Research Note: Aviation Learning Curves Are a Warning, Not a Shortcut
Published July 2026
Question
Spaceflight cost discussions often borrow the aviation analogy: fly more, reuse more, learn more, reduce unit cost. CosmosExplore needs to use that analogy carefully. Which parts transfer to launch, and which parts fail when every mission remains low-rate, high-energy, and high-consequence?
Source-Backed Data Points
- Learning-curve research traces the original aircraft-production learning curve to T. P. Wright's 1936 work. Source: Learning and forgetting in the jet fighter aircraft industry.
- SpaceX describes Falcon 9 reusability as a way to refly expensive rocket parts and drive down the cost of space access. Source: SpaceX Falcon 9.
- NASA OIG reported Artemis campaign costs projected at $93 billion through fiscal year 2025 and high SLS production and operations costs. Source: NASA OIG IG-23-015.
- NASA OIG projected about $4.1 billion per SLS/Orion launch through Artemis IV, illustrating the opposite of high-rate commodity production. Source: NASA OIG IG-22-003.
Reading
Learning curves are real, but they are not magic. They need repeated production, feedback, suppliers, inspection loops, and a stable enough design for accumulated work to matter. Aviation had these conditions at large scale. Some launch systems are moving in that direction through reuse and cadence, but lunar human transport still includes low-rate crew safety and mission-assurance layers.
The analogy helps when it reminds the model to track rate, reuse, refurbishment, manufacturing learning, and operations learning. It fails when it assumes that orbital launch will become air travel merely because both involve vehicles. A lunar route includes transfer, landing, crew survival, radiation, return, and authority domains that do not disappear with first-stage reuse.
For CosmosExplore, the useful output is a learning-curve lens with brakes. The model should ask which segment has repeated cycles and which segment remains bespoke. Earth launch may benefit from cadence before lunar landing does.
Model Rule
Cost curves should expose a "learning applies here" flag per segment: Earth launch, transfer, lander, crew system, operations, and regulation. A global learning-rate slider is too blunt for personal lunar access.