Wanted: Used Engineers
A FEB™ Big Why on Why the Best Companies Won't Hire New Grads for the Hardest Jobs
Read a materials or manufacturing engineering job posting at SpaceX or an Anduril-adjacent hardware company and a pattern jumps out fast: they are not asking what you know. They are asking what you have already touched, broken, and fixed.
SpaceX's composites materials role wants someone with hands-on experience who can walk in and immediately develop new processes and required training for prepreg systems, resin transfer molding, and automated fiber placement — not learn them. Their metallurgy role wants direct experience with in-house forming and casting methods, including additive, on the exact alloys the job touches. Look at who staffs the companies orbiting them, and the recruiting pitch says the same thing out loud: our people come from SpaceX, Anduril, Scaled Composites, Lockheed Martin Skunkworks. The company name on the resume isn't a line item. It's the credential.
That's the whole thesis in one sentence: the leading edge of American hardware isn't hiring engineers. It's hiring used engineers — people who already have the mileage on them, because nobody trusts a brand-new one to survive first contact with a real part.
That should stop you. Not because it's cynical. Because it's rational, and the reason it's rational is a story about why.
The Big Why
We've built the case for this in two earlier pieces, and it's worth stating the mechanism plainly instead of case-by-case.
Why can't a new engineer walk in ready? Because the barrier that would have prepared them acts upstream of the classroom, in domains the core curriculum simply doesn't require. Pull the actual degree charts. MIT's required mechanical engineering core is Mechanics and Materials, Dynamics and Control, Thermal-Fluids, Design and Manufacturing, Numerical Methods, and a thesis. Ohio State's core runs through Statics, Dynamics, Mechanics of Materials, Thermodynamics, Machine Elements, Fluid Mechanics. Neither required list contains composites or additive manufacturing. Where that material exists at all, it's an elective, or it's walled off in a standalone master's program built specifically because the undergraduate core doesn't reach it. The gap isn't a rumor. It's printed on the curriculum sheet.
Why does that gap matter now, more than it did ten years ago? Because the parts being built at the leading edge — printed titanium brackets, optimized composite structures, AI-generated geometry — fail in ways that a classical metals education has no category for. None of it shows up in a converged FEA plot. None of it shows up on a transcript.
Why does that push hiring toward "used" engineers instead of "new" ones? Because a resume can't show you what a person was never taught to look for, and neither can a portfolio of clean stress plots. The only thing that reliably signals someone has actually hit this wall and learned from it is having stood at the machine when a part came out wrong, or having sat in the room when a printed bracket failed a fatigue test the FEA said it would pass. That experience isn't teachable in a lecture. It's only earned by being present for the failure — which means the fastest, most legible proxy for "has the missing training" becomes "worked somewhere that already had to learn this the hard way." Pedigree-by-employer replaces credential-by-institution, quietly, without anyone declaring it as policy.
What the Curriculum Sheet Leaves Out
A few examples make the gap concrete rather than abstract:
Composite cure state. A datasheet material property describes an idealized coupon cured under lab-optimal conditions. The as-cured laminate in a production part — its void content, fiber alignment, degree of cure — can differ from that datasheet property in ways a standard FEA model never sees, and standard metals NDE (X-ray, dye penetrant) won't catch it. That requires C-scan or thermography, tools most metals-trained engineers have never been taught to ask for.
Printed-metal grain structure. A thin wall in an additive process isn't just a stiffness question — it's a different melt-pool history, which changes grain morphology and porosity in that specific feature. And fatigue behaves differently depending on which regime you're in: high-cycle fatigue is often driven by as-built surface condition and near-surface porosity, while low-cycle fatigue is governed by columnar grain structure and defect population. Two separate problems, not one problem with a different number plugged in.
Optimized robustness. A part topology-optimized against its design loads has incidental margin deliberately removed — that's the entire point of the optimization. The better it worked, the less is left over for the load case nobody wrote into the objective function: the hammer strike, the dropped part, the mis-rigged handling load.
None of this shows up in a converged stress plot. None of it shows up on a transcript.
What "Used" Actually Buys You
A used car, at minimum, comes with a maintenance record you can inspect. A used engineer comes with something better and worse at the same time: scar tissue you can't audit directly, but that you can trust probabilistically because you know which fires they were standing near. Someone who cut their teeth on SpaceX's composite shop floor has, almost by definition, been present for a resin-flow defect, a void-content rejection, a cure-cycle deviation that quietly changed the material state of a part that looked fine on paper. That's not something you can put in a course. It's something you can only get by being in the building when it happened.
This is the honest answer to "where will Anduril and SpaceX hire these skills" — not a pipeline, not a new degree track, but a small number of organizations functioning as the informal graduate school for the entire industry, training people the hard way and then watching everyone else recruit from their alumni.
You don't have to look far for a current example of exactly this kind of failure. Blue Origin's investigation into the May 2026 New Glenn pad explosion — traced to the main oxygen valve on one of the booster's BE-4 engines — is a reminder that this isn't a hiring thought experiment. Some of the most capable hardware engineers in the country, working at one of the best-funded and most scrutinized rocket programs on the planet, still had a component behave outside what the design and test program anticipated. That's not an indictment of Blue Origin's team. It's the pattern itself, in public, in real time: even at the leading edge, the barrier finds the gap before anyone else does.
San Bruno: The Steel Was Never the Problem
The pipe segment that ruptured under a San Bruno street in September 2010 had been in the ground since 1956. Fifty-four years of pressure cycling, soil movement, and ordinary wear, and the steel itself never gave out.
What failed was something that never carried a physical load at all: the paperwork. The NTSB's investigation found that the pipe's construction records didn't match what was actually buried in the ground — a welded seam type on file that didn't correspond to the pipe segment installed at that location. Nobody caught the mismatch for over half a century, because nobody had a reason to go back and check a record against the physical part it was supposed to describe. The record wasn't being tested. It was being trusted.
That's the second barrier, and it's the one this piece is actually about. A metallurgical failure gets inspected, tested, and replaced on a schedule. A records failure just sits there, silently correct-looking, until the day something else — a pressure transient, a control-room decision, an ordinary Tuesday — forces someone to finally rely on it. San Bruno's steel held for 54 years. Its records were wrong from day one, and nothing in the system was designed to notice.
The Question Worth Asking
Not "does this candidate have a degree in the right field." Not "does this vendor's design tool produce clean output." The question is narrower and it's the one nobody puts in a job posting:
Who taught this person, or this process, what a barrier looks like — and was it a classroom, or was it a failed part?
If the honest answer is "neither yet," you're not necessarily looking at a bad hire or a bad vendor. You're looking at someone who hasn't been used yet. That's fine — as long as you know it going in, and build the guardrails accordingly, instead of finding out the same way the last company found out.
Herbert Roberts, P.E. spent 30+ years in aviation R&D across two companies and analyzing accidents for attorneys under his PE license, translating engineering findings into legal language.
This piece extends a FEB™ application series — San Bruno (barrier acting through time), the metals-to-composites training gap (barrier acting through domain), and this one: the barrier acting through the labor market itself, and the quiet industry-wide answer to it — hire used.
I am an enginner and I wholeheartedly agree with your analysis. The classrooms do not impart the best practical experience.
At big companies, one mistake could result into a substantial loss before it can be corrected and worse of they lose credibility.
Dennis, this is the whole argument in three sentences. A classroom can hand someone the theory, but it can't hand them the moment when the simple side of the theory runs out and the deadline doesn't care. People's understanding matures once they've have failures ,big and small, with their name on them — that's the tuition no school charges. "School of hard knocks" is just forensic engineering without the fee.