing, which sounds like a gift until you realize inventory is also a buffer that absorbs demand spikes, and without it every spike lands directly on production. Forecasting stops working, because there is no SKU to forecast. Quality control by sampling stops working, because a sample of one tells you nothing about the next part, which is a different part. Batching stops working, which is the single largest source of manufacturing efficiency ever discovered. Returns and restocking stop working, because a returned object made from a stranger’s face has no second customer. Packaging stops being a solved problem, because the thing inside is a different shape every time. Even the box has to be designed for a family of objects rather than an object.

Every one of those is a hundred-year-old system quietly assuming repetition. Take repetition away and they do not degrade gracefully. They simply stop applying, and you are left standing in your basement at eleven at night with a tube in one hand and a label in the other.

The honest answer most of the industry gives is that this is why it is not done. The interesting answer is that the assumption is worth attacking, because an object made for one person is an object that person keeps. I still get messages from people asking for custom orders, years after I turned the thing off. Nobody sends an email like that about a vase they could have bought at a store. If we are going to spend material and energy making physical things at all, making fewer of them, and making them the kind people hold on to, seems like a more defensible use of the resources than making a great many that are returned, discounted, and eventually discarded. That is a more than good enough reason to take the logistics seriously instead of throwing up our hands and saying that is not how it is done.

Two halves of the same problem

Here is the thing I did not properly appreciate until recently. While I was away assuming somebody else would solve this, an industry was quietly forming around exactly that, and it arrived at the same insight from the opposite end.

Companies now exist whose entire business is lot size one. You send a file, any file, different every time, and they print it, box it, and ship it to a stranger under your name, at a quantity of one, with no minimum and no tooling. That is the logistics stack rebuilt from the ground up for non-repetition, and the people running it are a decade into the problem.

Read their published design guidance and it all reduces to a single idea: orient parts at forty-five degrees so they eject themselves from the bed, chamfer under overhangs so supports are never needed, keep the first layer simple, fillet the vertical edges. Every rule exists to eliminate a moment when a human hand has to touch the part. Cost at scale is a function of touches, not of plastic.

That is the same sentence as the printing was fine, the post office killed me. They found it from the factory side. I found it from the kitchen table. And the difference between me and the people who built those companies is not insight. It is that they treated the breaking point as the beginning of a company, and I treated it as a reason to stop.

But notice that this is only half of the problem. The back half. It matters only if something upstream produces a file worth printing, and that front half, turning an ordinary person’s phone photograph into a manufacturable, gate-passing, actually desirable object with no human anywhere in the loop, is still broadly unsolved. You can tell it is unsolved because these same companies now publish open invitations asking developers to please go and build it. They are shopping for the missing layer.

I built the missing layer in 2016, before the tools existed to make it easy, and I have spent this year rebuilding it properly. Which means both halves of this problem now exist in the world. They have simply never been joined.

The part I have to get right this time

The first Fahz did not fail financially. That is the detail I have to keep in front of me, because it is the one I am most likely to forget.

It failed because it took over my life. It went unsustainable long before it went unprofitable, and the thing that made it unsustainable was not that I was underpaid for my time. It was that there was no boundary on how much time it could take, and a thing with no boundary will take all of it.

So it is worth saying plainly: the economics are not the main driver here. This is an intellectual and design exercise that I love doing, and I want to do it well and prove the point. But that motivation is not a safety mechanism. It is exactly what I believed last time, and it did not protect me at all. Low financial expectations do not bound the downside. Only a bounded commitment does: the hours I will give it, and a decision made in advance about what happens when it asks for more.

Which is the actual reason for all of the automation. Not margin. Not scale for its own sake. Every automated step is an hour returned, and the hours are the whole ballgame. The pipeline that extracts a profile without Photoshop, the geometry that generates itself without Rhino, the partner who prints and ships without me touching a tube: those are not efficiency measures. They are the conditions under which I get to keep doing this at all.

Why bother

Because the question is still open, and it should not be. Because the technology has improved enormously in ten years and this remains elusive, which means the missing piece was never the hardware. Because I have a proof sitting in an archive folder that individuated objects can be made at volume by one person, and a very clear memory of exactly which part broke. Because somebody else is working on the other half and has been for a decade.

And because it is a genuinely good problem, and I like it. That is allowed to be a reason.