On April 10, 2026, the Orion capsule named Integrity fell out of the sky and hit the Pacific Ocean off San Diego with what NASA administrator Jared Isaacman called a “perfect” landing. Four astronauts — Reid Wiseman, Victor Glover, Christina Koch, and the CSA’s Jeremy Hansen — climbed onto the recovery deck of the USS John P. Murtha after ten days, 406,771 kilometres at the farthest point, and the first human transit of cislunar space since 1972. The Space Launch System hit its target orbit with greater than 99 % accuracy. Every system on the path performed.

That is the signal. The system underneath the signal is a more interesting problem, and it is the one a working systems architect should be reading.

←TODAY: Artemis II splashed down 5:07 p.m. PDT on April 10, 2026 — first crewed lunar flyby in 53 years, SLS at 99 % orbit accuracy, four green crew.
→3012: By the time off-Earth presence becomes routine, the single-rocket architectures will read like the mainframes of the 1970s: powerful, monolithic, and replaced.
Fulcrum: A successful test flight proves one path through the system works once. It does not prove the system works.

From Saturn V to a distributed graph

Apollo was a monolith. One Saturn V carried the crew capsule, the service module, and the lunar lander in a single stack. The architecture diagram fit on one page.

Artemis is not that. SLS plus Orion lifts the crew. A commercial lander — SpaceX’s Starship HLS or Blue Origin’s Blue Moon, both two to seven times larger than the Apollo Lunar Module per NASA’s Kent Chojnacki — meets them in lunar orbit on a separate launch. The Gateway station, when it exists, will be a third node. Crew, propulsion, and surface hardware travel in different vehicles, on different schedules, from different contractors.

As Ars Technica’s coverage of the post-splashdown briefing put it bluntly, Artemis II was “the lowest hanging fruit” of the programme. NASA’s associate administrator Amit Kshatriya said the same thing more diplomatically: “The work ahead is greater than the work behind us.”

The single points you don’t see until they stop

The architecture diagram of Artemis is a network. The dependency graph — the thing that actually fails — is something else.

Three quiet single-point failures visible from public reporting:

• One Interim Cryogenic Propulsion Stage left. NASA has exactly one ICPS upper stage remaining and an unresolved decision about whether to fly it on Artemis III or save it for Artemis IV. The new Centaur V upper stage will not arrive until Artemis V. A single unit between two flight assignments is a queue of one.
• One Mobile Launch Tower. The tower used for Artemis II sustained moderate damage and now has to return to the Vehicle Assembly Building at Kennedy for refurbishment before it can stack the next SLS. There is no spare.
• One European Service Module per Orion. Orion’s propulsion, electrical power, and life-support backbone is built by Airbus Defence and Space for ESA — a transatlantic dependency that does not appear in any photograph of an American rocket.

None of these are catastrophic on their own. Each is a quiet chokepoint that turns a programme schedule into a single-path graph the moment one of them slips.

Atelier: This is the same problem an architecture practice faces on a complex Wettbewerb-winning project with three structural consultants, two façade systems, and a single fabricator who happens to be the only one in DACH who can extrude the chosen profile. The architecture diagram shows resilience — many partners, parallel workstreams. The dependency graph shows the truth — one missing extruder kills the schedule. PAZ’s Atelier-Code work on building-system dependency mapping starts here: the building information model is not the dependency graph, and a BEP that does not name its single points is a BEP that lies.

The move

The trade-off worth stating plainly: a distributed architecture is more resilient in principle and more fragile in execution. Apollo’s Saturn V was a single point of failure, and that was also its strength — there was nothing to coordinate with anyone else. Artemis is harder because it asks NASA, SpaceX, Blue Origin, ESA, and CSA to all show up on the same week in 2028 with their respective hardware finished.

Do this in the next seven days, regardless of whether you ship rockets or housing: open your current project, set aside the org chart and the BIM federation diagram, and draw — by hand, on paper — the real dependency graph. One box per thing-that-can-stop-you. One arrow per blocking relationship. The exercise of finding the third single point you did not know you had is the whole point of the exercise.

Sources & Further Reading

• Primary: Ars Technica — The Artemis II mission has ended. Where does NASA go from here?
• PAZ context: Artemis II Is Home. The Architecture of What Comes Next Is the Hard Part.