On this page
When Apollo 11 was a few minutes from landing on the Moon, the guidance computer started flashing an alarm nobody in the lunar module had seen in training. The landing went ahead, and the reason it could is a piece of software design led by Margaret Hamilton. This post sets out what she did, using MIT, NASA and Computer History Museum records, works through the physics of why the computer had to decide for itself, and ends with habits a student can borrow.
Who was Margaret Hamilton?
According to her MIT News obituary, Hamilton was born in Paoli, Indiana, in 1936 and earned a BA in mathematics from Earlham College in 1958. At MIT she first wrote weather prediction software for the meteorologist Edward Lorenz, then worked on the SAGE air defence project at MIT Lincoln Laboratory from 1961. In 1965 she joined MIT's Instrumentation Lab, which had won the NASA contract for Apollo's on-board flight software, as the first programmer on the Apollo project there. By 1968 more than 400 people were working on Apollo's software. She died on 30 September 2026, aged 90, as MIT News reported on 7 October 2026.
The Computer History Museum records that she led the team that created the on-board flight software for the Apollo command modules and lunar modules, and directed the Software Engineering Division at the Instrumentation Laboratory. In 2016 she received the Presidential Medal of Freedom; the White House announcement credits her with concepts of "asynchronous software, priority scheduling and priority displays, and human-in-the-loop decision capability".
What was the "Lauren error"?
MIT News tells a story that shows how she thought. Her four-year-old daughter Lauren, playing with the command module simulator, started a pre-launch program called P01 while the simulator was in mid-flight, and crashed it. Hamilton proposed a software fix. She was overruled, because trained astronauts would never make that mistake. During Apollo 8 in 1968, Jim Lovell started P01 by accident and the navigation data vanished. After that, her changes went into the program. MIT News calls this an early example of defensive programming: software that expects errors and handles them.
What happened during the Apollo 11 landing?
NASA's Apollo Lunar Surface Journal (archived copy) gives the transcript. At mission time 102:38:26 Neil Armstrong reported "Program Alarm", and four seconds later, "It's a 1202." The journal explains that the alarm came from data overflow: an unexpected flow of rendezvous radar data. The computer had been programmed to treat that data as secondary and ignore it while it did more important calculations. At 102:38:53 Charlie Duke in Houston replied, "We're Go on that alarm."
A 2003 NASA press release (archived on the same journal) adds that about three minutes before touchdown the software overrode a command to give priority to a radar system whose switch had been turned on because of a faulty written procedure. "The action by the software permitted the mission to safely continue."
Why couldn't Houston just take over? A worked example
Radio waves travel at the speed of light, c = 3.00 × 10^8 m/s. NASA gives the average Earth-Moon distance as 384,400 km.
Convert the distance: d = 384,400 km = 3.844 × 10^8 m.
One-way delay: t = d / c = 3.844 × 10^8 / 3.00 × 10^8 = 1.28 s.
Round trip, for a question from the Moon and an answer from Earth: 2 × 1.28 = 2.56 s.
From the transcript, Houston's "Go" came 27 seconds after Armstrong first reported the alarm (102:38:26 to 102:38:53). Only about 2.6 s of that was radio travel; the other 24 s or so were people checking what the alarm meant.
During all of that time the lander was descending and the computer had to keep steering it. Even with no thinking time at all, a round trip of 2.56 s is far too slow for a control loop. The decision to shed low-priority work had to happen on board.
Common mistakes in this kind of question:
Forgetting the round trip. A reply needs the signal to go and come back, so double the one-way time.
Mixing units. Kilometres divided by metres per second gives nonsense; convert first.
Treating 384,400 km as exact. It is an average; the Moon's distance changes over its orbit, so give answers to two or three significant figures.
The same reasoning about signal speed and distance sits behind our post on Katherine Johnson's orbital calculations.
What can a student take from her way of working?
Plan for the mistake you were told can't happen. Hamilton's P01 fix was rejected because astronauts were too well trained; Apollo 8 proved otherwise. In your own work, check the "silly" errors, such as sign conventions and units, before the clever ones.
Rank your work by priority when time runs short. In a three-hour paper, leaving a hard low-mark question to secure the questions you can do is the Apollo 11 computer's logic applied to you.
Write things down for the next person. The 1202 alarm was cleared in Houston in under half a minute partly because, after a training simulation, flight controllers had been told to go back and study such alarms.
Software that tells a person what matters most at the moment is also an idea behind the tools you study with. If a chapter overwhelms you, Fermi's AI tutor can work out what you already know and build a plan for the session, so you start with a clear next step. For another computing pioneer who cared about making machines easier for people to use, read about Grace Hopper and the first compilers.
Try it now: work out the round-trip signal delay to Mars at its closest, using a distance you look up yourself.
