To the Builders of American Industry

To the builders, buyers, shop owners, foremen, estimators, machinists, engineers, quality managers, and program leads who still make real things — and to the programmers, data scientists, economists, politicians, investors, bankers, and operators who support them and depend on them, whether they know it or not:

We are living through an era of industrial confusion, and a fair amount of it is being misdiagnosed on purpose.

Every week the same story runs. The United States has a manufacturing capacity problem. Not enough shops, not enough labor, not enough machines, not enough training, not enough industrial policy. Add the missing inputs and we recover what we lost. I have spent the last several years inside this problem, and I think that story is incomplete in a way that reliably produces the wrong remedies. America does not have a capacity problem first. It has a coordination problem first, and beneath that, a trust problem.

You can feel it in the daily work. Buyers fire RFQs into the dark like flares and still cannot find the right shop in time. Shops spend unpaid hours decoding an ambiguous package, then lose the job to somebody faster or cheaper. A quality team builds a wall of paperwork after one bad supplier event and then wonders why lead times keep stretching. Engineers revise the model, then the drawing, then the note block, and everyone downstream pays the latency tax. Supplier qualification gets rebuilt from scratch, over and over, as though the last decade taught us nothing. None of that is a moral failure. It is mechanical, and it is what happens when a large system loses shared context and reliable signals at the same time.

The atoms versus bits debate is over

Manufacturing is not physical work with optional software bolted to the side. It is already a cyber-physical system. The cutting happens at the spindle, but the work exists in software long before the first chip comes off: CAD, CAM, tool libraries, post processors, probing cycles, inspection plans, CMM programs, travelers, ERP, MES, machine control, and now vision and sensors.

Plenty of shops still treat software as something to negotiate with or resist, and I don't think they're being irrational about it. They have been burned. Software people show up with a new standard, a new platform, a new workflow, and a promise that ignores how parts actually get made. They don't understand workholding. They don't understand setup risk. They don't understand that a tolerance isn't a number, it's a process — and honestly, plenty of mechanical engineers don't spend much time in the reality they're specifying either.

So the industry adopts software where it has no choice, in CNC and CAD, and resists it where it feels threatening, in quoting, supplier discovery, documentation, and data sharing. That half-adoption doesn't hold. It creates seams, and the seams create translation work, and translation work creates latency, and latency creates mistrust, and mistrust creates manual controls, which create more seams. Around it goes.

This series starts from a blunt premise: if we want coordination at national scale, we have to embrace software fully, and build it with respect for the machines, the process knowledge, and the incentives of the shop floor. Software as infrastructure, not as ideology.

The thesis in plain language

Here is the whole argument before I spend forty papers defending it. Distance and complexity broke shared context, which turned sourcing into a translation problem. Distance itself is cheap; uncertainty is what costs money, so the enemy was never miles, it was latency and ambiguity. That distinction matters, because it changes what you build.

We can't impose a single authoritative standard across this ecosystem, and I'd argue we shouldn't want to. But translation and structuring are genuinely solvable now with modern software, especially AI, in a way they weren't ten years ago. Trust has to be engineered rather than assumed, out of evidence-based identity, performance signals, and consequences that actually bite. Globalization amplified the breakdown and exposed us to regimes that don't play by our rules. Some of those nations coordinate faster than we do through centralized control, and we are not going to copy that.

Inputs matter — energy and labor are real constraints — but coordination is what determines throughput and resilience. Incentives and measurement drive where work goes, which means visibility is a form of governance whether or not anyone designed it that way. National boundaries turn out to be the practical container for aligned incentives and enforcement. And underneath all of it: coordination is computation. What cannot be represented cannot be routed, testability wins, and fallbacks are mandatory.

That's the case. It isn't nostalgia and it isn't grievance. It's system design.

Why this matters now

Two comforting illusions dominate the current conversation, and both of them cost us time.

The first is that we can outspend the problem. You can subsidize machines and training, expand incentives, stand up new programs, and none of that is wrong. But if the network stays noisy, untrusted, and slow, new capacity doesn't convert into throughput. It converts into more quoting waste, more mismatch, and more work leaking overseas the moment a deadline gets real.

The second is that national industrial policy will handle day-to-day execution. Government can set boundaries and enforce rules. It can use tariffs and procurement to blunt unfair advantage, fund foundational research, and build workforce pipelines. All of that helps. What government cannot do is run the coordination layer of American manufacturing. It cannot update at the pace of real programs, and it certainly cannot adjudicate every tolerance stack, every conflicting note block, every revision mismatch, every supplier dispute, every capacity shift, every late quote. Stake the industrial future on political timing and we lose, because the system moves faster than policy does. The durable version of this has to be built by industry, for industry, in a way that fits a free and decentralized country.

Why I'm suspicious of "one true format"

The obvious fix, and the one I hear most, is a single standard. One format for RFQs. One national data model. One enforced playbook.

It fails for a reason deeper than the technical difficulty: imposing a universal standard requires coercive authority we do not have and should not want. Our advantage was never that we can coordinate by decree. It's that we can coordinate by voluntary adoption, when a protocol is obviously useful, when it reduces work, and when it makes outcomes better. So I'm not chasing one perfect standard in these papers. I'm after protocols that translate between imperfect inputs — systems that accept messy reality and still produce routing you can trust.

What this series will do

Over forty papers I'll make the case for an American industrial renewal that is pro-market without being centrally planned, pro-small-business while staying honest about scale economics, and pro-national-resilience without the culture-war bait.

The structure is deliberate. Act I, Papers 1 through 10, is diagnosis — naming the failure modes in operational language: ambiguity, latency, noise, fragmented standards, repeated qualification, and the trust spiral. Act II, Papers 11 through 25, is the system — the universal request object, minimum viable work packages, AI-assisted structuring, verification, reputation, trust scoring as routing, visibility rules, domestic-first governance, and interoperability outside the ERP. Act III, Papers 26 through 40, is objections and governance — commoditization, platform capture, gaming, pay-to-play, IP leakage, quality risk, buyer inertia, and antitrust, ending in an industrial constitution: what has to be protected for any of this to stay legitimate.

Along the way I'll use evidence and numbers where they clarify something and leave them out where they'd just decorate the page: quoting cycle-time studies, supplier response-rate surveys, cost-of-quality research, lead time and expediting economics, and the documented failure patterns of marketplaces and broker models.

A note on motives and tone

I'm not writing this to win an argument on the internet.

I'm writing it because too many good people inside American manufacturing are burning their days on avoidable confusion, and the country pays for that in slow iteration, missed schedules, and supply chains that snap under the first real shock. Reasonable people will disagree with me about tactics. Some want heavier policy. Some want pure market forces. Some want to believe the past can be restored through willpower. This series isn't an attack on any of them. It's a claim that the mechanics are what they are, and that if you want different outcomes, you have to build different infrastructure.

So here is what I'd ask you to carry into Paper 1. Treat manufacturing as a directory problem and you'll create noise. Treat it as a marketplace problem and you'll optimize for price while damaging trust. Treat it as a coordination problem and you can move throughput without anyone seizing central control. My bet, and the reason I'm writing all of this down, is that the nation which builds the best coordination protocols will out-iterate the rest — even at higher wages.

Next: Paper 1, the Tower of Babel problem, and why translation is the root cause sitting underneath the pain.