brianchitester.com Posts Polymath Profiles Games RSS

Gerald Jay Sussman

Introduction

Gerald Jay Sussman (born 8 February 1947) is an American computer scientist who arrived at MIT as a freshman in 1964 and has never really left. He is the Panasonic Professor of Electrical Engineering there, a founding director of the Free Software Foundation, the co-inventor (with his student Guy Steele) of the Scheme programming language, and, with Hal Abelson and Julie Sussman, the author of Structure and Interpretation of Computer Programs, the “Wizard Book” that taught a generation of MIT students what programming was for (MIT CSAIL ). He also designed the Digital Orrery, a suitcase-sized special-purpose computer that he and the astronomer Jack Wisdom used to show that Pluto’s orbit is chaotic, and later rewrote graduate classical mechanics as a body of runnable programs. Off the clock he is a bonded locksmith, a life member of the American Watchmakers-Clockmakers Institute, an amateur telescope maker, and a ham radio operator.

The list looks scattered until you notice that everything on it is the same activity. Sussman’s career-long claim is that a precise program is an explanation: you understand something when you can build a mechanism that does it, and a computer is the strictest audience for checking that you have. He has described his own method as synthetic rather than reductive: faced with a complicated system, he asks how would I make something that does this? (AAAI oral history, MIT CSAIL ). A planetary integrator, a mechanics textbook that runs, a disassembled Rolex, and a telescope mirror cell are all answers to that question. He is a polymath of one institution and one method, which makes him an unusual case in this series: the breadth came from depth, and from staying put.

Time Management & Workflow

The record on Sussman’s daily schedule is thin. He has not written about his working hours, and no profile describes his mornings. What is documented is the larger shape of his time, which runs on the academic calendar more completely than almost anyone else’s in this series. For six decades the unit of his life has been the MIT semester, and his major projects line up with its rhythms: courses designed and redesigned over years, a sabbatical that produced a new field of work, and the January Independent Activities Period, which he still uses for things that fit nowhere else.

The habits he does describe were formed in the hacker culture of the 1960s AI Lab. As a freshman he fell in with physics students in Baker House, Alan Guth among them, and one of them showed him an underused computer on the ninth floor of Tech Square, run by Richard Greenblatt, who let him play with it (AAAI oral history ). The lab’s culture was nocturnal and collaborative, and in the same interview he recalls with obvious pleasure the late-night working sessions and dinners with brilliant colleagues. Bill Gosper, an upperclassman, taught him Lisp and, Sussman says, much of his programming style. Scheme itself came out of this mode of work: he and Steele built it while trying to understand Carl Hewitt’s actor model of computation, and discovered that what they had built to send messages to actors was the same thing as calling a function (AAAI oral history ). They wrote it up not as journal papers but as a run of AI Lab memos between 1975 and 1979, the “Lambda Papers,” with titles like LAMBDA: The Ultimate Imperative (Scheme Research ).

His working method in research is to take on problems he doesn’t yet understand. The right thing to do is always to take a hard problem that you don’t know much about, he said, defending Marvin Minsky’s habit of handing undergraduates problems that experts considered hopeless; as an undergraduate, Sussman himself was put in charge of the AI Lab’s summer computer-vision project (AAAI oral history ). It failed, as nearly everyone now knows it would, but it is the template for a career of walking into adjacent fields (astronomy, mechanics, horology) with the confidence that the programming habit would carry over.

The sabbatical is where that habit paid off most visibly. In 1983–84 he spent a year at Caltech, met Jack Wisdom, then a young dynamicist who had explained the 3:1 Kirkwood gap in the asteroid belt using a new mapping method, and decided the result needed testing by brute-force integration. No general-purpose computer could do the calculation at the precision required, so Sussman designed one that could (IEEE Computer Society ). The year away produced a collaboration with Wisdom that has now lasted four decades and given him a second career in celestial mechanics.

Daily Life Practices & Rituals

The best-documented ritual in Sussman’s life is a weekly trip to a watchmaker’s bench. A software developer who spoke with him at a conference in 2011 reported that Sussman traveled every week to the Stoneham, Massachusetts shop of one of the state’s few master watchmakers, doing work there in exchange for instruction in the craft, an apprenticeship paid in labor rather than money (Mario Aquino ). The same writer passed on two other details. Sussman had received a Rolex as a gift after his 1973 doctorate and still wore it; when he sent it to Rolex for cleaning it came back marked by their carelessness, so he took it apart and fixed their work himself. And he had no cell phone and didn’t watch television, because his free time was too valuable to spend that way. When the writer admitted to sleeping less than five hours a night, Sussman smiled and said he was winning, since it left more hours to be alive. These are secondhand reports from one conversation, but they fit everything else on the record.

The watch work is not a retirement hobby. He has belonged for decades to the American Watchmakers-Clockmakers Institute (as a life member) and the Massachusetts Watchmakers-Clockmakers Association, alongside the Amateur Telescope Makers of Boston and the American Radio Relay League, and he is a bonded locksmith (MIT CSAIL ). Radio is the oldest of these. He was a ham operator as a teenager in Brooklyn, and he remembers being drawn to an early computer partly because he had never seen a machine with 5,000 vacuum tubes in it (AAAI oral history ).

His reading habits point the same way. He says he reads little fiction, because the real world is interesting enough, and in a recent interview he was working through Carlo Rovelli’s textbook on quantum gravity at about a day per page (AAAI oral history ). There is no record of a diet, an exercise regime, a meditation practice, or any interest in optimizing his body; his energy goes into making and fixing things, and the record is silent on the rest.

One physical episode shaped his choice of work. In the lab’s robotics days he was nearly killed by an AMF industrial hydraulic arm whose position sensors were potentiometers glued in place with fast-curing epoxy. When one came loose, the arm swung violently back and forth until someone hit the stop switch, and a deadman switch was added afterward (AAAI oral history ). He soon left robotics, though he gives the reason as intellectual rather than physical: he found more to think about elsewhere.

Domains of Pursuit

Artificial intelligence and problem solving. Sussman’s way into AI came through a tic-tac-toe program. As a freshman he wired up a small neural network with random connections and set it learning; Minsky asked why it was random, and when Sussman said he wanted it to have no preconceptions, Minsky pointed out that a random net has preconceptions too, you just don’t know what they are (AAAI oral history ). The exchange became one of the canonical “AI koans” of hacker folklore, in which Minsky answers by closing his eyes, so that the room will be empty, and Sussman is enlightened (The Jargon File ). His 1973 doctoral thesis, A Computational Model of Skill Acquisition, built a program called HACKER that wrote and debugged its own plans in a world of toy blocks; its failure on one three-block stacking problem gave planning research the Sussman anomaly, still taught as the standard demonstration of why goals interfere with each other (Wikipedia ). Along the way he co-wrote Micro-Planner with Terry Winograd and Eugene Charniak, the language Winograd’s SHRDLU program ran on, and then the language CONNIVER (HOPL ; AAAI oral history ).
Programming languages. Scheme (1975) was Lisp stripped to a minimal core with lexical scoping and first-class procedures. Its design philosophy, which Sussman still quotes from the language standard, is that programming languages should be designed not by piling feature on top of feature, but by removing the weaknesses that make extra features seem necessary (IEEE Computer Society ). He and his students built Scheme chips in 1978, 1979, and 1981 (MIT CSAIL ).
Computer-aided engineering. In 1975 he hired a young Richard Stallman as a research assistant on a program to analyze electronic circuits (Free as in Freedom ). Their 1977 paper on the system introduced dependency-directed backtracking: when a chain of inferences hits a contradiction, the program traces back to the assumptions actually responsible rather than undoing its most recent guess (DSpace@MIT ). The idea runs through his later work on truth maintenance and, in 2009, on propagator networks with his student Alexey Radul: a model of computation as autonomous machines adding information to shared cells, so that partial knowledge from many sources can be combined (DSpace@MIT ).
Celestial mechanics and special-purpose hardware. The Digital Orrery, described in 1985 by Sussman and collaborators at Caltech, MIT, and Columbia, did about 10 million floating-point operations a second in roughly a cubic foot of space on 150 watts (NASA ADS/NTRS ). In 1988 he and Wisdom used it to integrate the orbits of the outer planets for 845 million years and found that Pluto’s motion is chaotic, with nearby trajectories diverging by a factor of e every 20 million years or so (DSpace@MIT ). Its successor, the Supercomputer Toolkit, built jointly with Hewlett-Packard, was a family of standard hardware and software components from which a scientist could configure a machine that reached supercomputer-class performance on one problem at a fraction of the cost of a general-purpose supercomputer (DSpace@MIT ). With it, Sussman and Wisdom integrated the entire planetary system for close to 100 million years and found the whole thing chaotic, with a divergence time of about 4 million years (DSpace@MIT ).
Physics as programs. With Wisdom and Meinhard Mayer he wrote Structure and Interpretation of Classical Mechanics, which teaches Lagrangian and Hamiltonian mechanics with every equation expressed as code that runs. The preface makes the case bluntly: in traditional notation symbols have ambiguous meanings that depend on context, and writing the methods in a programming language forces them to be unambiguous (SICM preface ). He and Wisdom teach it as a graduate course listed jointly in planetary science, electrical engineering, and physics (MIT OpenCourseWare ), and extended the approach in Functional Differential Geometry (2013, with Will Farr), which puts programs into the explanations of the mathematics behind general relativity (Schweitzer ).
Telescopes. He redesigned the mirror-support system of the 2.4-meter Hiltner telescope at the MDM Observatory on Kitt Peak, and with Julie Sussman wrote the first prototype software for the guider-probe controllers on the two 6.5-meter Magellan telescopes in Chile (MIT CSAIL ).
Horology. Sussman lectures on the mechanical watch as an electrical engineer would see it: a mainspring-driven oscillator whose balance wheel is an RLC circuit in brass. The talk explains why the escapement must deliver its push at the zero crossing of the balance’s swing, why the oscillator settles into a limit cycle of known amplitude, and how friction sets the Q and with it the timekeeping (RF Cafe ). He gave it again for MIT’s January term in 2026, with the watchmaker Jack Kurdzionak demonstrating on a real movement, followed by four weekend hands-on workshops (MIT Events ).

Employment & Economic Model

Sussman’s economic model is the simplest in this series: he has had one employer since he was seventeen. He took his S.B. (1968) and Ph.D. (1973) at MIT, both in mathematics, became an assistant professor in electrical engineering straight out of the doctorate, and has worked in MIT’s AI research since 1964 (MIT CSAIL ). Where Fuller lived precariously to keep his independence and Eno cross-subsidized his experiments with commercial production, Sussman got his independence from tenure and the institutional generosity of the old AI Lab, whose defense funding in the 1960s and 1970s let researchers follow their curiosity with unusually few deliverables.

The chair he holds is industrial money. The acknowledgments of the mechanics book thank the Matsushita Corporation for supporting him through an endowed chair (SICM acknowledgments ); Matsushita later became Panasonic, hence the title. Hardware projects brought in corporate partners on a project basis: the Supercomputer Toolkit was a joint MIT–Hewlett-Packard effort (DSpace@MIT ), and in July 1986 Abelson and Sussman recorded twenty SICP lectures for Hewlett-Packard employees, professionally produced by HP’s television unit and later released under a Creative Commons license (MIT ). The recordings have had a long second life online.

The books are the other income stream, though no figures have been published. SICP has been in print since 1985, in a second edition since 1996, and has been translated into several languages and into JavaScript (MIT CSAIL ). What makes Sussman’s economics unusual is how much of his output he gives away. He and Abelson released MIT/GNU Scheme as free software (Free as in Freedom ), the mechanics book is available free online, and he has sat on the Free Software Foundation’s board since its founding in 1985 (FSF ).

Family & Personal Relationships

Sussman is married to Julie Sussman, a computer programmer, and the marriage is also one of his longest working partnerships (Wikipedia ). She is credited on the title page of SICP (with Julie Sussman), and her role on the later books was more than polishing. In the mechanics book the authors thank her for careful reading and serious criticism that forced them to reorganize and rewrite major parts of the text (SICM acknowledgments ). The two also wrote the prototype software for the Magellan telescopes’ guider probes together (MIT CSAIL ). Little else about their life together is public. The public record does not say when they married or whether they have children.

His intellectual life has been built on a small number of very long partnerships. The first came before MIT: at Erasmus Hall High School in Brooklyn, he took a programming course offered through Columbia for high-school students, taught by Joel Moses, then a Columbia undergraduate and later an Institute Professor at MIT (MIT EECS ). In the Columbia library he found the manual for the IPL-V language and set about writing an interpreter for it while still in high school, and it was reading Minsky and McCarthy that convinced him MIT was the place to be (AAAI oral history ).

Minsky and Seymour Papert (his doctoral adviser) set his direction; Gosper set his style. Then came the collaborators of his own generation and the students who became collaborators: Guy Steele on Scheme, Richard Stallman on circuit analysis (Sussman later called him a clear thinker and a clear designer (Free as in Freedom )), Hal Abelson on SICP and the Toolkit for more than forty years, Jack Wisdom on the solar system and mechanics since 1984, Chris Hanson on MIT Scheme and Software Design for Flexibility (2021), a book about systems that can be adapted by adding code rather than changing it (MIT Press Bookstore ). He counts nearly fifty doctoral students and says the work he is proudest of is the education (not training) of students, many of whom went on to teach (IEEE Computer Society ). One of his more recent students, Leilani Gilpin, recalls a summer spent programming next to him, which is how she learned the way he thinks (MIT CSAIL Alliances ).

Philosophy of Life

The oldest statement of Sussman’s philosophy is in the preface to SICP: programs must be written for people to read, and only incidentally for machines to execute (SICP preface ). The same preface names the subject of the book as procedural epistemology, the study of the structure of knowledge from an imperative point of view, as opposed to the declarative what is of classical mathematics. That phrase is the key to the rest of his life. For Sussman a program is a way of knowing: a description of how to do something that is precise enough to be checked by running it. The mechanics book, the watch lecture, and the propagator work all apply the same epistemology to new material.

The second principle is that bugs are not failures. His thesis was about a program that debugged itself, and he has never stopped arguing that debugging is how design actually happens: bugs are necessary, he told an MIT interviewer, because finding them is how you refine what you meant (MIT CSAIL Alliances ). In the same conversation he argued for additive programming, extending a working system without editing what already works (one should not have to modify a working program), and against the industry habits that cut against it, such as rushing products out and treating programmers as interchangeable parts.

Third, he is openly unsatisfied with his own field. His 2011 Strange Loop keynote was titled We Really Don’t Know How to Compute! Its argument was that a human genome, about a gigabyte of information, builds a person from one cell, and that small edits to it yield a cow or a dog while larger ones yield a frog or a snake. Nothing engineers know how to write is that dense or that flexible, and new design principles and new linguistic support would be needed to get there (Strange Loop ). He takes from biology a lesson without imitation: mimicry is actually a bad idea, he says; birds teach us about flight, but airplanes shouldn’t flap (MIT CSAIL Alliances ).

His ethics follow from the epistemology. If a program is an explanation, then a machine that acts on the world owes the world an explanation. In a 2018 talk, “What, me worry?” or Should We Fear Intelligent Machines?, he called fears of AI as an existential threat a bit overblown and argued that the real problems are social and political. His conditions were concrete: autonomous agents must be able to explain their decisions and actions, must be accountable for them in adversarial proceedings, must be corrigible, and must run on software open to inspection by anyone (Clojure/Sync ). It is the free-software position extended from code to conduct. His definition of AI is deflationary in the same spirit: AI is the part of computer science we don’t know how to do (MIT CSAIL Alliances ), which is to say, whatever we can’t yet explain.

His advice to younger researchers is a summary of his own life: don’t chase what is rewarded; figure out what you will enjoy doing and do it! (IEEE Computer Society ).

Tools, Environment & Infrastructure

Sussman’s primary tool is a language he made for himself. Scheme is small on purpose (a few constructs and powerful means of combination and abstraction) because he wants a notation in which the ideas of a program are visible rather than hidden under conveniences. He likes to quote Alan Perlis’s line that syntactic sugar causes cancer of the semicolon (IEEE Computer Society ). MIT Scheme, maintained for decades with Chris Hanson, is the substrate under almost everything he has built since the 1980s, including Scmutils, the symbolic-algebra and numerical system that lets the mechanics book’s equations run. The acknowledgments credit Abelson, Hanson, and the engineer Dan Zuras with helping invent its organization (SICM acknowledgments ).

When his tools weren’t fast enough, he built hardware. The Digital Orrery was a parallel machine designed to do one class of calculation (orbital integration at high precision) better than any general-purpose computer of its day. It is now in the Smithsonian’s National Museum of American History, which credits Sussman as its donor and lists it at about 16 by 12 by 24 inches (Smithsonian ). The Toolkit generalized the Orrery’s idea into a kit, so that a scientist with a single hard problem could assemble a machine for it.

His physical tools are the ones of the crafts: the watchmaker’s bench, the locksmith’s tools, the telescope mirror cell. The watch lecture is in effect a demonstration of how these two toolkits meet. He uses close-up video of a working movement to map each part onto its counterpart in an electrical circuit (RF Cafe ), so that the escapement becomes something an EE student can reason about with equations they already know.

His larger environment is MIT itself, which he has inhabited the way some people inhabit a family house. He came up through the AI Group of Project MAC, a small community where, by his own account, he once helped Tom Knight and Stewart Nelson with a stunt against the time-sharing system by depositing a value into memory while the operator wasn’t looking. Knight was banished from the building for a while (AAAI oral history ). He is still in the descendant of that lab, CSAIL.

Tradeoffs & Costs

The central trade in Sussman’s life is depth of place for breadth of experience. He has spent his whole adult life at one institution, which gave him unusual freedom (to move from AI into astronomy in his late thirties, to keep teaching watch workshops at seventy-eight) at the cost of never seeing his ideas tested in a different environment. His one year away, at Caltech, produced the most consequential science of his career, which is a hint of what he gave up by not doing it more often.

The second cost is that his teaching outlived his curriculum. SICP was taught in its present form from fall 1980, to six or seven hundred students a year (SICP preface ), and remained MIT’s introductory course until 2007, when it was replaced by a Python course built around robots (Where’s Walden ). Sussman’s own explanation, given at the 2009 International Lisp Conference, was not bitter but it was pointed. In 1980, he said, good programmers spent a lot of time thinking and built systems from small parts they understood entirely; now you have to do basic science on your libraries to see how they work, which is a fundamentally different job that needed a different course. Python had been chosen, he guessed, because it probably already had a library for the robots (Where’s Walden ). The style of thinking he spent his career teaching, understanding every part, had become a minority practice in the industry it trained people for.

Much of his technical work has had the same fate: influential among specialists, little known outside. Dependency-directed backtracking, truth maintenance, and propagators shaped AI research and constraint solving without becoming household tools; Scheme became the language of programming-language research and of a famous textbook rather than of industry. His choice to publish in AI memos and give software away traded commercial reach for clarity and openness, and he has never suggested he regrets it.

There is one reputational cost of a different kind. In 2011 Sussman attended the “Mindshift” conference in the U.S. Virgin Islands, hosted by Jeffrey Epstein and Al Seckel; he has said he learned of the abuse allegations against Epstein only afterward, and he declined to comment publicly on his relationship with Seckel (Wikipedia ).

Legacy & Influence

Sussman’s most widespread legacy is a way of teaching programming as a liberal art. SICP’s argument, that computer science is about managing intellectual complexity through abstraction, and that the deepest program a student can write is an interpreter for the language they’re writing in, has outlasted MIT’s use of the book. It has been translated into several languages and rewritten in JavaScript, and its teaching approach won him the ACM’s Karl Karlstrom Outstanding Educator Award (1990), MIT’s Amar G. Bose teaching award (1992), the IEEE Educational Activities Board’s Major Education Innovation Award (2023), and the IEEE Computer Society’s Taylor L. Booth Education Award (2024) (MIT CSAIL ). The 1986 lectures continue to find new audiences online.

His second legacy is Scheme and the line of ideas that came through it: a deliberately small language that became a standard vehicle for programming-language research and teaching, as well as for SICP itself. His third is in planetary science: the Pluto result and the chaotic-solar-system integrations changed how astronomers think about the long-term stability of the planets, and the Digital Orrery stands as an early proof that a small team could build a special-purpose computer to answer one scientific question.

His fourth legacy is an argument still in progress, that notation is a tool for thinking, and that writing physics, mathematics, and engineering as executable programs exposes confusions that traditional notation hides. Structure and Interpretation of Classical Mechanics and Functional Differential Geometry are the demonstrations. The free-software movement, which he has backed from the FSF board for four decades, is the institutional side of the same idea: explanations should be open.

In January 2026, at seventy-eight, he was back in a lecture hall on Vassar Street explaining a balance wheel as a circuit, with a watchmaker beside him holding the movement, before a weekend of workshops at the bench (MIT Events ).

Sources: