Praetore is a manufacturer — not a marketplace, and not software. One engineering core writes complete, machine-ready Job Packets. A network of independent American shops runs them on one common standard. One company is accountable when a date slips, and it is us. This is the long version of why that shape and not another, and what we have promised never to become.
It starts from a fact that surprises people: American machining is not uncompetitive at cutting metal. US spindle time is roughly at parity with anywhere in the world. America loses in the hours around the cut — and that is a fixable, structural problem rather than a wage problem.
The hours America actually loses are quoting, DFM review, CAM programming, fixture design, setup, first article, inspection documentation, and the coordination between all of it. Those are skilled-engineering hours: exactly the input where low-wage countries win, and exactly the input this country has a worsening shortage of.
A typical low-volume order spends about two days on a machine and about twenty-five days in the queue around it. That ratio is the entire competitiveness gap for low- and mid-volume precision work. Nobody closes it by making cutting faster.
America doesn't lose at cutting. It loses at everything around the cut. So we deleted everything around the cut.
Every prior attempt to organize American shop capacity is a reasonable answer to a different question. None of them removes the duplicated engineering, which is where the twenty-five days live.
They automated finding a shop.
Every shop on the other side still quotes, still programs, still designs its own fixtures, still writes its own inspection reports. The marketplace added a routing layer on top of an unchanged, inefficient unit. The RFQ still goes out — it just goes out faster, to more people, and each of them still burns the same engineering hours to answer it.
It makes one programmer faster.
AI CAM copilots are real and they work. Sold one seat at a time, they improve shops individually and change the industry's structure not at all. The same engineering work is still done thousands of times over, in thousands of buildings, for parts that are frequently near-identical. Faster duplication is still duplication.
They centralize everything at once.
Engineering and machines together, in new, capital-intensive buildings. It works. It also scales at the speed of construction loans, and it competes with the existing industrial base instead of upgrading it. There are already tens of thousands of American shops with capable machines and skilled people in them. Building new ones is not the constraint.
The unexploited move is to split the two halves of the job: engineering in one place, cutting everywhere.
One engineering core — human experts leveraged by AI CAM, simulation, and automated fixture design — produces complete, machine-ready Job Packets. Independent shops, standardized on common workholding, receive packets and contribute what they are genuinely world-class at: spindle time, machine ownership, and craftsmanship. No RFQs. No quoting. No programming. No sales. Load the pallet, run the verified program, probe the part, ship.
It standardized the interface between cargo and vessel. Loading collapsed from days to hours, cargo became fungible across ships, trains, and trucks, and global trade restructured itself around the standard. The box was not the innovation. The agreement about the box was.
Machining's missing container is the Job Packet plus the workholding standard. Twelve mandatory artifacts, delivered so completely that a shop that has never seen the part can make it without calling anyone.
The twelve artifacts are the Job Packet's mandatory contents. Zero clarifying calls is the packet's release standard, not a measurement: if a shop has to ask a question to make the part, the packet failed.
Once every network shop runs the same modular fixturing and receives the same complete digital packet, three things follow that do not follow otherwise.
Any qualified shop runs any packet. A job can be rerouted mid-week when a machine goes down — which is the only honest way to guarantee a date. A date you cannot reroute is a hope with a font.
A shop joins by installing the workholding standard and passing a test-part protocol. Days, not months. Nothing about that requires the shop to hire an engineer or buy a machine.
Parts from different customers sharing material, stock, and fixture family run in one setup. No single shop sees enough order flow to do this. A network does. It is the machining equivalent of nesting many customers' parts on one sheet.
Not a marketplace. Not software. Praetore owns the customer, the engineering, the quality system, the schedule, and the guarantee. The network owns the machines. Both sides do only what they are best at, and only one of them answers the phone when a date slips.
This company was not buildable in 2019. The interesting question about any new industrial model is not whether it is a good idea but what changed to make it possible, and whether that change is durable. Three things changed, and none of them is reversing.
Toolpath generation that took a programmer hours now takes minutes with human validation, across exactly the 3-axis and 3+2 work that is most of the market. That is the change that makes a central engineering core economically possible for the first time — not a faster shop, a different shape of company.
Shops cannot hire programmers. It is not a hiring-market wobble; the people who know how to do it are retiring. A model where the shop does not need one is not a convenience. For a lot of shops it is the difference between continuing and not.
Defense, energy, and industrial buyers want domestic supply and cannot efficiently buy it from a fragmented base of tens of thousands of owner-operated shops. The demand arrived before the interface did.
The model only works inside an envelope where central engineering can be near-fully verified before a packet is released: 3-axis prismatic parts, a fixed material and stock library, defined tolerance classes. That is deliberately narrow, and it still covers the majority of machined-part demand.
Five-axis work, exotic tolerances, and unusual materials come later, funded and de-risked by the core. Fixture design, chip control, and first-article judgement stay human — the constraint set exists precisely so that human judgement scales. And the guarantee is only underwriteable inside a measured envelope, so every widening of the envelope has to be earned with delivery data rather than argued for in a meeting.
The envelope, stated as a spec sheetThe standard arrives as software before it arrives as iron. Version 0.1 of the workholding standard is a soft-jaw specification on vises every shop already owns: defined vise models, jaw blanks, and jaw-machining programs shipped inside the Job Packet itself.
Cutting the jaws from our program is simultaneously the fixture, the test-part protocol, and the proof that the shop can run packets — at about a hundred dollars a set of jaw blanks. Zero-point pallet hardware is a later upgrade, financed against routed work, and offered only after a shop has been paid several times.
There is no hardware ask in the recruiting pitch. The opening offer is a paying job with free engineering — a gift, not an ask. Trust always precedes iron.
We are publishing the plan rather than the projections. The revenue arithmetic behind it is a private document; what is below is what we are actually trying to do, in order, and what each phase has to prove before the next one is allowed to start.
Note 1 — Praetore has not shipped these jobs yet. Every figure that would describe our on-time rate, our yield, or the size of the network is unpublished because it does not exist. The Ledger explains what will be published, how it will be measured, and when.
Anyone can publish a mission. The useful thing to publish is the list of profitable things you have decided not to do, because that is the list you can be held to. These four are ours. If we ever break one, it will be visible from outside the company, and it should cost us the benefit of the doubt on everything else.
We never send a shop an RFQ. If a shop is being asked to price the work, nothing structural has changed and we have simply built a faster version of the thing that is not working.
The software exists to run our factory floor. It is not for sale seat-by-seat. The moment the product is the software, the incentive is to sell licenses rather than to ship parts on the date we promised.
We do not buy machines that compete with the network. Metrology and test capacity are the stated exception, because someone has to hold the master gauges and it should not be a shop we are also grading.
“Any part, any material, any process” is the epitaph of this industry. The resolution is two lanes: a wide door — the Sourcing lane takes anything, human-quoted and honestly dated — and a narrow factory, where the Catalog guarantees only what has been verified. The envelope grows only as fast as verification does, fed by Sourcing-lane evidence and never by marketing pressure.
The sentence
Praetore is building the shipping container for American machining: one engineering core, one standard, thousands of independent shops — so that buying precision parts domestically is faster and more certain than buying them anywhere else.
Neither of those is you? Write to us. This page recruits more than it sells, and it is meant to. If you want the short version, plus who is actually here and how early we are, that is on the About page.