A line from the September 26 dispatch reads like a technical footnote: the Crew-13 team arrived in Florida aboard two T-38 training jets. Canadian Joshua Kutryk and American Luke Delaney were at the controls; commander Jessica Watkins and Roscosmos cosmonaut Sergey Teteryatnikov rode in the back seats. Four days later, they would launch to the ISS aboard a Dragon.
That's not what hooked me. A month earlier, on August 27, NASA officially announced it was looking for a replacement for its T-38s. In other words, in the very same August, the agency publishes a document about retiring the aircraft and puts a crew bound for the station on it. Between these two events lies a question worth digging into: why is the road to space still routed through a machine designed in 1959, and why is it so hard to replace.
NASA JSC document number 80JSC026SFRTAM is an RFI — a Request for Information, not a tender. It contains a warning in all caps: "THIS IS NOT A REQUEST FOR PROPOSALS. NO PROPOSALS ARE BEING SOLICITED IN RESPONSE TO THIS NOTICE." Responses were collected through September 28, limited to no more than 15 PDF pages, and the agency stated upfront that it would not pay for information and might not respond to questions at all.
Then it gets interesting. The reason for the search is stated plainly: the U.S. Air Force is retiring its T-38s in the 2030–2035 window, and NASA has spent 60 years relying on a shared Air Force maintenance and parts base. When the fleet owner leaves, the entire ecosystem of warehouses, repair lines, and suppliers goes with them. NASA would be left with 25 aircraft and no logistics tail.
And here the document stops looking like a military procurement. The agency asks respondents about the possibility of "lease, lease-to-own, barter, or other arrangements" — and explains the reason almost matter-of-factly: "due to limited funding for initial acquisitions." A space agency planning a Moon landing is explaining to contractors that there's no money to buy aircraft right now and it's open to a swap.
The requirements for the aircraft read like a list of things the current T-38 no longer is. It needs "high reliability with established maintenance documentation" — a delicate way of saying the airframe's parts are long out of production. It needs "the ability to start engines independently" — at remote fields without ground air start carts, a T-38 can't start without one. And as a separate item: "it is highly desirable that the aircraft permit single-pilot operation from either seat." This requirement eliminates most modern training aircraft, certified under the "instructor in front, student in back" layout.
NASA also wants to know maintenance man-hours per flight hour, the percentage of serviceable aircraft, cost per flight hour, and willingness to provide "NASA pilots with a contractor responsible for maintenance."
To understand where these requirements come from, you have to go back to a document that reads like a verdict on simulators. In 2011, the U.S. National Research Council examined what to do with NASA's fleet after the shuttle program closed, and devoted a separate chapter to the T-38N. The main argument there isn't about G-forces — it's about irreversibility.
The formulation is almost philosophical: there is no "pause" button in an airplane. In a simulator, even on a motion platform, a person can stand up and walk out when the scenario goes bad. In a cockpit above 13,000 feet, he's strapped into an ejection seat, breathing through a mask, and can neither leave nor replay. Emergency procedures in simulators are practiced at a normal load of 1 G — conditions that won't exist in a real failure.
Further in the report is a table that reveals something people don't usually talk about: of all NASA's training resources, only the T-38 simultaneously provides motor skills, cognitive workload, and rapid reaction. The pool for EVA training provides 3 out of 4 categories; station simulators provide fewer. The "rapid reaction" column is unique to the T-38 in the table.
The G-forces are listed there too: the T-38 accelerates from zero to Mach 1.3, pulling up to +7.33 and down to −3.5 G. The shuttle pulled a maximum of 3. A ballistic reentry in a Soyuz — up to 8. And another detail: a fighter pilot experiences G-force as compression into the seat, while a rocket crew experiences it as pressure in the back, because they're lying down. These are different sensations, and training the second with the first works only partially. The report's authors acknowledge this in plain text.
The most surprising part of the report is the appendix with the training program. It reveals that mission specialists — the doctors, biologists, and engineers who will never sit at the controls — are not passengers in the back seat of a T-38. They are crew members.
The annual proficiency check requires them to: quote emergency procedures from memory, know limits for exhaust gas temperature, nozzle, minimum fuel remaining, oil pressure, and hydraulics; compute takeoff and landing data; check weather and NOTAMs; file flight plans and receive clearances; navigate by instruments; calculate landing speeds; know the ejection system, harnesses, parachute, and oxygen equipment. And at the very end of the list, in all caps with four exclamation marks: "RECOGNIZE ANY HAZARDOUS SITUATION OR CONDITION!!!!" The program includes simulated engine-out landings, no-flap landings, short-field takeoffs, and single-engine go-arounds.
Also in there is a number that explains why NASA clings to this aircraft. The report states directly: as soon as flight hours drop below the level set by the program, crew performance in the cockpit deteriorates — instructors observe degradation in hand-eye coordination and decision-making. Instructors call the current standards "minimally acceptable." The standard for pilots is 15 hours per month; for non-pilots, a minimum of 4.
The history of the T-38 itself is full of inversions. The prototype's first flight was April 10, 1959; it entered service on March 17, 1961; production closed in 1972 after 1,187 aircraft. Northrop won the competition not on flight performance but on lifecycle cost: the only rival was the two-seat F-100 Super Sabre, an aircraft that, as aviation literature notes, doesn't recover from spins. The key decision — two engines instead of one, for a safety margin.
Then comes the part that makes this story worth reading. In the 1960s, the T-38 was considered too easy to fly compared to combat fighters. By the 21st century, it had become the most complex aircraft in the U.S. Air Force fleet. The reason is technical: the wing was designed for 7.33 G and a fatigue life of 4,000 hours. That was enough for the training program, but under aggressive maneuvering, cases of wingtip separation in flight appeared, and the fleet later received new wings with thicker skin. The controls are hydraulic with no manual reversion: if both engines fail, the aircraft is uncontrollable.
Plus a short but telling detail from the registries: there are 7 of these aircraft in private hands in the U.S., 2 of them owned by Boeing and flying as chase planes.
Now for something rarely discussed in stories about a training aircraft. All NASA astronaut fatalities in the 1960s — 8 people. 4 of them died not in space and not on the launch pad, but in T-38s.
On October 31, 1964, Theodore Freeman was returning from St. Louis to Houston and on final approach in fog struck a flock of geese. A goose went into the left intake, the engine flamed out; canopy fragments went into both engines. Freeman steered the aircraft away from a residential area and ejected nearly horizontally, too low — the parachute didn't have time to deploy. His wife learned of his death from a Houston reporter who drove to their house. After this, NASA changed the procedure: astronauts notify the family of an astronaut's death, and as quickly as possible.
On February 28, 1966, Elliot See and Charles Bassett, the prime crew of Gemini 9, were flying to St. Louis for rendezvous simulators. The weather was atrocious: rain, snow, fog, ceiling at 1,500 feet. Both aircraft overshot the outer marker and went past the runway; the backup crew, Stafford and Cernan, went around for another approach, while See executed a left turn at 500–600 feet, engaged afterburner, and at 8:58 a.m. crashed into the roof of Building 101 at the McDonnell plant. 17 employees inside were injured. See and Bassett died 500 feet from the spacecraft they were supposed to fly.
Alan Shepard headed the commission; the verdict was pilot error. Deke Slayton in his memoirs called See's technique "girly," while Neil Armstrong and others defended him. The crash had no engineering consequences: the spacecraft was unharmed, only a piece of wing debris hit the unfinished Gemini 10. But the personnel consequences stretched over years. Stafford and Cernan's backup crew became the prime crew for Gemini 9A; their backup slots were taken by Jim Lovell and Buzz Aldrin — and through normal rotation, Aldrin got Gemini 12. Without the experience of that mission, he would have been an unlikely candidate for the Apollo 11 crew. The second man on the Moon got there through a chain that began with a crashed training aircraft.
On October 5, 1967, Clifton Williams, a Marine Corps pilot and member of the third astronaut class, was flying to visit his parents in Alabama. The controls jammed; the ejection seat didn't save him. He was the backup lunar module pilot under Pete Conrad. Al Bean took his place on the prime crew, and his mission became Apollo 12 — the second Moon landing. At Bean's suggestion, the mission patch got 4 stars instead of 3: one for Williams.
Replacing the T-38 at NASA isn't a new idea. In March 2008, the Flight Operations Directorate completed a months-long study and seriously considered 7 alternatives: the carrier-based T-45 Goshawk, the turboprop T-6 Texan, the business jets Premier 1A, Beech 400XP, and Learjet 60, and the light propeller-driven Cirrus and Lancair. The conclusion was unexpectedly stubborn: the T-38N delivered more of the required qualities than any single aircraft or any combination of them, and should be retained. The Learjet 60 had one noted advantage — side-by-side seating, as in the Soyuz and new capsules, useful for practicing teamwork. They even looked at the F-16: newer, but more expensive to operate.
Now there's a different candidate on the table. The Boeing T-7A Red Hawk passed Milestone C on April 23, 2026, receiving a $219 million contract for the first 14 production aircraft — including spare parts, ground equipment, and training packages, which is why naively dividing the sum by the count gives a figure half the actual aircraft cost. The first instructors from the 99th Training Squadron qualified in June. The target date for initial operational capability is November 2027; the full-rate production decision slipped to January 2029. The program has been delayed so much that the Air Force has had to push back its own Talon retirement timeline.
A separate thread the press noted but the document itself doesn't mention: the current NASA administrator founded Draken International — a company providing "adversary" services in private fighters, once the largest private air force in the world. And NASA's RFI explicitly invites discussion of barter and leasing aircraft from private owners. The coincidence is arithmetic, not legal — but it reads exactly the way it reads.
The first discrepancy I found at the very beginning and couldn't resolve. According to the FAA registry, NASA has 25 T-38s; this is the number cited in press analyses as well. An aviation encyclopedia, citing the same registry, puts the fleet at 32 aircraft. Both numbers circulate in publications, and no source specifies what it's counting: retired, flying, or stored.
The second is the accident statistics from the 2011 report. It lists rates of 1.09 for the Air Force, 1.28 for the Navy, and 1.90 for the Marine Corps, without specifying what they refer to. They appear to be accidents per 100,000 hours, but the available text doesn't spell this out, and I won't hazard a definition on my own.
Another minor thing that honestly points to the age of the sources: the T-38's maximum speed lives in 3 different values across various documents — Mach 1.08, 1.2, and 1.3. Formally these are different regimes and different modifications, but across the texts it looks like 3 different aircraft.
Back to September 2026. Consider what happened in the span of one month: NASA opened a search for an aircraft replacement, officially acknowledging the limits of its budget, and weeks later put 4 people on the very same aircraft to fly in for the launch. There's no contradiction here — there's a statement of fact.
Because the T-38 in this system doesn't function as a simulator. It functions as an entry filter and a common language. Everyone passes through it: the fighter colonel with 1,000 hours under his belt, and the biologist reaching for a navigation toggle for the first time. Both must be able to do things that a questionnaire can't verify. The ability to keep your head in a situation you can't walk away from — that's precisely the skill that no diploma can certify, and it's needed equally in the cockpit of a T-38 and in a burning station on orbit.
NASA's requirement for "single-pilot operation from either seat" looks like bureaucratic nitpicking, but it means one simple thing: the agency isn't buying an airplane. It's buying the preservation of the rule that the person in the back seat is not a passenger.