Makelab OS

A factory in Brooklyn that happens to have written a lot of software

Makelab OS is the software we built for our own floor and then opened up to the people we build for. Five printing technologies, machines that run overnight, a production floor in Sunset Park, and nine years of the same frustrations turned into something useful.

Manufacturing is usually a black box, and not because anyone is hiding anything. It is because most factories do not have the inside of it organised enough to open. You can only show a client what you genuinely have under control.

So this is not a feature tour. It is the argument for how this place runs, told through the things we built, because every one of them exists for a reason, and the reason is always the same kind of reason: we watched something go wrong, or nearly go wrong, enough times to build a road around it.

01

Ease and record keeping

Nothing about your job goes unrecorded.

Everything you have ever sent us is still here, and everything we have ever done to it is written down.

Every job on the floor has a named technician attached to it, and every part is initialled by whoever inspected it before it goes in the box. Not a team, not a department, not a ticket number. A person, whose handwriting is on it.

The software keeps the other half of that record. Every internal action on a job is logged as it happens: who touched it, what changed, and when. When a question surfaces six months after delivery, we can trace it back rather than reconstruct it from somebody’s memory of a Tuesday.

And none of it asks anything of you. There is no account to create, nothing to file, no reference number to keep. You sign in and the entire history of your work with us is already sitting there.

1 / 7

No account to create

You sign in. There is no form standing between you and work you have already given us.

02

Before money changes hands

We review your files before production.

Almost every expensive surprise in this business starts as a small geometric fact that nobody caught early enough.

Our engineers can tell within seconds of opening a file whether a wall is going to survive the process. For years that judgment lived in one person’s head and had to be passed on by hand. Same day, but still after you had already sent the file and started waiting.

So we encoded it. The check now runs at upload, in your browser, before anyone has taken your money, and the bands are named for what actually happens on the machine rather than for numbers, because “will likely fail” is the sentence the technician would have said to you anyway.

A thin wall found at upload is a five-minute CAD edit. The same wall found on the floor is a failed build, a remake, and a conversation about who pays for it. We would rather lose the order.

1 / 6

Thin walls, in colour

The whole part is shaded by thickness the moment it loads, so the problem is visible rather than described.

03

The number

A quote is a promise, not an estimate.

Most of the industry quotes from a rule of thumb and then spends the project defending it.

Every price here comes off the actual geometry of the actual file, volume, surface area, support burden, nesting density, per technology. Not a category the part resembles, not what something similar cost last quarter. Upload the same file twice and the number is identical twice, because it is computed rather than judged.

The date is held to the same standard. Capacity is checked before a lead time is offered, not after your order lands, which is the entire reason our on-time number holds. If the floor cannot hit a date, you never see that date.

1 / 7

A price in seconds

Computed from your geometry across FDM, SLA, MJF, Industrial SLA and FGF, rather than assigned from a category.

04

The same view we have

You get the tools our engineers use.

A manufacturer can only hand over its own view of a part if that view is worth handing over.

Cut a section plane through the model. Measure across any two points. Strip it to wireframe, put half an assembly away, snap the camera to a face. These are not simplified client-facing approximations of our tools, they are the tools, in the browser, with nothing held back.

Handing them over is not generosity. It is the shortest path to agreeing on what the part actually is before anybody spends money finding out, and the fastest way for you to catch something we might not have thought to ask about.

1 / 8

Cut straight through it

Drag a section plane through the model. Internal geometry is where the surprises live.

05

The mistake we refuse to let you make

Scale is never left to the imagination.

The most expensive ordinary mistake in this business is not a tolerance. It is somebody picturing the part wrong.

Nobody has an instinct for 148 millimetres. Everybody has one for a soda can. So a can stands beside your model at true scale, always, and it is the least sophisticated thing on this entire page.

It is also the one that prevents the most expensive category of mistake there is, because a part that is the wrong size is not a defect we can fix. It is a week, gone, and nobody finds out until the box is opened.

A soda can, always
1 / 4

A soda can, always

Sitting beside your part at true scale, because comparison beats numbers when the real question is whether it is the size you think it is.

06

What we watched happen

We built this after watching people do it over and over.

Repetition is not just tedious. It is where the quiet errors get in.

Forty models in a scene, configured one at a time, is forty opportunities for part thirty-one to end up quietly on the wrong material, and nobody notices until QA, or later. The cost of that is never the clicking. It is the drift.

So a decision made once is applied identically everywhere, and what you approved is what gets built, whether the order is one part or four hundred.

1 / 7

Set it once, apply it everywhere

Configure one part and push those settings to all of them, so nothing is left behind on a stale setting.

07

Ten at night

Things happen outside our business hours.

The measure of an operation is not the smooth order. It is what happens the first time something is unusual.

Tax exemption, a mutual NDA, a Net 30 application, a change order after the fact, a quote made out to whoever actually holds the card. None of it is clever. Each one is simply a thing that used to need a person, a thread and a wait, and now has a route that does not run through anybody’s inbox.

That is the whole argument, really. Every one of these started as a request somebody made more than once, often one of our own team answering it for the fortieth time that week, until it was obviously worth building a road for.

1 / 5

Send it to whoever pays

The person specifying a part is rarely the person holding the card. The quote can be handed over without being rebuilt.

Every one of these started as a complaint

None of this was designed in one sitting. Each item on this page began with somebody saying something was annoying. A client, or one of our own people answering the same question for the fortieth time that week. It got logged, it kept recurring, and eventually it was worth building a road for.

That is the only method here worth describing. If something still is annoying, tell us. It goes on the list, and the list is how this place gets better.