What do firearms and cameras have in common? One answer is shared vocabulary: load, aim, shoot. The etymological origins of this relatedness are murky. People likely borrowed language about firearms — which are over 500 years older than photography — to talk about cameras, which are operated in a similar manner.  But in the case of the movie camera, the connection is concrete.

The first movie camera is generally considered to be the kinetograph, invented in 1891 by Thomas Edison and William Kennedy Dickson, an inventor at the Edison lab. Two years earlier, Edison had travelled to Paris to the Exposition Universelle of 1889, a world’s fair, to showcase his company’s phonograph. There, he met with Étienne-Jules Marey, a French scientist who had begun his career studying blood circulation, and who had developed an interest in chronophotography partly out of his work on physiology. Chronophotography consists of taking multiple photographs in quick succession to capture movement. 

A famous early achievement of the technique is The Horse in Motion, a series of photographs taken by Eadweard Muybridge in 1878 that proved something Marey had asserted years earlier: For a brief moment in gallop, a horse has all four hooves off the ground. Encouraged by Muybridge, Marey developed an improved device to take chronophotographs. The resulting design, the “photographic gun,” featured a long barrel derived from existing firearms. We don’t know what transpired between Marey and Edison, but it seems likely that Marey’s design inspired the kinetograph. Coincidentally (or not), the Online Etymology Dictionary’s entry for “shoot” claims “the meaning ‘to photograph’ (especially a movie) is from 1890” — around the time of the kinetograph’s invention.

Marey himself had been inspired by the invention of another Frenchman, the astronomer Pierre Janssen. In 1874, hoping to capture images of the transit of Venus, Janssen created a device he called the “photographic revolver.” It looked like a large telescope. Inside, it hosted a complex camera that used a revolving apparatus to take several images in quick succession. The resulting eight seconds of the black dot of Venus moving across the sun disk are often recognized as the first motion picture in history. ¹

Prior to Jansenn’s invention, revolving components had not been used in photography or astronomy. They came from the Colt Paterson revolver, a handgun popular in the U.S. since its mid-19th century creation by Samuel Colt. Thus, we can reconstruct a genealogical succession of ideas: Colt inspired Janssen, who inspired Marey, who likely inspired Edison (who then inspired the Lumière brothers and further developments in filmmaking). 

We can go further back still. Colt improved the revolver and made its mass production commercially viable, but he did not invent it. Nor was it a particularly recent development: Guns with a revolving barrel have existed since the 16th century. A German revolver dates from 1597, but there are even older specimens. The xun lei chong, a Chinese revolving musket, also dates from this period.

It makes sense: The 1500s were a time of great innovation in firing mechanisms for muskets and arquebuses, with developments like the wheellock (which used a spinning metal wheel against pyrite to create sparks), snaplock (which struck flint against steel using a spring-loaded arm), and flintlock (which improved on the snaplock with a combined hammer-flint holder). Volley guns made of multiple cannons existed in England since at least the 1330s. These were descendants, in turn, of simpler cannons that stem from the invention of gunpowder in China in the 800s. It stands to reason that eventually someone in Europe would think of making a spinning mechanism to allow a weapon to fire multiple shots in quick succession. 

And yet, if you’re anything like me, most of what you just read was probably surprising. We don’t typically associate revolvers with the 1500s, or, for that matter, with movie cameras. 

The reason I learned about this — and dozens of anecdotes around other inventions — is that I have been working on a quixotic project to place all major technological innovations in history on a timeline, together with the connections between them. The goal is to situate stories like that of the revolver and camera in their historical context, notice patterns, and understand the logic of the history of technology. The result is an interactive visualization that I call the historical tech tree. 

The tech tree can be traced to the 1980 board game Civilization, widely popularized by game designer Sid Meier’s video game version in 1991. The tech tree in Civilization functions as a branching pathway of scientific discovery that unlocks new units, buildings, and abilities — from primitive tools to space-age technologies. The story goes that Meier threw the tree together on only cursory research, with the intention to eventually come back and make the “real tech tree” later. But this initial version was so fun that it stuck. 

Sid Meier’s Civilization and its sequels went on to sell over 70 million copies, entrenching the tech tree as a defining feature of 4X strategy games. Their validity is also a staple of game criticism in blog posts, Reddit threads, and even academic papers. Mention anywhere online that you consider Civilization to be educational, and you are certain to attract comments on how the game promotes a problematic model of history — not least because tech trees reinforce the idea that technological progress is linear, deterministic, and teleological. 

This criticism is valid. But it applies mostly to design decisions needed to make a functioning game. On a much more expansive tree, one meant to represent an accurate history of technological development, rather than give players interesting choices, we aren’t so constrained. 

This is because technologies really do follow a tree structure — or, to be more mathematically precise, a directed acyclic graph. 

Inventions don’t spontaneously spring from the head of a genius like Thomas Edison in complete form. They always descend from something else. Sometimes that’s an aspect of nature (legend has it that Cai Lun, a court official in the third century Eastern Han Dynasty, invented paper after watching wasps build their nest). More often, technologies come from other technologies: earlier versions, prerequisites, or components that can be combined into something new. 

At the time of this writing, the historical tech tree contains over 1,550 technologies, from stone tools to robotaxis, and more than 1,700 connections between them. This is a good start, but only a start. Many technologies, and particularly the ways they influenced each other, still need to be added. 

For this initial version, I relied primarily on Wikipedia, where contributors have collated immense data about inventions and discoveries. ² When claims in Wikipedia seemed doubtful or unsupported, I consulted scholarly articles and books. Eventually, every piece of data in the tree will have a reference to a primary source. 

Though I used plenty of LLM automation to create the interactive visualization, I decided against mining Wikipedia or other digital sources to find the technologies and connections. Mostly, this decision is because I had to build my own mental model of technology history first, instead of automating a visual from online databases. So, I explored rabbit hole after rabbit hole, recording every case of an invention described as an improvement, an evolution, an application, or a consequence of an earlier one. In the process, I made three important design choices for my database.

What actually constitutes a technology? Spears, clocks, and light bulbs seem obvious, but what about democracy, or the concept of infinity, or rituals? The definition I landed on is “a piece of knowledge created intentionally by humans for a practical purpose (not for its own sake) and is implemented in some kind of physical substrate.” 

Like most definitions, it is imperfect: Each part is contradicted by certain things that are widely considered technologies. Some inventions were serendipitous rather than intentional, like rubber vulcanization, a process that hardens rubbers, and Czochralski crystal growth, used to build semiconductors. Some inventions blur the line between practical and intrinsically valued, like musical instruments and game consoles. Some belong more to the realm of ideas than of physics, like scientific discoveries, laboratory techniques, or software. Some primitive technologies are also created by non-human animals, while future ones may come from artificial intelligence. But overall, this definition is useful and justifies why works of art, games, sports, biologically evolved traits, philosophical and theological ideas, political events, and business models are not in the tree. ³ It matters that this is a tech tree, not an “everything tree” — otherwise the project would become intractable. 

Innovation is always a process, but a constraint of putting technologies on a timeline is that they must be reduced to discrete events. This step turned out to be less of an issue than I anticipated. We are already used to thinking about innovation in terms of specific “inventions” and “discoveries.”. Even though there were many precursors and prototypes to the light bulb, we generally date its invention to 1879, when Edison and Joseph Swan made the first practical versions.

Still, what level of discretization should be used? Too zoomed out, a tree won’t be useful. Too zoomed in, and it becomes impossible to complete, its core information drowned in minutiae. The heuristic, here also, is to use Wikipedia: If a technology has its own Wikipedia page, then it probably deserves to be a node in the tree. 

The invention also has to be sufficiently innovative. Something is an invention only if it allows some new capability rather than tweaking an existing design. This approach comes with limitations. The tree doesn’t tell us much about tech diffusion (like the spread of the compass in the Middle East and Europe after it appeared in China) or incremental improvements (like new car models after the Benz Patent-Motorwagen in 1885) — except where the diffused or improved versions can themselves be considered new technologies (like the dry compass in 13th-century Europe or the hybrid electric car in 1900). 

The other big constraint of the timeline format is that everything must be assigned a date. This is not always straightforward. 

Our knowledge of prehistoric, ancient, and medieval technology relies on scant archeological and documentary evidence. By necessity, the tree places inventions like the bow and arrow, the codex book format, or the horse collar at the earliest date we have found them, with the understanding that this is an approximation that could require a later update.

For more recent tech, a choice must be made between unimplemented ideas, prototypes, failed commercialization attempts, or patents. By default, the preferred date is the first practical version of the invention, which is not always easy to determine. Were light-emitting diodes invented in 1927 (first working version by Soviet scientist Oleg Losev), 1962 (reinvented by Texas Instruments for infrared and General Electric for visible red light), or 1968 (first practical displays by Hewlett-Packard)? 

There is also the common problem of independent inventions. To reduce information overload, the tree only shows the first invention, except when multiple versions of an invention had different consequences on the history — as in the case of the three or four independent inventions of writing.

What does all this work give us? First, a new way of noticing. The tree helps us identify unforeseen connections and interesting patterns, like the revolver and camera. It also gives us a general understanding of how technologies arise, often much earlier or later than we’d imagine.

For example, when were drones invented? I would have naively guessed sometime in the last few decades. If we’re talking about the familiar quadcopter camera drone, this is correct: it was first commercialized in 1999. But unmanned aerial vehicles are far older, dating to only a few years after the invention of powered flight by the Wright brothers in 1903. During World War I, the British and Americans both experimented with radio-based remote control — which had been prototyped by Leonardo Torres Quevedo in Spain and Nikola Tesla in the U.S. using recently discovered principles of radio — to fly unmanned airplanes. The word “drone” itself, originally meaning worker bee, comes from a particular unmanned aircraft flown for target practice in 1930s Britain, called the “Queen Bee.” 

Meanwhile, hourglasses are much younger than I would have thought. They are similar in principle to water clocks, known from ancient Egypt at least 3,500 years ago. Hourglasses are made of blown glass, a technology developed in the Middle East around the dawn of the Roman Empire, 2,000 years ago. Yet, as far as anyone can tell, they date from the 14th century, with the earliest evidence being an Italian painting from 1338. This is a few decades after the invention of fully mechanical clocks in Europe. 

We can keep going. Did the bicycle, automobile, or motorcycle come first? Trick question: They are all from the same year, 1885. To be fair, I’m cheating a bit with discretization here: I’m referring specifically to the modern design of bicycles, the “safety bicycle.” With a more expansive definition, cycle-based vehicles (like the draisine and the penny-farthing) go back decades earlier. ⁴ Indeed, the bicycle craze and the myriad tricycles and quadricycles that came out of it enabled the first motorcycle automobile, once inventors made versions powered with internal combustion engines. ⁵  

Bicycles are at the root of most modern transportation, including even the airplane. The Wright brothers were heavily influenced by their experience in the bicycle retail and repair business when they invented the Wright Flyer, which used a chain drive borrowed from bicycles to power the propellers. 

Such stories of technologies leading to others in an unexpected way are perhaps the most interesting bits in the tree. For example, the invention of Scotch tape in 1930 inadvertently enabled the discovery (and extraction) of graphene, a two-dimensional lattice of carbon atoms known for its strength and electrical conductivity, in 2004. 

Silk, produced since at least 6500 B.C. in China, reached Europe in the Middle Ages and became a major industry around Lyon, France. There, a succession of inventors made strides in automating silk weaving during the 18th century. One of the key inventions, the punch card, brought together the technologies of paper and musical organs: The punch card allowed looms to be programmed to weave specific patterns in silk. Around 1800, this led to the development of the Jacquard loom, which later inspired Charles Babbage’s difference and analytical engines, early designs for a mechanical calculator and computer, respectively. (Through another branch of the tree they also inspired floating-point arithmetic, also developed by Torres Quevedo.) Herman Hollerith’s tabulating machines, which processed punch card data, were used in the 1890 U.S. census. Hollerith’s company was an ancestor of IBM. From there, the seeds of computers in the 20th century were sown. 

Meanwhile, a parallel thread starts from silk and runs through artificial silk, also called rayon or viscose. Its manufacture is very similar to — and led to — the invention of cellophane. And from there to Scotch tape and graphene. 

Inventions in one field creating an unexpected, new field is a common pattern. Weather forecasts weren’t possible until the invention of the telegraph. Before that, communications — by sea or land — were too slow compared to the movements of air masses in the atmosphere. 

The development of meteorology also owes much to balloon flight, which developed suddenly in a “balloon craze” after hot-air balloons were first demonstrated near Paris in 1783. Within months, aeronauts filled balloons with lighter-than-air gases; this in turn inspired Jean-Pierre Minckelers in Leuven to investigate flammable gases, invent coal gasification, and prototype gas lighting in 1785. After a few more innovations (though whether they can be attributed to Scottish inventor William Murdoch or French engineer Philippe Lebon is debated), public gas lamps and a thriving “town gas” industry was born in England. (Natural gas, which replaced town gas in the 20th century, is “natural” because it does not come from coal gasification.)

And speaking of gas lighting, I would be remiss to not mention the Geissler tube, a curiosity from the 1850s consisting of a gas-filled piece of blown glass that lit up in interesting colors. It left a prodigious legacy, including fluorescent lighting, neon lighting, the discovery of cathode rays (which turned out to be electrons), x-rays, cathode-ray tube TVs, and last but not least, the humble thermionic diode, the device whose invention in 1904 is at the root of the entire electronic industry, including, when combined with the mechanical machines mentioned above, all but the earliest computers. 

A historical tech tree is a good source of anecdotes. But you can get anecdotes anywhere: books, blog posts, documentaries, Wikipedia. Is there a particular reason to arrange these anecdotes in a big interactive diagram that takes a while to load? Yes, and it has to do with the structure of technological history itself. 

One of the influences for this project is the 1970s British TV show “Connections,” presented by the science historian James Burke. The book version (also called Connections) has proved a valuable resource in finding nodes and links for the tree. Yet one gets the subtle sense that Burke was constantly being limited by his chosen format. The story he tells constantly wants to swerve suddenly, to jump ahead and back, to branch out into ten different plotlines after a single significant event. Some inventions and discoveries pop up again and again, like the astrolabe or the vacuum, since they play a role in multiple intersecting narratives. 

Even my short anecdotes need contain many parenthetical statements (including some that my editors made me delete). It’s hard to resist the temptation to veer into tangent after tangent. Start writing about how blue LEDs were invented by Shuji Nakamura in 1993, completing the color spectrum and enabling LED lamps in the early aughts — and suddenly you find yourself on the other side of an em-dash, pointing out that, by the way, research on blue LEDs also led to the first blue lasers, and by extension the Blu-ray format for optical storage, since the shorter wavelength of blue light allows higher data density than CDs and DVDs. 

These tangents are tech history wanting to escape a linear narrative. The story of technological progress calls for something grander, more encompassing, more highly dimensional. 

James Burke himself seems to have recognized that the narrative format of “Corrections” could not quite capture the history of science and technology to his full satisfaction. In the 2000s, he launched the Knowledge Web, a project not dissimilar to the historical tech tree: a vast graph showing the connections between all manners of people and knowledge. Its current fate is unclear. There is a website, including narrative descriptions of paths from Mozart to the helicopter, or from cornflakes to communism, but the app itself seems to have become unavailable. Only fragments of the Knowledge Web remain, viewable in decade-old YouTube videos. 

The historical tech tree is in many ways a successor to Burke’s project. It avoids, I hope, some of the Knowledge Web’s pitfalls. By ignoring non-technologies, it is less ambitious; by incorporating a timeline, it is more structured. But like the Knowledge Web, the tree manages to show the non-linearity of technological history and gives a view of the data that is distinct from a collection of individual stories. It provides a different model of innovation.

Building such a model is valuable. Technological history has arguably been neglected in popular perception, compared to political and military history. As historian Stephen Davies claims in an article for Works in Progress, we tend to recognize the dates of important battles, wars, and revolutions much more readily than the dates of landmark inventions and discoveries, even though the latter arguably play larger roles in our lives. Davies speculates that this neglect causes us to pay less attention to technological solutions to our problems, and therefore, contributes to the relative slowness of current technological progress, what has been called the “Great Stagnation.” 

Relatedly, the tree can help manage the burden of complexity. In the 1970s, Burke was already concerned that our lives depend on technological systems that very few people deeply understand. It is, of course, possible to live without comprehending how computers, money, or airplanes work. But when everything around us feels vaguely magical, reliant on experts whose actions we have no way of verifying, it’s easy to lose trust in technological solutions to our current problems. To solve obesity or climate change, many people are instinctively drawn to lifestyle interventions or ending capitalism rather than GLP-1 receptor agonists or geoengineering. There are many reasons for this, but I think it’s at least in part because few of us have a good grasp of the mechanics of technological progress, including its tendency to surprise us with unforeseen solutions. The complexity of technological development also makes it easier for harmful fears to take root, as in the case of the anti-vaccine movement — which slows down the search for new solutions and limits the reach of the existing ones.

I’m confident the tree can reduce complexity, because I felt that myself. I’m a history nerd (as evidenced by the fact that I compile lists of historical inventions for fun), and I thought I had a good handle on the story of human progress. But this project made me realize how little I knew about the objects around me. I didn’t really know that “electronics” meant controlling the flow of electrons with vacuum tubes or semiconductors, or that refining petroleum into kerosene uses fractional distillation, or that WiFi and bluetooth are just the use of certain radio frequencies that can be detected by a specific kind of chip. I haven’t become an expert in any of these things. But I now feel I have a reasonable understanding of how the modern world works, and more importantly, how much of it is due to the tinkering of inventors and scientists. 

Can the model of technological history embedded in the tree help actually speed up innovation, as progress studies people like Davies might hope for? Probably not directly. We can’t extrapolate from the tree to determine what the next big technology will be, in the same way that understanding political history doesn’t necessarily help predict current events. But studying history puts the present into perspective. The same is true of studying technology. Its builders, like its critics, benefit from having a general understanding of the story so far. 

The true motivations behind the tech tree are not that different from those behind playing Civilization: The tree is fun, and in a way, beautiful. There is a kind of awe in seeing links and patterns emerge. This is probably what drove Burke to work on "Connections” and the Knowledge Web. 

In the realm of history visuals, I’m particularly fond of Emma Willard’s “Temple of Time,” a crazy, 3D, classical-temple-shaped diagram that attempts to summarize all of history from creation (4004 BC, around the invention of the wheel, plumbing, and metal mirrors) to 1846, when Willard drew it. 

The caption at the bottom of the Temple suggests that it “saves great labor of thought, and may suggest new ideas, even to the learned.” I’m sure it does, but mostly, the Temple is just cool. It’s beautiful, and like a real temple, serves as a monument to the people and events that made the world what it is. 

The historical tech tree aims to be such a monument. It celebrates the creativity and resourcefulness of humans across history. I think it comes at a particularly interesting inflection point, too: We may be on the verge of automating further technological development, thanks to AI. If that’s the case, then the tech tree takes extra significance, as a record of everything that had to be invented and discovered before a new tree branches out from it — perhaps one whose connections are going to be more opaque to us. 

In Civilization, completing the tech tree usually means we’re nearing the end of a game. There’s no reason to think we’re anywhere close to completing the historical tech tree, but maybe it’s meaningful that I made it now, just as AI is becoming good enough to code most of it on its own.