The founders of Hera come from three generations of manufacturers. My grandfather ran a process equipment factory. My father runs a paper mill. I grew up watching physical things get made, and watching how much of that work is spent not on making things, but on fixing misunderstandings about what was supposed to be made.
That gap, between what someone intends and what actually gets built, is where I've spent my whole life. It's the family business. Hera is how I'm going to close it.
Hera Mission
General intelligence for manufacturing. Every part built right the first time.
The World Runs on Custom Parts
Look around you. Almost everything within reach was mass produced. But the machines that mass produced them were not. The pressure vessel at the refinery, the conveyor in the food plant, the tooling in the auto factory: each one was engineered to order, built once, for one customer, to one specification.
Engineer-to-order manufacturing is the industry that builds industry. It's the layer underneath everything, and almost nobody is building for it. Software eats the world in order of how easy the market is, and ETO is hard. Every job is different. Every customer is different. There is no "average user." So the companies that make everything else possible still run on email, tribal knowledge, and paper.
Intent Becomes Instruction
Every physical object begins as intent in someone's head. Between that intent and the finished part sits a chain of translations: requirement to design, design to drawing, drawing to quote, quote to plan, plan to machine, machine to part.
Every translation is a chance to lose meaning. And in manufacturing, lost meaning isn't a bug ticket. It's scrapped metal, a stopped machine, a blown ship date, a customer who doesn't come back. The physical world charges interest on misunderstanding, and the rate is brutal: an error that costs a dollar to fix at a desk costs ten on the floor and a hundred in the field.
The companies that win the next fifty years of manufacturing will be the ones that lose the least meaning in translation.
The Retiring Brain
For a century, the thing holding these translations together was people. A senior engineer who had seen thirty years of jobs and knew what breaks. A shop foreman who could look at a spec and know it would fail before anyone cut metal. Judgment, carried in heads.
Those heads are retiring. The average skilled manufacturing worker in America is pushing sixty, and the knowledge walking out the door was never written down. You cannot hire your way out of this. The apprenticeship pipeline that built that judgment took decades and no longer exists.
But for the first time, machines can read the way those people read. Not scan. Read. Understand a document the way a person who has seen ten thousand of them understands it. That changes what's possible for every factory that can't find its next senior anything.
Full Stack or Nothing
We don't believe you fix manufacturing by selling software to it. Software companies that sell tools to factories learn the industry secondhand and price themselves as a line item. We think the endgame is owning outcomes: taking a customer's intent and handing back a finished part, with intelligence running every translation in between.
We can say this because we're not tourists. Our families run factories. We have a manufacturing network behind us that can build what our software understands. The plan has always been to fuse the two: the intelligence layer and the physical layer, one company.
Where We Start
You start where the meaning gets lost first, and that's the drawing.
The 2D engineering drawing is still the legal contract of manufacturing. Not the 3D model. When a machinist decides what to cut and an inspector decides what ships, the drawing is the law. And today most drawings get reviewed by whoever has twenty spare minutes, which means they don't get reviewed.
Hera's first product reads every drawing before it reaches the floor and checks it the way the best senior checker would: against ASME and ISO standards, against GD&T, against the model, against manufacturability itself. The checker who never retires, never rushes, never has a bad Friday.
Taste Compounds
But standards checking is table stakes. The real reason we start at review is taste.
Ask a great senior checker how they know a drawing is wrong and half the time they can't tell you. They just know. That knowing is taste: thousands of drawings, thousands of parts that failed or didn't, compressed into judgment about what will actually build. It's DFM and DFA as instinct rather than checklist. In mass production you can learn it from one part repeated a million times. In ETO you can only learn it the way checkers did, by reviewing everything.
The review is where that judgment gets exercised and corrected. A checker flags something, the floor pushes back or the part fails, and the judgment updates. That loop, review, feedback, revision, is the only place in ETO where taste gets built. It's how humans built it for a century, and it's how a machine has to build it now. Sit anywhere else in the workflow and you see fragments. Sit at review and you see every drawing, every mistake, every correction, across every job.
So drawing review isn't just our first product. It's our training ground. Every review makes the next one sharper, and taste is the one thing in manufacturing that compounds.
Bad drawings make bad parts. We fix the drawing. Then we build the part.