A Corroded Nut on Omelek and a Wave Called Dilep
We followed SpaceX from Falcon 1's corroded fuel-line nut on Kwajalein in 2006 to the tower catch of Ship 42 at Starbase, and set the record against Marshallese wave piloting.
“Our reviewer found a formatting tag stranded on a procedure line of a Marshallese ground-segment manual for a satellite operator.”
At 2230 GMT on 24 March 2006, the first Falcon 1 lifted off from Omelek, a 7-acre island on Kwajalein Atoll. SpaceX used Reagan Test Site range facilities for the launch. Elon Musk followed the countdown from a windowless concrete control room on Kwajalein Island, after 4 scrubs since November 2005. The payload was FalconSAT-2, a US Air Force Academy satellite flown under the DARPA and Air Force small launch vehicle programme. About 30 seconds into the climb, the Merlin engine shut down. Falcon 1 fell onto the reef less than a kilometre from its pad.
SpaceX had fewer than 200 employees in 2006. Its launch staff lived on Kuwajleen (Kwajalein) and took a daily boat out to Omelek. Staff sleeping overnight on the island earned a t-shirt modelled on the Survivor logo. In 2005 the Omelek crew went on strike after a boat with food failed to arrive. The launch director sent an Army helicopter with supplies that night. By the third flight, the crew had a kitchen and a refrigerated sea van. Some 13,500 Marshallese citizens live on the atoll, most of them on Ebeye.
DARPA investigators examined recovered parts next to a failed aluminium B-nut on the fuel pump inlet pressure transducer. Telemetry showed a fuel-line pressure below the expected figure near that nut. The investigators named inter-granular corrosion cracking as the only plausible cause. Leaking RP-1 fuel coated the engine and caught fire at ignition. Musk later said a line on the data plot rose during the static fire and stayed flat on launch day. His team knew the fault from that line. The nut had corroded in the salty, humid air.
On Flight 3 in August 2008, the spent first stage struck the second stage after separation. The sole change for Flight 4 was a longer pause between first-stage burnout and stage separation. Falcon 1 went into orbit with a 165 kg mass simulator on 28 September 2008. It was the first privately developed liquid-fuel rocket to orbit Earth. Of the 5 Falcon 1 flights from Omelek, 3 had failed. The last Falcon 1 lifted off on 14 July 2009 with RazakSAT for Malaysia, the first commercial launch in SpaceX history.
Residents of Starbase voted 212 to 6 on 3 May 2025 to incorporate their community as a Texas city. Cameron County confirmed the result on 20 May. Next to its Starfactory, SpaceX is building a Gigabay integration hall of about 700,000 square feet. The company targets completion for the end of 2026. On 12 June 2026, SpaceX listed on Nasdaq at $135 a share. The company sold 555,555,555 shares and raised $75 billion, the largest IPO on record.
Measuring 407 feet (124 metres), the V3 Starship first flew on Flight 12 on 22 May 2026 from a new pad at Starbase. On Flight 13 on 24 July, the upper stage deployed 20 Starlink V3 satellites and splashed down intact in the Indian Ocean. SpaceX spokesperson Dan Huot said it was the first intact Starship the company had put in the water. Flight 14 on 28 September 2026 sent a Starship into orbit for the first time, with 26 Starlink V3 satellites aboard.
Flight 15 lifted off on 13 November 2026. The Super Heavy booster aborted its tower catch. SpaceX lost the booster offshore in the Gulf of Mexico. On 14 November, the tower arms at Starbase caught Ship 42 after about 22.5 hours and 14 orbits. It was the first catch of a Starship upper stage. SpaceX had caught its first Super Heavy booster on Flight 5, on 13 October 2024. Starlink provides internet service in the Marshall Islands 20 years after Omelek.
Across the Marshall Islands, pilots found low atolls by sensing the refraction and reflection of the ocean swell around islands. They combined the swell with the stars and dead reckoning. Learners floated blindfolded in the ocean to memorise the sensations of waves and currents beneath them. A Penn Museum writer reported that Marshallese seafarers knew this wave behaviour better than scientific oceanographers. In the writer's account, oceanographers caught up through aerial photographs of swell action.
Stick charts gave the pilots models of the sea, built from palm strips and cowrie shells. The mattang (teaching chart) set out the general behaviour of swells around an island. The meddo (sea) showed particular islands and their swells. The Penn writer called the mattang a science model and the meddo a set of piloting instructions. Navigators memorised the charts and left them ashore. Musk's account of the flat telemetry line describes a crew comparing a flight against a memorised model of the engine.
In the 2000s, Captain Korent Joel sought help through Waan Aelõñ in Majel (Canoes of the Marshall Islands) to revive wave navigation. Joel was one of the last traditional navigators. He described the dilep (backbone), a line of swell between 2 atolls. Heaped waves along it are booj (knots). Anthropologist Joseph Genz and a team of oceanographers conducted the research with Joel in spoken Marshallese. The English version of the dilep troubled the oceanographers. To the scientists, Joel's explanation seemed to contradict basic wave dynamics.
Testing the pilots' signals in a 2009 paper, Genz and 5 co-authors used wave buoys and wave model simulations. Their data accounted for an island-finding signal. The authors linked another pattern to concepts not easily translated into oceanographic terms. In 2006, the year Falcon 1 first flew from Kwajalein, Genz tested Joel at sea. He changed the yacht's course during Joel's nap below deck and kept him away from the GPS. On waking, Joel did direct the boat back toward the islands.
Our reviewer found a formatting tag stranded on a procedure line of a Marshallese ground-segment manual for a satellite operator. The reviewer asked the translator to delete the tag and compare the line with the English source. A technician following that procedure would find the stray code in the middle of an instruction. We have also set an AI engine's Marshallese beside a reviewed human version, line against line.
Alson Kelen directs Waan Aelõñ in Majel. He said Genz brought a GPS on their voyages. By Kelen's account, the crews never needed it. Joel died in January 2017, aged 68. Marshallese navigators call a navigator's coming of age ruprup jo̧kur (breaking out of the turtle shell). Physicist John Huth has studied the physics of wave piloting alongside Kelen. Huth aims to turn that research into a layperson's manual for teaching in Marshallese schools.