The Sucking Sound Reaches Washington
What I diagnosed inside one engine program, the White House just diagnosed for the whole country
When I wrote The Great Sucking Sound in Engineering, the core finding across all four parts came down to one sentence: the system drove those outcomes. Not one bad hire. Not one bad quarter. A system that selected for speed, rewarded extraction, and treated deep knowledge as overhead. I wrote that from inside a specific lineage of an OEM engine programs, watching specialists leave faster than their knowledge could transfer, watching validation quietly shift from correlation — prediction checked against test data — to seniority, where the most confident person in the room became the de facto ledger.
I didn't write it as policy. I wrote it as an autopsy.
In July 2026, the White House Office of Science and Technology Policy published Science: A New Golden Age — a 123-page report to the President from Michael Kratsios, the nation's Chief Technology Officer. Reading it, I kept finding my own sentences with the serial numbers filed off. Not because anyone read my Substack. Because the failure mode I was describing at one company scales to an entire national research system, and enough people inside that system have apparently been living it too.
Four diagnoses, side by side
The system drove those outcomes. That was never really about any single engineer's choices. It was about incentives that made speed and extraction rational and depth irrational. The OSTP report names the same structural problem under a different label: "The Incumbency Tax" and "Misaligned Incentives," diagnosed in Chapter II as funding and career structures that reward novel claims and short-term output over durable, patient work. Their fix is telling — long-horizon grants explicitly modeled on the NIH Director's Pioneer Award, designed to reward the kind of depth that short grant cycles currently punish.
Validation by seniority replaced validation by correlation. I watched the empirical ledger — the record linking a prediction to what the test actually showed — quietly disappear from engineering organizations, replaced by whoever had been in the room longest. The OSTP report calls this the reproducibility crisis, and it isn't vague about the cost: a majority of researchers now acknowledge the crisis is real, and one study found irreproducible biomedical findings alone misdirect an estimated $28 billion a year. Their fix is the Restoring Gold Standard Science Executive Order, signed May 2025, mandating reproducibility, transparency, falsifiability — and, direct hit on my own thesis — explicit "acceptance of negative results" across all federally funded research.
Short tenure is the mechanism, not a side effect. People have to stay long enough to become the institutional record. When they don't, the knowledge doesn't transfer, it just leaves. The report's Chapter IV, "The Marriage of Science and Craft," makes almost the identical case at the level of an entire economy: technology lives "not just in papers and patents, but in the tacit knowledge passed from mentor to mentee," and decades of offshored manufacturing severed exactly that transfer. Their fix: practitioner-in-residence programs embedding machinists and technicians alongside PhD researchers, national fellowships for skilled craftspeople, and a real push to modernize apprenticeships in science and technology fields — not just the trades.
A confident number with nothing standing behind it. That line was written about a specific failure mode I kept running into — a result stated with total confidence, no verification underneath it, and no one positioned to check. Chapter V of the report warns that AI is about to make this failure mode worse at industrial scale, warning it will soon be "easier to generate plausible-sounding scientific results than to verify them." Their fix: build verification infrastructure with the same seriousness as generation infrastructure — AI-enabled replication systems, machine-auditable replication packages, and real incentives for researchers who replicate or disprove influential results, instead of the current system where confirming someone else's work carries no reward at all.
The part that closes the loop
Here's what ties this back to the other half of what I've been writing. Two of these fixes — Gold Standard Science's "acceptance of negative results," and a separate FY2028 budget directive requiring agencies to fund the curation of data that's otherwise routinely abandoned — are, functionally, the shelf-and-bank mechanism I've been describing for shelved technology programs. Same idea, applied to a different failure mode: instead of a negative result or a failed trial getting buried because nobody had an incentive to keep it, it gets documented, banked, and made searchable for whoever needs it next.
I wasn't wrong about what the system was doing to engineers. I just didn't expect Washington to write the same diagnosis, eighteen months later, backed by an executive order and an actual budget.
The sucking sound reached the White House. Now the question is whether the fixes reach the engine programs where it started.
Herbert Roberts, P.E. — 30+ years in aviation R&D across two companies. 8+ years analyzing accidents for attorneys under my P.E. license.