The last day

Adobe killed Flash on 31 December 2020, and the browser lost the largest games platform it had ever had

The date had been public knowledge since July 2017, when Adobe and its partners — Apple, Facebook, Google, Microsoft, and Mozilla — issued a coordinated announcement that Flash Player would reach end-of-life at the close of 2020. Three and a half years of warning is an unusual luxury in technology. It made almost no difference. When the day came, tens of thousands of games stopped working, and most of them had nowhere to go.

Adobe's mechanism was not subtle. The company had baked a kill switch into every version of Flash Player distributed after a certain point: on 12 January 2021, the Player began refusing to run Flash content at all, displaying instead a static prompt to visit Adobe's own end-of-life page. The practical effect was immediate for anyone who still had the standalone projector or the browser plugin installed. For users of a major browser, the situation had already been resolved: Chrome, Firefox, and Edge had removed Flash support before the deadline arrived, and the mobile web had effectively been Flash-free since Apple declined to support it on iOS, a position Steve Jobs set out publicly in April 2010. The kill switch was therefore aimed at the tail — the enterprise intranets, the educational platforms, the holdouts — but it swept the games along with them.

The scale is difficult to state precisely. BlueMaxima's Flashpoint, the most systematic preservation attempt, had catalogued well over a hundred thousand pieces of Flash content by the time the deadline passed, and games formed the majority of it. The Internet Archive had been running Flash content in the browser via emulation since 2020, using a framework called the Emularity to load the Ruffle emulator or, earlier, Gnash. Neither approach was complete. The gap between what was catalogued and what was actually playable, and between what was catalogued and what existed in the first place, remains unmeasured.

How a deadline becomes a cliff

The coordinated 2017 announcement created an uneven response. Large studios ported their properties or abandoned them; the long tail of hobbyist work, educational software and portal-era games received no such attention. Portals like Newgrounds, Miniclip, and Kongregate had built their entire catalogues on the Flash runtime, and while Newgrounds moved relatively quickly to support HTML5 alternatives and to fund the Ruffle project directly, the transition was never clean. Ruffle, an open-source reimplementation of the Flash Player written in Rust and compiled to WebAssembly, executes ActionScript 1 and 2 reasonably well but ActionScript 3 — the version used by most games made after roughly 2006 — was still incomplete at the end of 2020 and remains a work in progress.

The economics had already moved. After Adobe acquired Macromedia in 2005, Flash became the de facto standard for interactive browser content, and the plugin install base reached figures that no other optional browser technology matched. By the early 2010s, that position was eroding from multiple directions: the rise of iOS and then Android without plugin support; the arrival of WebGL and the canvas element as native alternatives; and a sustained, well-documented record of security vulnerabilities in Flash Player that made enterprise IT departments hostile to keeping it installed. Adobe's own Creative Cloud had shifted revenue toward subscription software long before end-of-life was declared. The plugin had become a liability.

What that meant for games was that the platform's death was being engineered from outside the games industry, by people whose primary concern was browser security and mobile compatibility. The games were not the product; they were content that happened to use the runtime. When the runtime was withdrawn, they were withdrawn with it.

What the kill switch actually killed

Bit rot — the degradation of software as the environment it depends on changes around it — usually operates slowly. A program that ran in 2005 will often still run in 2010; by 2015 it may need a compatibility layer; by 2025 the dependencies have moved too far and the path back is long. Adobe's kill switch compressed that process to a single calendar date. Content that was functional on 30 December 2020 was non-functional on 1 January 2021, not because anything in the files had changed but because the interpreter refused to interpret.

This is why the preservation argument shifted from archiving games to archiving runtimes. The logic, as the Internet Archive and projects like the Emularity made clear, is that if you can accurately emulate the Player, you preserve everything built on it simultaneously. The Video Game History Foundation and the Software Preservation Network have both argued in submissions to the US Copyright Office that security exceptions for preservation are insufficient — that the legal framework around circumventing access controls creates structural obstacles to exactly the kind of emulation work that would otherwise solve the problem. The Flash end-of-life was their clearest recent case study: a platform with tens of thousands of titles, most held by no institution, most with no surviving source code, made dark by a single corporate decision.

The content that fared worst was not the commercially visible games but the hobbyist and semi-professional work that had made the portal era genuinely interesting: the experiments, the weird things, the games made by one person on a slow connection who posted to Newgrounds in 2003 and never updated the file. BlueMaxima's Flashpoint ran on volunteer effort and disk space and the willingness of people to download a launcher that weighed tens of gigabytes; the Internet Archive ran on server capacity and legal uncertainty. Neither is a permanent solution. Both are better than nothing, which is the alternative that arrived on 31 December 2020 for everything that neither project reached.

The text-based internet that preceded the plugin era — the MUDs, the talkers, the shell-hosted games — survived its own platform transitions more easily, partly because plain text is more portable than compiled binary content and partly because communities of interest kept running servers long after the original institutions stopped caring. MUD1, written by Roy Trubshaw and Richard Bartle at the University of Essex, is still accessible. The Flash canon, by contrast, required a plugin that only one company could supply, and that company decided to stop.

Three and a half years of warning is an unusual luxury in technology.

The last day was a Thursday. Nothing dramatic happened to mark it; the browser just quietly stopped being the platform it had been.