Imaginary interview with Margaret Hamilton
by Charactorium · Margaret Hamilton (1936 — ?) · Technology · Sciences · 6 min read
Cambridge, Massachusetts, a winter evening in the late sixties. The neon lights of the MIT Instrumentation Laboratory hum over desks strewn with annotated listings. Margaret Hamilton, cold coffee within reach and her daughter Lauren asleep in a nearby armchair, agrees to set aside the stacks of code for a moment to tell how you teach a program never to panic.
—How would you describe the machine you were writing for, to someone who has never seen a computer from that era?
Imagine a 32-kilo block with only 4 KB of RAM, in which you had to fit navigation, lunar landing, emergencies, and the return. The program wasn't loaded like today: it was woven. Women workers passed a copper wire through a tiny magnetic ring for a 1, beside it for a 0 — the core rope memory. Once sewn, the software became physical, etched into matter, impossible to correct without reweaving everything. It forces a particular humility: every line you validate goes to live 380,000 kilometers away, and you can never touch it again. You didn't debug in flight. You debugged before, obsessively, thinking of the failure no one had yet imagined.
The program wasn't loaded, it was woven: a copper wire for a 1, beside it for a 0.
—With so little memory, how do you guarantee that a machine reacts in time during a flight maneuver?
Real time is a cruel master. The Apollo Guidance Computer had to respond in milliseconds while a module descended toward the Moon — there is no second chance, no engineer on the ground hitting pause. So we built the entire system around a simple-sounding but hard-to-achieve idea: fault tolerance. The software had to continue functioning correctly even when the hardware failed, even when asked the impossible. That meant planning not for the nominal scenario in the manuals, but for dozens of catastrophic scenarios everyone deemed unlikely. I spent my evenings at the Cambridge lab precisely looking for those holes, the failures we hadn't yet covered. A space mission does not forgive optimism.
A space mission does not forgive optimism.
—It's said your daughter Lauren was behind an important discovery. What happened?
I used to bring her to the lab nights and weekends — she played while I ran endless test sequences. One evening, Lauren, who must have been four, started pressing keys on the simulator console and accidentally triggered a program that had no business being there during flight. Everything collapsed: the navigation was wiped. I thought immediately: if a child can do it, an exhausted astronaut will too. I wanted to add a safeguard. NASA refused — their men were professionals, they would never make such a blunder. I slipped a note into the documentation anyway. During Apollo 8, in 1968, Jim Lovell inadvertently selected exactly that program and wiped his navigation data. The fix I had prepared anyway allowed the mission to recover.
If a child can do it, an exhausted astronaut will too.
—What do you take away from that disagreement with NASA about human error?
That overconfidence is itself a bug. I was told the pilots were too well trained to make mistakes, and behind that phrase I heard a dangerous certainty. An engineer does not design for the ideal human; they design for the human at three in the morning, under adrenaline, hundreds of thousands of kilometers from home. The Lauren episode served as my argument for years: it wasn't an astronaut's fault, it was a design fault if the system allowed such an error to be fatal. After Apollo 8, no one asked me why I wasted time on 'impossible' cases. Error detection and recovery became a requirement, not a whim. The machine must assume the human will err, and catch them before the fall.
Overconfidence is itself a bug.
—Let's go back to July 1969. Three minutes from the lunar surface, the alarms go off. What do you feel at that moment?
The 1202 code. Executive overflow — the AGC's processor was receiving more tasks than it could handle in the allotted time. At that exact moment, the entire edifice we had woven for years was staking its reputation in seconds. But I wasn't afraid of the 1202, and that's the whole point of our work: the system was designed for this. Instead of giving up, it automatically shed secondary tasks and focused its meager resources on the essential — landing the module. Priority scheduling did exactly what we had imagined it for. Armstrong continued his descent. The software didn't crash; it decided, on its own, what should live and what could wait. That's the moment software engineering stopped being a promise and became proof.
The software didn't crash; it decided, on its own, what should live and what could wait.
—How can a program make such a critical decision on its own, without ground intervention?
Because we accepted an uncomfortable truth early on: during descent, the ground can do nothing. The delay, the radio silence, the urgency — there is no time to ask permission from Houston. I wrote about this in our reports from the Draper Laboratory, as early as the early seventies: the system detects errors and recovers in real time, ensuring critical tasks always come first. It all comes down to hierarchy. You number each task by importance, and you give the system the right to sacrifice the secondary to save the essential. The 1202 was not a failure; it was the software doing its job: recognizing its own overload and choosing. Designing a machine capable of intelligent sacrifice — that's what separates a program from an automaton.
Designing a machine capable of intelligent sacrifice — that's what separates a program from an automaton.
—You are credited with coining the term 'software engineering'. Why did you feel the need to invent a word?
Because without the word, you don't exist. In the sixties, writing software was seen as tinkering, a second-class activity next to real engineering — rockets and engines. When I told my aeronautical colleagues that my team was doing software engineering, some laughed outright. The term sounded pretentious. But that was deliberate: I wanted our programs to be held to the same reliability standards as a load-bearing structure, because our code, too, carried human lives. Putting a name to a thing gives it a place in the world, demands that it be taken seriously. Decades later, when I received the Presidential Medal of Freedom in 2016, that once-mocked term had become the name of a global discipline. Skepticism had switched sides.
Putting a name to a thing gives it a place in the world.
—What, in your view, distinguishes an engineer from a mere programmer?
Responsibility, and the documentation that follows from it. Anyone can make a program work on a lucky day; the engineer guarantees it will work on the worst day, in the worst conditions, with no one to catch it. That's why I imposed testing and documentation standards so strict they annoyed people. Every assumption had to be written down, every failure case traced, every emergency procedure validated before entering the astronauts' manual. Later, when I founded Higher Order Software in 1976, I pushed this idea to its limit: prevent error at the design stage, through mathematical foundations, rather than hunting it afterward. The dream of software that is correct by construction. Engineering is not the art of fixing bugs; it's the art of building a system where certain errors simply cannot arise.
Engineering is not the art of fixing bugs; it's the art of building a system where certain errors cannot arise.
—You led a largely female team in a male-dominated field. How did you experience that singularity?
I was often photographed in a printed dress, amid a sea of white shirts, in front of the consoles at the MIT Instrumentation Laboratory. The contrast amused journalists; to me, it was just my office. I had a team where women were numerous, at a time when we weren't expected there. The difficulty was not so much the scrutiny as the constant need to prove myself. Faced with skeptical aeronautical engineers, I couldn't just claim my software was reliable; I had to demonstrate it, figure by figure, test after test. Perhaps that's what saved us. That demand to always be right under criticism forged a rigor from which the entire program benefited. They looked down on us; we responded by letting no flaw through.
They looked down on us; we responded by letting no flaw through.
—That famous photo of you next to a stack of listings as tall as you — what does it really represent?
Those printed code listings, stacked up to my shoulder, are all of Apollo's software made visible, tangible. Back then, we didn't have screens everywhere; we read code on paper, annotated it by hand, line by line, in the lab at night. That stack is not a trophy; it's the sum of thousands of hours of manual verification, of doubts set down in black and white. People like to say that no critical bug occurred in flight across the program's seventeen missions. That silence, that absence of accident, is invisible — so the photo finally gives it a body. When the image went viral years later, people who didn't know my name suddenly understood that going to the Moon had also been a matter of patience, paper, and stubborn proofreading.
That stack is not a trophy; it's the sum of thousands of hours of doubts set down in black and white.
This imaginary interview was generated by artificial intelligence from sources documented in Margaret Hamilton's profile. It dramatises what the figure might have said based on what we know about them, but does not constitute attested historical testimony. For primary sources and factual documentation, refer to the full profile.