Withdraw With Interest
Why Congress Shouldn't Pick the Technology — and What the White House Just Funded Instead
I've watched Congress question people like Mark Zuckerberg and Elon Musk — grilling them on what they're building, how they're spending their money, where the technology is headed. And every time, the same thing is obvious within the first few questions: the people asking don't understand what they're asking. The questions are staffer-written, read off a page, and it shows the moment a follow-up strays off script.
That's not a knock on any one senator. It's structural evidence of something more useful: Congress is tuned to governance — international affairs, state relations, the rights and needs of constituents. It is not, and was never built to be, tuned to the daily pulse of technology. So when government tries to pick the technology of the year — decide by committee what deserves funding and what doesn't — it's asking the wrong experts to make a technical call.
For most of my career, that's exactly the trap researchers operated inside. Chase the government's priority list, reshape your own research to match whatever the administration decided was the trend of the year, and hope the money follows. Miss the window, and the work — and the money already spent on it — just disappears.
Here's the distinction that matters: the government can help lead into the future through funding and technology banking. It does not have to steer the path that brings out the innovations. Those are two different jobs, and confusing them is exactly how you end up with a senator reading a staffer's question about a technology he's never used.
The Shelf, Not the Scrap Heap
Here's the reframe I've been working through: not every idea that stalls has failed. Sometimes the technology isn't ready for the need. Sometimes the need isn't ready for the technology. Either way, shelving the work — instead of cancelling it outright — lets the two catch up to each other later.
The difference between a shelf and a scrap heap is documentation. If you record, at the moment something gets shelved, why it was shelved — which side of the gap failed, technology or market — the archive becomes searchable instead of buried. In a digital age, with AI able to continuously scan that archive against current conditions, what comes back off the shelf isn't the original idea. It's a 2.0 derivative — improved by whatever changed while it waited.
Think of it like a bank. The money doesn't vanish because you didn't spend it this quarter. It sits, and it earns interest. Withdraw it later, and you get back more than you put in.
Three Ways a Shelf Gets Built
I've watched this play out three different ways across a single materials lineage in aerospace, and each one teaches something different about why things sit.
Supply-chain desync — ceramic matrix composites. GE Research Center and GE Energy both knew the physics case for CMCs decades before the technology reached the aviation world: run turbines hotter, keep parts in service for twenty or thirty years instead of throwing away worn metal, push T4 higher and get more efficiency out of the whole engine. The idea wasn't missing. What was missing was fiber available at volume, and a manufacturing base sized to actually produce at the scale GE Energy needed. That's not a science problem. That's an industrial capacity problem, and no amount of research funding fixes it faster than the supply chain can physically build.
Domain-transfer desync — PMC fan blades. Polymer matrix composites had already been proven, for years, in marine and boating applications before they ever made it into a jet engine. The material science wasn't in question. What took so long was qualifying that same material for a completely different stress environment — bird strike testing, fatigue certification, everything that separates a boat hull from a rotating blade under load. The technology was ready. The pathway to this specific application wasn't.
Sequential capability unlock — 3D printing. This one's the most interesting because it isn't one idea waiting on one thing — it's a whole stack, each layer waiting on the one below it to mature. Stereolithography and early CAD-to-plastic printing proved you could go from a digital model to a physical part, but the parts were floppy, imprecise, structurally unconvincing. Fiber and powder-based plastic printing solved precision — now the process itself could be trusted. Powder metal printing — Pratt was a leader here with titanium — then asked whether that same trusted process could carry a real structural material, and the answer helped make SpaceX's Raptor engine possible. Now composite and microfiber printing is the current unlock: an entire pre-existing database of laminate designs, sitting ready, waiting only for the printer to catch up to the material. Nobody has to reinvent that science. It's already banked. It just needed the process to mature enough to carry it.
This Isn't a What-If — It's Already Funded
Here's the part I want to be precise about: none of this is hypothetical, and the infrastructure it depends on is already law and already funded. In November 2025, President Trump signed Executive Order 14363, launching the Genesis Mission — a whole-of-government effort directing the Department of Energy to build the American Science and Security Platform, connecting the nation's supercomputers, AI systems, and scientific datasets into a single discovery engine, with the explicit goal of doubling U.S. science productivity within a decade. By December 2025, DOE had already signed agreements with 24 organizations, including AI companies, chip makers, and cloud providers. A companion piece of that infrastructure, the American Science Cloud, exists specifically to curate and distribute data currently "hidden behind government bureaucracy."
And critically, the FY2028 R&D budget memo — an actual annex in the OSTP report, not aspiration — directs every federal agency to propose funding specifically to curate data that is otherwise routinely abandoned for lack of funding or recognition: experimental records, negative results, operational data from failed trials. That's the closest the government has come, in concrete funded form, to the bank-and-scan mechanism I'm describing.
To be precise about where the report's own language stops and my framework starts: the government's current mandate is built around data — failed trials, negative results, records that would otherwise be lost. It is not yet built around shelving whole technology programs — the way GE Research Center sat on CMC knowledge for decades waiting on a fiber supply chain, or the way PMC blades waited years for an aviation qualification pathway. That synchronization mechanism, applied one level up from data to entire programs, is my own extension of what the government has already funded and launched.
Good — Now Take It Further
To that first step: genuinely, good. As Musk has put it, the worst thing you can do is optimize a bad idea. What this report gets right is recognizing the opposite mistake is just as costly — killing a good idea because it wasn't ready on this year's schedule. The Federal Government is finally proposing to operate the way GE Research Center already did with CMCs: it didn't throw away decades of ceramic matrix knowledge just because the fiber supply chain wasn't there yet. It kept working the problem at the margins, kept the knowledge alive, and waited for the rest of the world to catch up. Eventually it did.
Thank goodness somebody with some forethought is finally at the steering wheel of technology.
But there's a next step that would multiply the value of the exact same infrastructure investment, without appropriating a single additional dollar for R&D itself. The report's own numbers make the case: American industry deploys roughly $700 billion annually in R&D, more than triple the combined spending of government and higher education. The Genesis Mission and the American Science Cloud, as funded today, only reach federal data. That means the shelf they're building covers a fraction of the nation's actual R&D pie — the smaller fraction.
The fix is to extend the same tagging discipline outward:
A standardized Readiness-Gap Tag — the same documentation the report wants agencies to apply to abandoned datasets, offered as a voluntary standard private industry could apply to its own shelved programs: what failed to sync, tech or market, and on what specific dimension.
A real incentive to make tagging worth a company's time — an R&D tax credit multiplier for documented, taggable shelved work, or an IP safe harbor that lets a company disclose that something was shelved and why, without disclosing the trade secret itself.
A federated private-sector mirror of the American Science Cloud — companies keep the actual work behind their own walls, but the readiness-gap metadata becomes searchable, so a national AI scan could flag that a company's shelved program and another company's new capability are a sync match, without either party exposing anything they haven't chosen to.
Build the tagging standard once, extend it across both federal and private shelves, and the same dollar already being spent on Genesis Mission's infrastructure now indexes a $900 billion pool instead of a $200 billion one. That's more power per dollar spent — not by spending more, but by finally being able to find what's already been paid for.
Let the people who understand governance handle governance. Let the people who understand technology handle technology. Government's job is to fund the bank and keep the shelf standing — not to steer which withdrawals get made or which path innovation has to take to get there. And let the shelf — federal and private alike — do what a good bank does: hold the value, quietly earning interest, until someone's ready to withdraw it and build something better than what they put in.
Herbert Roberts, P.E. — 30+ years in aviation R&D across two companies. 8+ years analyzing accidents for attorneys under my P.E. license.