Everything in the archive on one continuous line, oldest first. Narrow it by period, platform, company,
region or category, or search for a word that appears anywhere in an entry.
Physicist Edward Condon and colleagues at Westinghouse built a relay-driven cabinet that played the counting game Nim against visitors to the New York World's Fair, signaling its moves with light bulbs rather than any kind of screen.
Long before anyone spoke of computer games, a large manufacturer decided that a game was the clearest way to show the public what automatic machinery could do. That instinct, using play to make an unfamiliar technology legible, recurs in nearly every later demonstration of computing, from exhibition halls to chess matches carried on television.
Two engineers at a television manufacturer applied for a patent on a missile game played with knobs against a cathode-ray tube, with targets printed on a plastic overlay taped to the glass. Nothing was ever manufactured.
Here the idea of using a television tube for play is written down, dated and witnessed, twenty years before anyone built a product around it. The patent also shows how much early game hardware came out of wartime electronics, since the mechanism is essentially a radar display with the targets drawn on plastic.
Josef Kates built a tic-tac-toe machine taller than the people playing it for the Canadian National Exhibition, using a vacuum tube of his own design. Visitors queued to take a turn, and it was scrapped when the fair closed.
Among the earliest machines built so that members of the public could play a game against electronics, and one of the first to treat difficulty as something an operator could tune to the person in front of it. It also shows how quickly the exhibition hall became the standard place to explain computing to an audience.
Ferranti's NIMROD, a cabinet of lights built to play Nim, went on show at the Festival of Britain as a lesson in how a computer follows rules. Visitors were mostly interested in beating it.
One of the first computers built to be played with rather than used, and an unusually clear case of a company discovering that people will engage with a machine through competition when they will not engage with an explanation. It is often listed among the earliest computer games, with the caveat that its makers did not think of it that way.
Arthur Samuel's checkers program for the IBM 701 improved by playing against itself, and a televised demonstration introduced American viewers to the idea of a machine that got better at a game over time.
This is where machine learning acquired a public face. A program that improved through experience, shown to a television audience, made an abstract research idea concrete, and its use of self-play as a training method returned decades later at the center of the most capable game-playing systems ever built.
Staff at Brookhaven National Laboratory built a tennis game on an oscilloscope for the laboratory's public open house, with a net, a ball that arced under simulated gravity, and a control box for each of the two players.
Earlier machines played games of logic. This one simulated physical behavior and asked two people to compete in real time, with enjoyment as the entire purpose and no educational or promotional argument attached. That makes it the closest thing in the 1950s to a game built simply because it would be fun.
Ralph Baer, a division engineer at a New Hampshire defense contractor, wrote a short proposal for games played on ordinary television receivers. His management funded the work quietly, and the first home console followed.
The home console begins as a memo. Its premise, that the screen most people already owned was the platform, separated Baer's approach from every laboratory game before it and from arcade machines, which brought their own hardware. That premise governed console design for the next forty years.
A programmer at UCLA typed the first characters sent between two host computers on ARPANET, addressed to a machine at Stanford Research Institute. The connection failed after two letters, and the network it began carried everything afterward.
Online play needed a network before it needed anything else. This is the first working demonstration of general-purpose communication between distant host computers, and every later form of connected gaming, from university dungeon worlds to global shooters, rests on the approach proved here.
Midway rebuilt Tomohiro Nishikado's Western Gun around a microprocessor rather than fixed circuitry, producing the first arcade video game to run software. Programmable hardware changed how games were made and what they could contain.
Software replaced soldering. That shift separated game design from electrical engineering, made libraries of titles possible on shared hardware, and set the arcade on its path from single-purpose novelties to a business of annual releases. Every later leap in arcade ambition depended on it.
Atari's tank game drew its battlefield as glowing wireframe outlines viewed through a hooded periscope, giving arcade players an early taste of moving through a three-dimensional space rather than watching one from outside.
Few machines of the period asked players to imagine themselves inside the picture. This one did, and it demonstrated that a convincing sense of space mattered more than surface detail. The wireframe look also became shorthand for computer vision itself, borrowed by films, television titles and advertising for years afterward.
Sega's racing game rendered cars and circuits as flat-shaded polygons calculated on the fly, with camera angles players could switch between. It made three-dimensional graphics a paying arcade attraction rather than a laboratory demonstration.
Real-time polygon rendering stopped being a technical curiosity and became something audiences would pay for repeatedly. That shift reset expectations across the entire industry within roughly three years, and the design questions it raised, camera placement above all, are still being worked out.
Microsoft released the first version of DirectX, letting Windows programs reach graphics, sound and input hardware almost directly. It ended the game industry's dependence on MS-DOS and made Windows a coherent target for commercial development.
A software layer decided which platform the computer game industry would live on. By giving Windows the low-level access games required, Microsoft consolidated a fragmented field around one operating system and one hardware interface, a position it has held for roughly three decades.
Boards built on the Voodoo chipset from 3dfx Interactive reached buyers during 1996, moving texture drawing off the main processor and onto dedicated hardware. An accelerated computer suddenly outclassed the new consoles.
Consumer hardware acceleration reset the argument between computers and consoles. Instead of being a compromise platform, a personal computer with the right card became the place where games looked best, at the cost of a purchase console owners never had to make. That trade-off has defined the market ever since.
The GeForce 256 went on sale on October 11, 1999, marketed as a graphics processing unit because it handled geometry and lighting work the main processor had previously done. The term outlasted the product by decades.
A category name entered general use and stayed there, describing a component that grew from a game accessory into one of the most consequential pieces of computing hardware in the world. Moving geometry and lighting onto dedicated silicon also set the architectural direction later consoles and modern parallel computing both followed.
Microsoft's depth-sensing camera for the Xbox 360 read body movement and spoken commands with nothing held or worn. It sold at record speed, then struggled to find games that justified the hardware.
Real-time body tracking had escaped the laboratory and become a mass-produced consumer part, years before such sensors were ordinary. The launch also delivered a blunt lesson about hardware without software: a remarkable sensor cannot carry a platform if designers never work out what to build with it.
Sony released the PlayStation 5 Pro with a stronger graphics processor and machine-learning upscaling, priced above the original console and shipped without a disc drive included. Mid-cycle revisions were now routine.
Console generations have become cycles with upgrades inside them, closer to how phones and computers are sold than to the fixed hardware model that defined consoles for thirty years. Machine-learning upscaling also became a standard part of how rendering performance is achieved rather than an experiment.
