This article takes a look at browsers on video game consoles in detail, though only covers official web browsers. Many consoles have browsers installable via custom firmware and homebrew, but they’re beyond the scope of this post, as are non-web systems such as Satellaview and online services that didn’t provide a browser, such as XBAND, Sega Meganet, and Sega Channel.

Game console browsers were of interest to web developers for a period while personal computing devices and mobile browsers were still establishing themselves. Overall, the development of console browsers provides an insight into a juvenile web, slowly growing and establishing itself, as well as an insight into game console user interfaces.

CD-i

The Compact Disc-Interactive format and hardware created by Philips and Sony was an ill-fated attempt to bring interactive multimedia to the masses. Development on the project started in the mid ’80s, and home players arrived in 1991.

The CD-i’s inclusion as a ‘game console’ here is debatable as it was designed for and touted with much broader capabilities. However, towards the end of its life it was marketed much more as a game console, and it is as a game console that it is best known today – especially thanks to the infamous Mario and Zelda games.

With a modem and a CD-Online disc (known as Web-i in the United States) released in late 1995, users could access the web in a very rudimentary manner. The term ‘internet-lite’ is seen paired with the CD-i frequently, for not only was the internet and indeed the entire World Wide Web burgeoning, but the CD-i was a rather limited machine that wasn’t well equipped for supporting the full-scale web.

The browser worked and had links to various web portals but was very limited, even given the primitive web of the time. The CD-i’s limited RAM meant that it could store very little, and that simply using the browser would overwrite other values in memory, such as preferences and game saves. The idea was that the CD-i would be a cheaper, TV-based computing device, available at a price point lower than typical home computers that could make it the gateway to the internet for the less technologically literate.

During 1996, the CD-i KeyControl keyboard was released, as were additional CD-Online discs. The idea being that new discs would be released periodically with additional games, software, and peripheral support. Records indicate six discs were released in total, with later versions including the ability to develop and deploy your own homepage. By late 1998, CD-Online was winding down on the CD-i, with a version of the service launching for PCs during the turn of the millennium. Throughout the early 2000s it fizzled out and domains went offline, with everything CD-Online coming to a complete close in the mid-2000s.

Sega Saturn

Sega’s 1994 console, the Saturn, gained internet access in late 1996 with the release of Sega Net Link – a tiny device with a custom chip and 28.8kbps modem that fit into the cartridge port.

For the Net Link, United States based company PlanetWeb provided their PlanetWeb Browser as the ‘NetLink Custom Web Browser’, an efficient and lightweight browser with a bespoke engine which was tailored and released for various devices in collaboration with manufacturers. Reading PlanetWeb’s frequently asked questions page from 1997, it is clear their browser was built carefully to target television-based devices and existing consoles, making it optimal for the Saturn.

Due to televisions having low-resolution screens compared to dedicated monitors, the PlanetWeb browser made use of ‘proprietary software technology for displaying clear, readable text and sharp images on an ordinary television.’. In other words, ‘the Browser features such advances as anti-aliased fonts’.

For its time, and given it is on an extremely limited console, it is remarkably fully featured. The browser has an inbuilt magnifier that can zoom to multiple levels, image support, history, bookmarks, an address book, the ability to download files temporarily, and full parental controls for filtering.¹ That is in addition to all the configurable options and settings available in a menu accessed by pressing Start. There is a wide selection of themes with different cursors and sounds – any of which wouldn’t be out of place in Kid Pix, with blips and bloops and springy sounds of all varieties.

Each time PlanetWeb boots, a random splash screen is shown. Net Link was tied in to the Saturn’s wider online gaming capabilities, allowing ‘On-Line’ battles. Beyond direct multiplayer, it also built a community around the console. For a while PlanetWeb hosted a game exchange mail-in forum allowing users to email in their Saturn game saves via the browser to share them with other people. PlanetWeb also launched Planet WebMaster, a web development and hosting platform that people could use with their browser.

In the United States of America, two versions of the browser were widely distributed. Version 2 and version 3. Version 3 brought greater optimisation, more legible fonts, and IRC support. Neither of these versions supported frames, however, instead displaying them as text links. Version 1 was pressed but never released due to a flaw in parental controls – all copies were destroyed.

The final version of the browser was Beta 4.035, which was never released physically and was only available as a downloadable update. Version 4 introduced full support for displaying frames inline, more IRC commands, support for embedded wav and aif audio, more shortcuts, faster input, different bullet shapes, outlined fonts, better image format compatibility, and many bug fixes. Beta 4.035 even introduced SSL for e-commerce.

It is likely version 4 would have seen a complete and full release had Sega not discontinued the Saturn in 1998 after rapidly losing market share to the Nintendo 64 and PlayStation, which dominated the market. Following the discontinuation, some users took it upon themselves to splice bits of version 4’s beta and previous stable versions together to compile a more complete and final browser experience.

Apple Bandai Pippin

Apple Computer Inc released the infamous Pippin in collaboration with Bandai in 1996. The Pippin is a fairly standard PowerPC computer under the hood and runs software built atop Classic Mac OS. As such, several variants of mainstream desktop browsers were ported. These browsers never saw major utilisation, however, as the Pippin was a failure, and people noted that the Sega Saturn was a far cheaper gateway to the web, even factoring in the additional cost of the Net Link attachment.

For the North American market the @WORLD Browser, based upon Spyglass Mosaic, was released in 1996, bundled with @WORLD branded consoles sold in the United States. @WORLD browser came with plug-ins providing support for QuickTime and Macromedia Shockwave. It had filtering control via SurfWatch ProServer and support for some of Apple’s StyleWriter set of ink-jet printers. In mid-1997, a beta version titled @WORLD Online Suite Premium had an terrifically limited release. In early 1998, version 3 of the browser released, but details on what it changed are scarce, and by that time Bandai were announcing they were dropping support for the Pippin due to extremely poor sales.

In April 1997, SurfEZ! (also called the Katz Media browser) was released for the Katz Media Player 2000 – an improved version of the Pippin distributed under licence in Canada and Europe. The console boasted mildly improved specs, and the browser supports HTML, plug-ins and Java, with capabilities for developers to customise the interface for specific solutions.

In Japan, a disc titled ‘Internet Kit’ was distributed with Pippins. Version 1.0 was based on Netscape Navigator 1.12 and released in March 1996. It was superseded by version 1.1 in June of that year, which added support for saving images to floppy disc. Version 2.0 released in December 1996 and updated the browser to a Netscape Navigator 2.01 base while adding support for some StyleWriter printers. A Japanese/English machine translation plug-in was available for this version but required a 2MB memory module to use.

J-DATA, a company that offered internet services for the Pippin in Japan, released NetCruiser (ネットクルーザー) 2.0 in a bundle with other software for the Pippin. NetCruiser was originally a Windows browser, before it was ported to the Macintosh and then to the Pippin via a process Apple called ‘pippinizing’, which was possible due to Pippin’s similarity to the Mac. J-DATA also published a monthly J.D.Press magazine which included listings of website addresses as barcodes, which could be scanned via a barcode reader device (titled the ‘Super Cat’) to bypass the laborious task of typing out a URL.

In 1997, the final Pippin browser for the Japanese market, WebViewer, released in Japan. WebViewer, instead of being built upon Netscape Navigator, was instead built upon Internet Explorer 3.0. It brought much-improved performance but required an 8MB memory module to function. In some cases it was bundled with NetCruiser 3.0, an updated version of J-Data’s browser.

Apple wouldn’t release their own browser, Safari, until several years later in 2003, long after they had stopped supporting the Pippin.

Nintendo 64

The Nintendo 64, as it released in 1996, never received a browser, but the 64DD attachment, which released in 1999, brought internet capabilities to the console and a browser via Randnet Disc (ランドネットディスク) which released in early 2000. Randnet was Nintendo’s dreadfully unsuccessful Japan-exclusive online service. The Randnet Disc includes a general-purpose browser for the system, capable of rendering various webpages. While the browser can be controlled via the system’s normal controller, it also has support for keyboard and mouse input. Details about Randnet, and the browser particularly, are far from abundant.

Game.com

It is no fault of your own if you’ve never heard of the Game.com, for it was a poorly, dot-com doomed handheld game console from Tiger Electronics, who you may know for their licensed LCD handheld games. In 1997, with the release of the Game.com, they attempted to compete with the Game Boy to dire results one might expect.

A large part of their marketing was that the Game.com had internet capabilities. Indeed, by purchasing the Internet Cartridge and connecting the device to a modem (either an existing one or the Game.com-branded model), you could access a non-graphical, text-only browser. It was very bare bones and feature-poor.

Game Boy Colour

The Game Boy Colour released in 1998, and in Japan exclusively received what one could charitably call a browser with Mobile Trainer in 2001. Using the Mobile Adapter GB, which is an accessory that connected to mobile phones, the Game Boy Colour could play games online. The Mobile Trainer cartridge, however, provided an email client and web browser.

Given the Game Boy Colour’s meagre 8-bit processor, even calling it a browser is a stretch. It can only access an official Nintendo website, which included news, contests, and information about some downloadable content – arbitrary web addresses cannot be input. On account of the system’s limitations, it only supports a few HTML elements, like paragraphs and lists. Impressively though, it does support images, albeit in black and white.

This system also worked on the Game Boy Advance by virtue of its backwards compatibility with the Game Boy Colour.

Sega Dreamcast

What would be Sega’s last console, the Dreamcast, released in 1998. It didn’t get one browser like its predecessor, the Saturn, but instead received three distinct lineages of browsers. Dream Passport, various versions of the PlanetWeb Browser, and Dreamkey.

Dream Passport (ドリームパスポート) version 1.01 was bundled with Dreamcasts sold in Japan on launch. It lacked support for many web standards and common formats, so it was quickly obsoleted by the updated Dream Passport 2. Owners of the first version of Dream Passport could exchange their copy at some retailers for Dream Passport 2, free-of-charge, between the 5th and 31st August, 1999. After that time it had to be purchased online via Sega Direct. Very briefly in 2000 a version of Dream Passport was bundled with the Dreamcast LAN Adapter, titled ‘Dream Passport 2 for LAN’, before being replaced with Broadband Passport (ブロードバンドパスポート), a version bundled with the Dreamcast Broadband Adapter. Later in 2000, Dream Passport 3 released, again with improved support for web standards.

The final technically improved version of Dream Passport to release was Dream Passport Premier in 2001. In addition to being usable on the Dreamcast, it can also be used to install versions of Dream Passport on Windows and Macintosh.

There were also variants of Dream Passport produced, such as the themed Hello Kitty Dream Passport 2 (ハローキティのドリームパスポート2), Sakura Taisen Dream Passport 3 (サクラ大戦 ドリームパスポート3), and Dream Passport 3: Urban Style. The latter was part of an Internet Starter Kit aimed at non-tech-savvy folk – particularly, women. In addition to the special version of Dream Passport 3, the kit included a translucent keyboard and mouse, a mousepad, some manuals, and a guide: The Internet for Housewives.

Also in Japan, Microsoft made an attempt to port the first generation of WebTV to the Dreamcast via the Microsoft WebTV Connection Kit (マイクロソフト ウェブ・ティービー接続キット). It was possible by virtue of the Dreamcast’s Windows CE abstraction layer and had an Internet Explorer-based browser. There were only two versions: the initial release version and Microsoft WebTV Connection Kit: 128bit SSL Compatible Version, which added exactly what the title suggests.

In the United States of America, PlanetWeb returned with their PlanetWeb Browser under the name ‘Web Browser’. It first released for the Dreamcast in September 1999. A small version 1.2 update released afterwards, and in 2000, Web Browser 2.0 released with better JavaScript support and support for Macromedia Flash. It also added support for uploading and downloading Dreamcast save games, downloading and playing MP3s, and included a full copy of the puzzle game Sega Swirl. Web Browser 2.62 released in 2001 as a successor to 2.0 and included demos of StarLancer, Speed Devils Online, Typing of the Dead, and 4x4 Evo. It was replaced later the same year by Internet Browser v3.0, which brought support for Macromedia Flash 4, again improved JavaScript support (JavaScript 1.2), and the ability to play Java 1.1.8 applet-based web games with PlanetWeb’s own Java Virtual Machine (JVM).

The interface of PlanetWeb on the Dreamcast is very similar to the interface that was on the Saturn, albeit themed differently. It has the same menu, bookmarks system, address book, and zoom tool. Unfortunately the theming functionality wasn’t carried over. PlanetWeb also hosted a Dreamcast Fansite and web portal.

Outside of Japan and the United States (mainly Europe and Australia), rather than Dream Passport or a version of PlanetWeb Browser, Dreamkey was bundled with consoles. Dreamkey was built on NetFront, a bespoke browser developed and provided on license by the Japanese company Access Co Ltd. The first version was 1.0, which was followed by 1.5 to add regional internet settings for the Republic of Ireland. Version 2.0 released in 2001 and updated the browser engine slightly, improved the on-screen keyboard, added support for the Dreamcast Mouse, and for a messaging service called Dreamnote. Dreamnote allowed sending messages, including voice messages via the Dreamcast’s microphone, and had a feature where Dreamnote users could see other Dreamnote users on the same website and communicate with them.

In 2002, Dreamkey 3.0 released, removing restrictions surrounding internet service providers which had been imposed by Sega since version 1.0. A final version 3.1 was released exclusively in Spain and Portugal to fix regional issues in 2003. There is a wide variety of miscellany hidden in Dreamkey, including hidden 3D models, images, videos, demos, ‘Sonic the Hedgehog’ listed as a contributor in the credits, and the ability to use the Dreamcast’s light gun controller as a pointer, shooting links to click them.

WonderSwan

Launching in March 1999, the WonderSwan was a handheld game console that released exclusively in Japan. A tiny 16-bit console with a mere 16KB of RAM. In 2000, the MobileWonderGate launched – a collaboration between Bandai and the mobile network NTT DoCoMo.

MobileWonderGate is a physical adapter using a phone connection that provided the WonderSwan with an internet connection and an extra 512KB of RAM. It allowed downloading additional content for some games, email, and an exceptionally limited browser.

Its rudimentary engine has support for only a small subset of HTML 3.2. The browser itself has a bookmark system, and support for displaying GIFs, with settings to disable images, change their display, and configure the overall presentation of webpages. To bypass some of the severe limitations of the WonderSwan, web pages were proxied through a service to improve compatibility, not unlike Opera Mini.

PlayStation 2

The PlayStation 2, which released in 2000, got a fairly substantial web system. The PlayStation Broadband Navigator (PlayStation BB, PSBBN) was available in Japan and, provided the user had a hard drive in their Expansion Bay, a memory card, and a network adapter, allowed users to take their PlayStation online.

The PlayStation 2 only had a very rudimentary system interface, and PSBBN brought many more features. For the time, PSBBN was quite an advanced system. A pre-release version was released in early 2002, being replaced with v0.20 later in the year. v0.20 brought the ability for the software to update itself from the internet and manage save games. In v0.30, which released mid-2003, an email program was incorporated, as well as a service to handle monthly billing for some online games. A final minor update was released later in the year with v0.31 to fix an exploit.

In addition to various ‘channels’ where game trailers, demos, and teasers were showcased in manners similar to websites, PSBBN had NetFront Browser 3.0. It supported HTML 4.1 and JavaScript. NetFront Browser 3.0 was also distributed for Linux for PlayStation 2.

Also for the Japanese market, PlanetWeb produced EGBROWSER, which was released by Ergosoft Co Ltd in April of 2001. It includes Ergosoft’s Japanese/English machine translation and allows playback, though not saving, of MP3 files. It is a continuation of PlanetWeb’s bespoke browser system, functioning much the same as the releases on Sega’s consoles with a slightly tweaked interface. It supports HTML 3.2, CSS 1.0, various audio and image formats, SSL 3.0, Macromedia Flash 3.0, and JavaScript 1.1. It doesn’t support videos, Java applets, or PDFs.

Following the release of PSBBN, an improved version titled EGBROWSER BB was released in 2002. It improved support for CSS and achieved compliance with HTML 4.0 while bumping Macromedia Flash to version 4.0. It also made use of the PlayStation Broadband Navigator to store more data, such as emails and cookies.

In PAL regions, a Network Access Disc was distributed which had a browser that could access Central Station, the predecessor to PlayStation Network. It didn’t present any browser controls needed to access arbitrary websites, but people ‘hacked’ it to display sites of their choosing by pointing its configuration at an IP address other than Sony’s own.

Nintendo DS

For their 2004 console, the DS, Nintendo partnered with Opera to bring version 8.5 of the Opera browser to the console. Releasing in 2006, the browser was sold with a Memory Expansion Pak required to use the browser, which fit into the console’s Slot-2 and is usually reserved for Game Boy Advance games. It provided 8MB of RAM, increasing the DS’ total memory from 4MB to 12MB.

The browser was released twice, once for the original model of the DS, and again for the sleeker 2006 hardware revision, the DS Lite. The only difference between the two being that the DS Lite version had a smaller Memory Expansion Pack that sat flush with the Lite’s smaller chassis. The original Expansion Pak works in the Lite but is the same size as a Game Boy Advance game and thus protrudes.

Though limited, the DS’s browser fit a niche for mobile internet access in a world where smartphones were not yet ubiquitous.

For a tiny browser on a very limited console, it is reasonably fully featured. It has support for JavaScript, images, history, bookmarks, searching, a reasonable degree of settings, zoom options, and display toggles. It doesn’t have support for videos or other complex multimedia, though. Inputs can be performed using the DS’s on-screen keyboard or by writing characters on the touchscreen to be interpreted via handwriting recognition. The browser lacked any inbuilt filtering system, instead advising people to find and configure a proxy. A password can also be configured to be required on launch. The browser manual notes that the browser might run out of memory sometimes and humorously notes that if it does, you should turn your DS off and then on again.

Two presentation modes were available. The overview mode showed the complete page on the bottom screen, which could then be selected to view it on the top screen. The fit-to-width mode made the site take up the entire horizontal width of the display so that the user doesn’t need to pan around the page. In a time before responsive web design – which is a term that would only be coined in 2010 – this made websites designed exclusively for desktop use somewhat usable. While possible to use the interface without a stylus and to navigate using only the buttons or a finger, a stylus is the much more precise control method.

DSi

The DSi released in 2008 and had a browser included with its system software in most cases and was available to install for free through the DSi Shop if not. It forewent the need for a Memory Expansion Pak like the base DS thanks to the DSi’s inbuilt 16MB of RAM. The browser was much improved from the DS’, being a stripped-down version of Opera 9.50.²

Introduction of HTML canvas support, particularly, was major, and various sites including DSiPaint, DSiCade, DSiPlaza, and Social Neko, were built specifically for the DSi browser, taking advantage of the newly introduced support for canvas and other web technologies to facilitate more interactive experiences. The browser only supports one font and three font sizes, with content transformed to adhere to the limitation.

Reviewers at the time were critical of its lack of support for Adobe Flash, lack of video capability, and its semi-regular errors messages about running out of memory, though they considered it a major improvement from the prior attempt to bring a browser to the DS. It still lacked capability for playing videos. The browser got a single update with version 1.4 in August 2009, which slightly reduced the browser’s size on the system.

Despite having increased RAM, the DSi isn’t compatible with the DS browser due to a hard check for the Memory Expansion Pak, which cannot be installed in the DSi due to lacking Slot-2.

Wii

Continuing their partnership with Opera, the Internet Channel was the browser for Nintendo’s 2006 console, the Wii. The browser first saw a release for free in a trial capacity in late 2006 before launching fully in mid-April 2007. Starting in July of 2007, it became paid, costing 500 Wii Points.

It remained at this price until the start of September 2009, when it became free once again, and switched from Macromedia Flash Player 7 to Adobe Flash Lite 3.1, based on Flash 8. Starting in October 2009, users who had purchased the Internet Channel during this paid period were offered a free Nintendo Entertainment System game valued at 500 points.

On October 10th of 2007, the Internet Channel saw an update, which brought USB keyboard support; interface improvements; the ability to favourite 56 pages, up from 48; functions to copy text from webpages into the search field; and further integrations into the Wii system so that users could send links to users in their Wii Address Book via the Wii Message Board.

Interestingly, trying to navigate to /. takes you directly to the website Slashdot, a long-standing shortcut in Opera. I’m surprised Nintendo allowed such a thing, which leads me to assume they were unaware. Google released Google Reader Play, a version of Google Reader specifically for use on the Wii, that later expanded for use on other consoles.

As a very brief aside, the Wii’s precursor, the GameCube, didn’t have a web browser per se (at least nothing internet-enabled) but did have a browser engine. Various game demo discs were built with HTML using a browser by who else but PlanetWeb. Research into this is detailed in The GameCube’s Lost Internet Browser Discovered.

PlayStation Portable

The PlayStation Portable was released in 2005 and gained a free system-integrated web browser in September of that year with PlayStation Portable System Software 2.00. It used NetFront’s browser engine and got additional updates over time. In October 2005, 2.50 brought Unicode character support and the ability to save form inputs. 2.70 in April 2006 introduced support for Macromedia Flash 6. Trend Micro Web Security and Trend Micro Kids Safety were made available on the PSP in April 2009 to allow for content filtering.

The browser was rather fully featured, with support for creating/editing/deleting bookmarks, managing history, changing text sizes and how the page displays, JavaScript, images, using a proxy, cookies, cache management, and up to three tabs. The PSP even had comprehensive support for RSS on a system level.

Three view modes are made available by the browser. ‘Normal’ displays the page as served, ‘Just-Fit’ attempts to shrink some elements to fit the small screen without impacting layout too much, and ‘Smart-Fit’ forcefully wraps any content that horizontally overflows so it appears on the next line.

The life of the PlayStation Portable’s Internet Browser was cut a tad short due to relying on SSL 3.0, which was quickly fled from by the bulk of websites when a major vulnerability was found in the protocol in late 2014.

Xbox 360

Despite releasing in 2005, the Xbox 360 didn’t receive a browser until 2012, when it got a modified version of Internet Explorer 9. Initially it was only available to users paying for Xbox Live Gold (something that people were very unhappy about), but this restriction was later relaxed.

Microsoft went as far as to publish a full paper about developing for Internet Explorer for Xbox, covering everything from how the input works to design considerations for televisions to browser feature support and testing.

Though the browser has support for favouriting sites to return to later, it lacks any support for tabs or any other similar multi-tasking functionality. In the settings, toggles are present to enable requesting mobile versions of webpages, as well as to enable support for Microsoft’s SmartScreen filter to protect against malicious sites.

Using the SmartGlass console companion app for Windows, Windows Phone, Android, and iOS allowed navigating and interacting with Internet Explorer on the Xbox 360 via a mobile device. An experience much nicer than trying to use the on-screen keyboard. The Xbox’s browser interface is peculiar due to it not entirely fitting the 360’s Metro Dashboard aesthetic and taking design cues from the flatter version of Metro seen on Windows 8.

Being released so late in the console’s life cycle, it was never updated beyond Internet Explorer 9.

PlayStation 3

The PlayStation 3 released in 2006 with NetFront’s browser running on NetFront’s proprietary engine, before later being replaced entirely with WebKit.

It has a wealth of features: disableable JavaScript support, the ability to change the page resolution, a zoom function, multiple display sizes so pages could conform to standard desktop sizes or take up the whole screen, website screenshot functionality, options to copy the address, a bookmark and history system, the ability to save images, and tabbed browsing support via windows which can be slid between with the shoulder buttons.

In November of 2007, version 2.00 was released which introduced Trend Micro Web Security and Kids Safety as options in the browse settings. Version 2.20 from March of 2008 allowed playback of some videos while they download. In July version 2.40 launched, which added an internet search option in the XMB and search as an option in the browse mode menu of the Internet Browser. However, this version was withdrawn later the same day due to instability, and the changes were more widely distributed a few days later in 2.41.

In October of 2008 support for Adobe Flash 9 was introduced with version 2.50, and in December this was expanded to allow full-screen display of Flash content with version 2.53. Version 2.70 in April of 2009 brought several features pertaining to the system’s web browser. An internet search function was added to the games menu as a shortcut for performing a Google search on the item’s title; the ability to copy and paste text from websites was added; Internet Security was added as an option; and functions to copy the current page’s address and the address of a link were incorporated into the file menu. An estimate of how long a download will take was also added.

In version 3.00 which released at the beginning of September 2009, the ability to take screenshots of web pages in the browser was added, and with version 4.10 in February 2012, the browser got a substantial update in that it moved to WebKit. This brought huge load speed, performance and stability improvements, support for some HTML5 features, and page layout accuracy benefits. In addition to these more notable changes, the Internet Browser also saw occasional performance and stability improvements throughout the console’s life.

The PlayStation 3 also had support for connecting to printers and media servers, downloading content, and other advanced functionality not typically seen in game consoles. It is perhaps the closest a game console got to the ’90s dream of a game console serving as a computer on your television.

PlayStation Vita

Sony’s second (and far less successful) swing at the handheld gaming sector came in 2011 with the PS Vita. From launch it had a non-uninstallable web browser app. The browser is WebKit-based and has no support for Adobe Flash. Though it lacked tabs, the Vita browser did have support for up to eight tab-like windows, and links on a page could be opened in new windows at will. Attempting to open more than eight closes the last of the existing windows. Images can be downloaded from websites, and a comprehensive bookmarks system is included, allowing the creation, deletion, editing, and sorting of bookmarks, and even a way to keep them organised by filing them in nested bookmark folders.

Some models of the Vita had support for 3G networking, making the browser usable while away from a Wi-Fi connection.

Open windows appear on the browser’s LiveArea screen, allowing quick access to them. As on the PlayStation 3, a button is present when viewing games in the system menu to perform a search for the game’s title in the browser. Many system apps also have a help action icon in their LiveArea which opens the application’s user guide in the browser. Via the Vita’s Parental Controls app, the browser can be disabled entirely, or Trend Micro Kids Safety and Web Security could be subscribed to for the browser. Alongside the PlayStation Portable and PlayStation 3 services, the Trend Micro services reached end-of-life in June 2022 and end of support a year later.

System software version 1.80 which released in late August of 2012, brought the ability to scroll pages using the Vita’s rear touchpad and added a button to immediately jump to the top of the page. Version 2.00 in November 2012 improved the browser engine, let the browser use more GPU resources, allowed accessing the browser while in-game, and allowed selecting links with a pointer while holding down the shoulder buttons, in addition to via the touchscreen.

In April of 2013, with version 2.10, support for displaying videos was added, provided the Vita has a memory card installed. That update also brought support for viewing HTML emails to the email app. The aforementioned video support was later enhanced in August with version 2.60 to support more videos. The next update came in October of 2014 with version 3.30, which allowed closing all open tabs at once and improved HTML and JavaScript support.

In 2013 Sony released the short-lived PlayStation TV, a variant of the Vita designed as a home console. This brought the Vita experience, including its web browser, to the television.

3DS

The Nintendo 3DS launched in February 2011 but didn’t receive its ‘Internet Browser’ until system update 2.0.0-2 in June 2011. Internet Browser uses the WebKit-based Netfront Browser NX v1.0. It is more like the later-released Wii U Internet Channel than the previous DS Browser, with a much more minimal interface that is in keeping with the 3DS’s bubbly, glossy system theme.

The browser didn’t support Flash, nor video playback, and saw minimal updates throughout the system’s life, beyond patching exploits. The stylus can be slid across the lower screen to scroll pages, and unique to the 3DS Internet Browser is the ability to show .mpo image files in stereoscopic 3D on the top screen, though this doesn’t work inline.

The browser offers a choice between Google and Yahoo! as search engines. To allow sharing images from the console, Nintendo provided the 3DS Image Share Service. It made use of the browser’s ability to download to and upload from the 3DS’s System Memory or SD Card.

New 3DS

The New Nintendo 3DS (a name which time has been rather unkind to) released in 2014 with a variety of hardware changes, though relevant to web browsing are the improved processors (four cores, up from two) and RAM (256MB, up from 128MB). Owing to this improved hardware, it received an improved version of the browser.

Most notably, this improved release shipped with video playback support and a newer WebKit version with support for HTML5. It was based on NetFront Browser NX v3.0. Adobe Flash remained unsupported. It came with an entirely new user interface with options to explicitly request the mobile versions of pages and a toggle to enable more aggressive text wrapping.

For Japanese models of the New 3DS, a mature content filter was implemented that could be disabled by purchasing a small paid verification DLC with a valid credit card.

Wii U

The Wii U released in 2012, and the launch system software was without a browser. However, the very first system software update, 2.0.0, which released on the same day as the console, immediately introduced the ‘Internet Browser’ application along with other network-connected apps. Initially shipping with NetFront Browser NX v2.1, the version of WebKit provided saw consistent updates throughout the Wii U’s life up until v3.0.4 in 2017. It doesn’t have support for any plugins such as Adobe Flash.

The Wii U’s Internet Browser has a lot of the charm you’d expect from Wii-era Nintendo. The interface is displayed on the GamePad while the television displays spotlights over a stage curtain which can be opened and closed from the controller via a button. If the button is held down, the curtains are opened with a drum roll. If the curtain remains closed for a few seconds, a Mii appears in front of it and cycles through a collection of idle animations with various props. When the curtain is open, the page currently viewed on the gamepad takes up the entire television display.

The bookmark menu is a literal book, and pages of bookmarks can be flipped through. There are sixteen pages with five bookmarks each. The button for the bookmarks page is the user’s Mii holding a book, and as you flip through the pages of your virtual bookmarks book, your Mii can be seen flipping the pages in the bottom left corner of the GamePad.

By depressing the Wii U GamePad’s shoulder buttons and tilting the controller back and forth, the webpage can be scrolled vertically. The GamePad provides complete control independent of the television display, allowing for actions such as adding videos to the queue without interrupting playback. The Internet Browser represents one of the rare few cases where the Wii U’s GamePad is genuinely an enhanced input method and not just a tacked-on gimmick.

The browser only saw one major update, which came with version 4.0.0 in September of 2013. It introduced skipping and speeding up video playback with the shoulder buttons, saving usernames and passwords, PDF viewing capabilities, and the ability to access the browser while a game is suspended, so users can reference it for guides and use it without losing a save, or use the Wii U Image Share service which also launched with the update and brought the ability to quickly share screenshots to social media (it was discontinued alongside the 3DS version in October of 2022).

It also brought an update to the settings. Previously, the settings only provided the ability to choose between Google or Yahoo! for search, manage saved data, and configure a proxy. However, the update added functionality to change the user agent and allow computers on the local network to access the browser’s developer tools. Nintendo even provided rather comprehensive developer documentation for the browser.

For someone so inclined, the Wii’s Internet Channel can be run in the Wii U’s Wii Mode, but only if re-downloaded in the virtual Wii environment, as the Internet Channel can’t be transferred from the Wii to the Wii U with Nintendo’s Wii System Transfer tool.

PlayStation 4

The PlayStation 4 released in 2013 with an inbuilt browser referred to creatively as ‘Internet Browser’. The browser uses WebKit, like Safari. It has multi-tasking functionality so that a user can switch between the browser and a game or application without either program fully closing and having to be reopened.

It has web filtering support, bookmark functionality, and support for ‘Frequently Used Pages’ which presents the eight most recently accessed websites, and it can manage up to six windows but lacks support for viewing PDFs and Adobe Flash. The browser does support WebGL and JIT inside the PS4 store and platform content, but it isn’t exposed for external developers.

As with many consoles before it, the PlayStation 4’s browser proved a solid vector for exploits to attack so that users could install custom firmware on their console.

Xbox One

The Xbox One released the same year as the PlayStation 4, 2013, with its own inbuilt browser, an updated version of Internet Explorer. Sporting a modified interface, it – like the Xbox 360 version of Internet Explorer – was designed with support for the Xbox Kinect, keeping an interface that could be controlled with physical gesture controls and voice commands. It also brought the ability to use Internet Explorer in a split-screen view, so the user could display the browser alongside games and applications.

In 2015 Internet Explorer was removed from the Xbox One with the ‘New Xbox One Experience’ update. In its place, the then new Microsoft Edge. This was the version now known as ‘Microsoft Edge Legacy’, built upon the EdgeHTML browser engine forked from Internet Explorer’s MSHTML (Trident) engine. It was no longer designed for the Kinect, instead having the interface of the desktop Windows 10 version.

In 2020, a completely overhauled version of Edge, colloquially referred to as ‘New Edge’ and built upon Google’s Chromium began rolling out. This updated version reached the Xbox One in September 2021. As Edge is on a Windows 11 desktop, Edge is on the Xbox One. The experience is identical, down to the last pixel on the Microsoft Fluent-styled interface. Some specific features, namely browser extensions and downloads, aren’t supported.

SteamOS

Valve’s SteamOS operating system released in 2013 with an inbuilt web browser interface. This browser could also be accessed via Big Picture Mode on the desktop version of Steam. In late 2015 an update to the browser allowed playback of some DRM-protected content in the browser.

The Steam client is built on the Chromium Embedded Framework (CEF), so is itself already a modified browser instance. In addition to the web browser made available in Big Picture Mode, a browser window is accessible in the in-game overlay of Steam’s desktop interface. It supports multiple tabs, bookmark management, and offers minimal browser settings, mainly allowing you to change the browser home page.

In 2022 Steam released SteamOS 3.0, which came with an updated Big Picture Mode interface. However, since the release of 3.0 and as of publication, this updated interface exposes no general browser interface, in-game or otherwise. Being Linux-based, any browser that supports Linux can be installed trivially through external package management but can’t be installed directly via Steam itself.

Xbox Series

The two Xbox Series consoles, the Xbox Series S and the Xbox Series X, were released in 2020 and initially had Microsoft Edge Legacy – the same as the Xbox One. However, this was replaced with the new Chromium-based Edge at the same time as it was replaced on the Xbox One in September 2021.

Both Nintendo and PlayStation have refrained from releasing browser applications for their consoles following the Wii U and PlayStation 4, respectively. Neither the Switch, the Switch 2, nor the PlayStation 5 has a general-purpose, user-facing browser. Browsers are present in the systems for the purpose of facilitating login flows, content display, and web applications, such as Nintendo’s eShop, but they are restricted and don’t allow arbitrary navigation.

Despite this, people often configure proxies, follow trails of links, or message themselves URLs to break outside of intended pages. Unfortunately performance using these methods is sub-par, as a result of the browsers not being built for general use.

The Nintendo Switch and Switch 2 both use the WebKit-based NetFront NX browser, though it is worth clarifying that the similarity between the name of the browser, NetFront NX, and the Nintendo Switch’s development codename, NX, is pure coincidence. The PlayStation 5 uses Sony’s own wrapper around WebKit.

While certainly a welcome addition to consoles, especially if wanting an exploit to enable running arbitrary software, general-purpose web browsers on game consoles are no longer as coveted of a feature as they once were. Time has moved on, and with it has brought cheap web-capable devices and budget television dongles with pre-installed browser apps. Developers can still take away lessons from the era of game console browsers, but a game console likely isn’t a person’s only method of getting online anymore, as it might have been in the ’90s with the Saturn.

Instead, it is harder to log off from the web than it is to log onto it in our current era. No longer is the web a place one must venture to, for it is instead with us at all times.

This is a non-exhaustive list of these technologies for reference. Though I have attempted to be relatively comprehensive in covering usage, accessibility often necessitates varied usage and unique assembly of technologies, such that this post does not and will not cover every case.

Screen Readers

A screen reader is an output method that audibly relays the content on a screen via non-visual methods. In addition to reading text and content that is directly upon a screen, a screen reader also often conveys semantic meaning. For example, though the word ‘navigation’ may not be displayed visually, if a section is designated as ‘navigation’, a screen reader might convey that. Screen readers are extremely configurable, like many other accessibility tools, and are commonly modified for specific user needs. Often, synthetic speech output from screen readers will be at such a pace that it cannot be comprehended by people unacquainted.

They are most commonly associated with people with visual disabilities, though are also used in many cases by people who are illiterate or have learning disabilities.

Screen Magnifiers

Screen magnifiers are an output technology which enlarge sections of a display for easier viewing. This can be achieved in technology with often toggleable enlargement of specific screen sections or with physical magnifying lenses.

Most commonly, screen magnifiers are used by people with visual impairments that are significant enough to impact computer use but which don’t necessitate a screen reader. Magnifiers are, however, also used by people who lack fine motor control and thus need enlarged inputs.

Braille Displays

Electronic braille displays, also known as refreshable braille displays or braille terminals, are dynamic braille output interfaces. Braille is a tactile writing system where characters are represented as embossed dots on a 3x2 grid known as a ‘braille cell’. Different characters have different dots within the cell embossed and in different arrangements, such that a person with a visual impairment can identify a character by feeling it with their finger.

Electronic braille displays take output from a computer and mechanically displace a surface into braille cells for the user to read. On most devices the braille is displayed in a single strip between 40 and 80 cells wide, though for mobile readers it can be narrower. Depending on the device, the current cursor location might be indicated by the dots vibrating or additional dots above or below the cell. These displays are often used for the same purpose as speech synthesisers, and sometimes both tools are used by the same individuals who switch between them depending on circumstance. Braille displays are especially valuable for people who are both deaf and blind.

Tactile Graphics Displays

Functioning very similarly to braille displays, tactile graphics displays are an output technology used to convey more graphical concepts, such as spatial layouts, charts, geometric shapes, or maps. They’re usually much denser than braille displays, with many more pins. Instead of the tactile area being a single braille cell tall, they are usually much taller. In some cases pins are not a binary flat or raised, but instead can be fixed at a range of heights so depth can be more fully conveyed.

Tactile graphics displays are mainly used by people who are blind.

Sip-and-Puff

Sip-and-puff (SNP) is an input method which sends signals based on air pressure. The user sips (inhales) or puffs (exhales) on a device with their mouth to regulate this air pressure. The device is commonly worn as a headset, but it can also be mounted directly to a surface such as a wheelchair or table. Different inputs are achieved via sips and puffs of different strengths, which are calibrated based on the individual user. These strengths are often referred to as ‘hard’ and ‘soft’.

Most users of sip-and-puff machines have severely reduced motor function.

Gyroscopic Head Mice

Often worn as an earpiece or glasses, a gyroscopic head mouse is an input device which uses head motion. Depending on the level of motor control one has, it can be used to directly control an on-screen pointer akin to a mouse, or it can be used for directional input. A tilting of the head right to indicate a right input, and so forth. It can also be used in a method more akin to a joystick, which is sometimes taken advantage of for gaming.

Ergonomic Mice

We are all familiar with the standard computer mouse, which is an input device one glides along a surface to manoeuvre a pointer. However, there are many more variations on the popular computer mouse that are used. These variations are often used by people with carpal tunnel, arthritis, or repetitive strain injuries.

• Roller bar mice are a bar which usually sit bellow the keyboard. One rolls the bar towards and away from themself to move the pointer up and down, and slides it laterally to move left and right. Clicks can be achieved by either pressing down on the bar, or using dedicated buttons usually located bellow the bar.
• Trackball mice do not refer to the pre-laser tracking technology of the ’90s but instead to mice which are stationary on a surface and have a large trackball which can be directly rotated to move a cursor. Large trackball mice also exist so that the ball can be manipulated reasonably with a closed fist, elbow, foot, head, or other appendage.
• Vertical mice are very similar to typical mice but are oriented such that the buttons and scroll wheel are perpendicular to the surface it operates on. This places the arm in a more natural resting position similar to that it takes during a handshake, without the ulna and radius bones being crossed.

Mouth and Tongue Joysticks

Positioned close to the mouth, similar to a sip-and-puff, mouth and tongue joysticks are input devices which rely on oral control of a joystick. A sip-and-puff machine is often integrated into the joystick for further input controls.

In progressively-onset motor diseases, the mouth is often one of the last external body parts one retains control of. If the option for a joystick of this nature is available, people often find them especially desirable for motorised wheelchair movement.

Tongue Drive

Similar to but distinct from tongue joysticks, tongue drive systems are an input method which uses the movement of the tongue inside one’s mouth using magnets. A magnet is attached to the tongue via either adhesive or a method similar to a cosmetic tongue piercing, and the position of this magnet is identified by a specialised dental retainer worn on the roof of the mouth.

Tongue drive systems suited for applications similar to sip-and-puff devices.

Chording Keyboard

A chording keyboard is an input device which involves pressing down multiple keys simultaneously to create a ‘chord’. Different combinations of keys trigger different inputs based on a user’s pre-defined dictionary. If you’ve ever wondered how court room stenographers type so fast, it is because stenography is a chording input method.

Due to minimal reaching or stretching, and the reduced number of keys needed for a chording keyboard, it can be advantageous for someone with wrist or arm problems. The combinations also allow for configurations where only one hand is needed for input, which allows more effective input for people with only a single arm.

Switch Control Devices

Switch control devices are input devices which send single binary inputs. Commonly, they’re large buttons which are placed within the range of motion of someone with limited motor control, such that they can press them when needed. The exact placement depends on the individual case and what movement is possible. Sensitive microswitches can also be used and attached in specific locations such that they can pick up tiny movements like twitching a muscle or blinking. Typically, a device configured for someone using switch control devices will cycle through input options with the user activating the switch when it highlights their choice. The exact configuration depends on the number of inputs and the fidelity of movement the user is capable of.

Dwell Clicker

A dwell clicker is an input method where hovering in a single place for a chosen period executes an input. It is often used in tandem with other accessibility technologies, such as head-and-mouth sticks and eye trackers, but is also used by people who can operate more common pointer devices like mice but are unable to click.

Head/Mouth Stick

Head sticks and mouth sticks are input devices either attached to one’s head with a harness or held between one’s teeth, respectively. They can be used to interact with touchscreens like one would with a finger or to press physical keys. Trackballs are also often popular, as they can be nudged to move a cursor. Which of these methods is chosen depends on if the user can maintain a bite grip or suffers from jaw fatigue and what is most comfortable for them.

They’re mainly used by people with quadriplegia/tetraplegia.

Expanded Keyboards

Expanded keyboards are input devices which resemble standard keyboards. However, they are designed for easier input by people with reduced motor control. They are physically larger, with significant spaces between keys to give a larger margin of error, and have a physical keyguard. The keyguard sits above the keys, making them recessed, and gives users a place to rest their hand or arm without triggering inputs. It also gives users more control over which key they’re pressing, so one can more deliberately strike a single key through the hole in the keyguard with a finger or a pointer stick attached to one’s head or held in one’s mouth.

Expanded keyboards are used by people with many accessibility considerations. Conditions including cerebral palsy, Parkinson’s disease, multiple sclerosis, or essential tremors can cause unintended tremors and spasms, which the keyboard mitigates by preventing unintended strikes. Typing on a typical keyboard requires one’s hands to rest lightly above keys, so people with muscle weakness who would otherwise accidentally trigger the keys use expanded keyboards for their keyguard. Due to expanded keyboards commonly having large, clearly labelled keys, people with low vision and cognitive disabilities also benefit. As aforementioned, it can also help head-and-mouth stick users.

Eye Gaze

Eye gaze or eye tracking is an input method where a person uses their eyes to navigate a screen. Tracking either one or both eyes, the cursor follows wherever one looks. To make a selection, one must either use dwell control and hold their gaze, blink, or pair eye gaze with another input method, like a sip-and-puff.

Similar to gyroscopic head mice, input is sometimes alternatively done by gestures in a direction, such as looking right to navigate right. Tracking is usually achieved with a specialised infrared eye tracker and is paired with software for configuring and managing control.

Even in cases of total paralysis, control of eyes often remains, which makes eye gaze input one of the last input methods available in many cases.

Foot Pedals and Mice

Foot pedals and mice are input devices which one operates with their feet. A primary and secondary click can be controlled with one’s feet, as well as mouse pointer input, but keyboard input is often considered to require too much fine motor control. Foot control is often very capable, though is often not quite as precise as control with arms.

Immersive Readers

Immersive readers are an enhancing output method that blocks content on a page so that only what the user must see is exposed. For example, on a website, sidebars and navigation systems might be hidden, while only core article content is left undisturbed. Sometimes the user will also change the font in use, adjust the text size, and use a line reader to highlight their current point on the page or obscure everything else.

Immersive readers are commonly used by a wide variety of people but are of particular importance to people with cognitive fatigue, ADHD, autism, and dyslexia.

Speech Input

Speech input is an input method where a user speaks commands at a microphone, which are then processed by a computer and treated as controls. Speech input takes many forms and is very configurable. Depending on configuration, a user might make a hissing sound to scroll or read off text for a button to click it. Often, optical character recognition (OCR) and overlay¹ systems are used so that any item on screen can be referenced with speech. A user also can write text via dictation.

Speech input is commonly used by people with motor problems.

Gesture Recognition

Gesture recognition is an input method where a camera and/or depth sensor is used to track body or hand movements without physical contact. Physical movements are bound to computer actions in software, so that, for example, a nod of the head triggers a click. Gesture recognition systems are often used by people who cannot hold or grip physical devices, such as those with severe chronic pain or motor control limitations.

Every child will grow knowing no world other than one where an infinite library awaits them at any moment. Sure, personal computers are expensive now, but with all the efficiencies the internet will allot us, the price will fall swiftly. Think of the productivity increases. No more waiting days for responses in the mail, or having to wait for the morning newspaper to hear yesterday’s news, or having to make expensive long-distance phone calls during business hours. The length of work weeks will fall as productivity increases reduce the time required for labour. Entire industries will transform, and automation will position people to do not stressful work under difficult circumstance but, instead, fulfilling work under positive contexts.

We will be able to connect on a scale as yet unseen. Love and hope among humankind shared the world over and perhaps eventually even with other worlds entirely. With everyone connected at once without limits, borders will fall. A new age of global relations will blossom and bloom. Xenophobia may take a while to fall, but I’m sure it will, for a world where we interact with everyone from every background at every moment will be one where trivialities such as race, gender, sexuality, and ethnicity are irrelevant.

Books will become antiquities, as digital goods take over physical ones. A digital page can have the text expanded to sizes appropriate for the reader’s sight, and images can be expanded for inspection. A ‘book’, in the concept it will be known, won’t be limited to static material but will instead include interactives and videos. Concepts and ideas will be hyperlinked, such that a network of interconnected pages will be established. Project Xanadu and other display means like it will change how we view, interact, visualise, and manage content.

Entertainment will grow more ambitious. Individuals will have the knowledge to create and the means to distribute, such that limits are no longer for their creative realisations. They will be able to establish themselves independently on a global scale for a global audience. Enterprises will be inundated with a wealth of new self-taught talent, leading to greater ambitions in the face of competing with the swathes of indie genius. New and groundbreaking media will be created with quality as yet unseen, out of necessity to compete.

Everyone will have a website – a special type of place on the internet where a person can upload and publish things – to interact and grow on the web. People are already establishing their presences and carving out their own cyber homes. They’ll all be linked and organically self-categorising, so discovery shall be no blocker. Some websites will be dedicated entirely to the categorisation and curation of great works, but no one curation site will dominate. Ideas and concepts will move so fast, and the lack of an initial centralising source will breed a thousand self-sustaining niches capable of thriving in isolation and association.

Wrongs will be righted, as every website exists on an equal playing field, and citizen journalism explodes with power and ubiquity. Every injustice will be documented by a chorus of voices and amplified by a magnitude more. There will be no monopolies, for there will be no monoculture. Everyone will specialise and experience their interests, such that they are fulfilled. The open-source movement which has been gaining traction will explode, as anyone anywhere can contribute. Software will become a collaborative, living canvas, freed from corporate gates and improved daily by the global collective. Perfection will be reached towards out of passion.

A thousand thousand colonies will swarm around the hive and between others in a rush to create and polish. The goal is to make things better for everyone. Everyone will share what they know, however they can. Knowledge will not be gated, but open to anyone. With the sharing and accessibility of information, everyone will upskill. In a world where everyone can know and learn without hindrance. Uncommodified information and communication.

The barrier to entry will be a device and a connection. The devices will evolve, becoming a part of every activity and endeavour, and the connections will improve. Connections will be faster, will carry more data, and will become more widely available. Eventually, there will be no tether, as infrastructure is established and wired becomes wireless, allowing anyone anywhere in the whole world to log on.

Natural disasters and events of terror will be avoided, for the threat can be broadcast with ease and everyone will be elevated to conditions such that they don’t feel they must resort to extremes. Those that are bedbound or limited, unable to physically access the world, will be reconnected with it. They might be unable to go to where they wish, but they can visit through cyberspace. Doctors will be able to assess situations remotely, for people far away or restricted in movement. The rapid evolution of the technology will bring more people power and be the great bringer of equality.

Of course, every technology carries with it new abuses, new threats, and new cruelties. Yet even so, the promise outweighs the fear. The foundation is open, and collectively regulators and we as a people would have to allow the closing and control. We would have to roll over and allow their foot to crush us, and that is something we will not do, for our independence and control are our mostly strongly held values as humans.

The internet, and the web that it is establishing, is in its juvenile infancy. I’m certain there will be some pains, but this is sure to be big. It is certain to change the world, and I welcome the change with open arms. I’m excited for this great new world, even if I may only live long enough to see the first reaping of crop. The internet is sure to be a greatness and I welcome its advancements.

In media, you often see quick and easy injections. Take James Bond or The Hunger Games. Both are popular works which depict the protagonist being swiftly implanted with a chip deep into the forearm. Both are also pure works of fiction, with many outlandish concepts. Implants, as they and other media depict, are no more anchored in reality than the smart blood and nanobot viruses also seen in Bond or the fantastical prosthetics and mutant animals also seen in Panem.

Technology

When you see microchips, such as the ones placed in pets or the ones offered by companies like Dangerous Things, you’re seeing passive chips. They have no power source of their own and only operate when interacting with a close-range reader.¹ Microchips cannot remotely track you.

The first thing to consider is power. How do you power a tracking chip? A tiny little battery? Some sort of kinetically charged solution like you sometimes see in watches? Location tracking is notoriously power-intensive, so a small battery is sure to run out of charge swiftly. Implantable devices like pacemakers have large inbuilt batteries, and even Apple’s AirTags – which are also of considerable size, mind you – only have an expected battery life of one year.

Then, you must consider the complexities of broadcasting a signal. Even if you do somehow have power, how does the tracking get done and then sent to the malicious actor? You need an antenna of some description, which certainly takes up more space, and then must transmit/receive a signal through your flesh and then possibly your clothing and then possibly through a building, avoiding all forms of interference along the way. Humans are made mostly of water, and water blocks high-frequency radio waves incredibly well. Even just holding a dedicated GPS with a significant internal antenna incorrectly can hinder its accuracy, and location tracking within buildings is a notoriously tricky problem. Burying an antenna inside your body completely suffocates the signal.

Further, the more complex these supposed implants become, the larger they’re required to be. You can make some pretty tiny tech, but even the smallest powered trackers are still of decent size. If we hypothetically say that tech itself manages to be impossibly small, a coating is still required and needs to be of a reasonable thickness. Coatings are needed to prevent the implant itself from interacting with one’s own flesh, and they must be of a thickness such that they both provide a sufficient barrier and are impervious to expected impacts and forces. For microchip implants, glass or resin is usually used.

Tiny little chips (especially ones doing complex work like real-time tracking) also get really hot. Implants are also extremely visible in x-rays – even shining a bright light through your flesh will usually show them off. The idea of a small, working, discreet, tracking microchip is nothing more than science fiction. If real-time tracking in a small implant were viable, they’d be rather popular with pet owners for finding lost animals in the same way close-range identification chips are.

Healing

Let’s suppose that some impossible chip becomes possible. Somehow, a tracking chip is made at the scale of currently available commercial NFC and RFID offerings. In such a case, it must still be implanted, which isn’t something that can be done discreetly. To implant something, an incision must be made. You need to make a hole in the skin and a pocket in the flesh so that an implant can be placed within it. You can’t just teleport an implant into somebody.

The larger the implant is, obviously, the larger the required pocket and opening must be. A safe placement must also be found. Microchip implants are always done rather shallow (which means you can usually visibly see and feel them). If you go too deep, they stop working. If you don’t go deep enough, your body forces them out. In general, it is an effort to stop your body pushing out an injection. Wounds heal in such a way that they zipper out anything stuck inside.

Your body doesn’t want to trap things inside your body, so trapping only happens in the case of material forced so deeply and with such a minimal entry hole that the wound heals before it can be pushed out, as sometimes seen with bullets. Even once a wound is healed, the body will still attempt to push foreign objects out. There are many cases of metal fragments working their way out of people decades later.

Healing takes multiple weeks. I covered the healing journey of my xG3 v2 bio-magnet. My hand swelled up, and I had to use medical strips to hold it in place – even then, my body pushed it out a bit towards the opening. My hand turned colours of yellow and green around the implant site, which it remained for multiple days, and it hadn’t healed back to normal until weeks afterwards. Even then, the entry point of the syringe was still not completely healed, and I’ll have a small but visible scar there for the rest of my life. All that was for my positively tiny xG3 v2, which is only 3mm by 15mm. Anything larger or deeper requires a scalpel incision and takes much longer to heal.

An implant is not installed casually or discreetly. Healing takes time, and implantation is an effort too. You need a sterile environment with a professional capable of an install. It isn’t happening subtly.

Viability

As we’ve established, the idea of an implanted tracker microchip is pure science fiction and falls apart in multiple places. However, let’s again suggest that by some marvel it is possible. Why would some super powerful entity choose microchips over any other less intrusive and more technically viable system?

You likely are being tracked everywhere you go, though not by a magical implanted tracker that exists beyond the bounds of known technology. There are more subtle methods, like cameras with facial detection and even gait recognition, your card payments at stores, your public transport usage, and the likes. Then there are your own devices. Your phone connecting to local towers or communicating with GPS systems and your car likely doing the same. Otherwise, social engineering. One has much less privacy than they think. You might share content which hints at your location. A single photo, even one which to the untrained eye seems innocuous, can be used to place you exactly.

For what reason would an adversary with technology beyond the bounds of known science spend such resources on a medically risky and complex, physically limited, and easily detectable surgical implant with healing periods when they can use open-source intelligence (OSINT) tactics for negligible cost? The infrastructure around us is already doing the work for free.

Recall Ockham’s razor: ‘When presented with competing theories, the simpler explanation is to be preferred’.

If you think you’ve been microchipped against your will, that is incredibly unlikely. You should seek professional guidance from a registered medical expert. You can reference Dangerous Thing’s So you think you’ve been implanted against your will? for further reading. Do not message me if you’re having delusions about being microchipped or having your mind read.

Know The Problem

You should ask for the problem ahead of time, ideally while booking the appointment and the issue is fresh in the client’s mind. People are forgetful – the elderly doubly so. I have found clients on more than one occasion befuddled as to exactly what they needed help with.

If possible, have them text through their problem with as much detail as possible. Otherwise, call them and ask them questions (ensuring you write it down). Even just knowing the general realm of the problem is useful for getting the process started. Otherwise, run through the typical list:

1. Malware
2. Email
3. Printer/Scanner
4. Microsoft Office (Word, Excel, PowerPoint, etc)

Help Articulate

Your job isn’t just to fix problems; it is largely to understand what the problem is in the first place. If someone is having trouble with simple functions of their computer, there is a decent chance they don’t know all appropriate technology. You must explain jargon, for throwing around technobabble is useless. Much like when speaking to someone in another language, if there is a communication barrier you should be ready to use different terminology to convey a point.

Refer to it not as ‘Google Chrome’ but as the coloured beach ball and not as ‘settings’ but as ‘the grey cog’. Icons and colours are more memorable and identifiable. These associations are much clearer than terms which are undecipherable without context. Always prioritise clarity over technical accuracy.

If receptive, you can gently educate the client on the correct terms. Don’t force it, however. Some people don’t want to learn; they just want the problem to go away. I’ve personally found a lot of older folks want to understand computers but think they’re too old to comprehend them. If you work through it with them, they’ll probably appreciate it.

Some people are very curious and wish to learn. Since you’re going through the motions anyway, it is worth explaining the steps you’re taking. You might educate them such that they can fix the same problem occurring again in the future, thus removing a chance for you to get another call back, but you make up for this by establishing a rapport with the client that will make you their first pick if another problem occurs.

Provide More Help

While you’re working on a computer, there can be periods of dead air. Waiting for something to install, for software to update, for files to delete, etc. If someone is struggling with their computer in one way, there is a decent chance they’re struggling in multiple ways. Ideally, ask at the head end of the job so you can fill up this dead time with solving their other problems.

My go-to is a quick disk cleanup and checking that software updates are installed. If those are already done, I’ll write down some documentation for the client which they can reference and refer to when fixing problems in the future. It can also be a good opportunity to pick some fluff out of charging ports, speakers, or fans, or to install a remote access solution.

Configure Remote Access

It isn’t rare for someone to recall something they meant to ask or request after the fact, or to need help with a quick or minor issue. For such cases, a remote desktop setup can be very effective. You can save a lot of travel time by performing remote support. My personal go-to is RustDesk, though any remote desktop solution will work. RustDesk can be self-hosted, and it is trivial to apply your own branding.

A remote desktop is also effective in cases where you are interacting with people who use eye-tracking or other assistive input methods, even if you’re physically present in the same location as them. Physically interacting with their input device often means making them uncomfortable and taking away their means of communication, even if temporarily. It’d be like someone reaching over and disabling your mouth so they can fiddle with it. On top of that, it can take a lot of time to get some input methods tuned in, so disrupting their device can be a major inconvenience.

It can be tempting to exclusively rely on remote device control, but in-person interactions are valuable. Especially the first interaction should optimally be conducted in person, as it builds a lot of trust. If you’re troubleshooting physical items such as printers or scanners (I hate printers), make sure that is in person too. You can’t remotely plug in a USB cable or press the reset combination on a printer.

Tools

If you’re doing tech support, there are some tools you should possess. A Flipper Zero or another BadUSB device can also be very useful for running commands or actions. A set of precision screwdrivers can come in handy, both for dismantling parts of devices if necessary and for more menial tasks like digging crap out of a device’s orifice or general MacGyver behaviour. You’ll want a variety of cables, as you should never assume anyone charges their devices, much less that they know where their device charging cables are. A pen and notepad (especially one with perforations so you can break off pages) are very useful for leaving details.

You’ll also want some USBs. You’ll want some blank ones for backing up or copying data (keep these in a state such that you can sell them to a client) and some with utilities such as memory checkers and Linux live USBs. Live USBs are great for checking if a device problem is a hardware fault when Windows is acting up. If you’re so inclined, install media for the latest version of Windows is also a pretty good idea.

Using a laptop for searching seems much more professional than your phone. Of course, most laptops don’t have their own cellular plans, and you shouldn’t expect a connection at a client’s home. Bring something capable of providing a mobile hotspot. Sometimes you’ll find yourself on your back under a desk fishing for a port or loose cable, so a flashlight is a must-have (the one in your phone should do the job).

Never Assume

Don’t assume anything. As the saying goes, ‘to assume makes an ass out of u and me’. There will forever be some random person who keeps their most important files in the recycle bin for safekeeping. You must never put too much faith in people using their computers correctly.

Likewise, don’t rearrange or modify anything more than is necessary. People settle into understanding that a program is located in a certain position on their desktop. Don’t move it. Every change breaks someone’s workflow.

Store Client Details

Set up a secure place to store client details, and ensure you communicate that you’re storing them to the client. People sometimes forget their own emails, or you can forget their email. Especially if you’ve just created a new address for them. Don’t expect anyone to remember anything. Write it down. A lot of people will go off to perform other tasks while you are addressing their problem. Bringing them over to request their phone number or email or code or password or whatever else is a hindrance that stretches out the interaction and inconveniences you both.

This information also makes booking appointments quicker and more seamless. Asking a repeat client for their address each time they’re booking an appointment is repetitive and suggests that you’re either disorganised or not caring about the client. A much nicer interaction (which saves a back and forth) is ‘Are you still on Street Lane?’. It is still important to ask as people change addresses.

Ignore Adult Content

You will probably discover NSFW content. Whether that be saved nudes or pornography in the browser search history, you’re best to just not mention it. Be professional and gloss over it. People find it embarrassing, so you should move on quickly. I’ve worked on a lot of personal computers, and regardless of age, people have NSFW content saved. If they’re viewing their history or saved photos, look away if appropriate to do so.

If you’re dealing with malware from a malicious adult site, edge around the topic while still informing them of the risks posed by less scrupulous sites. How much you avoid the topic depends on your client. Some people will get very defensive and dismissive if you mention their consumption of pornography, even if in a non-confrontational way. One method of skirting the topic is to mention dodgy download links on gambling websites and other ‘mature’ sites.

Full send it, I say. I went and picked it up locally, and the seller even threw in a pristine hardshell case. Lovely.

The Patient

The laptop is a 2020 13″ MacBook Pro. It is the model with two Thunderbolt 3 ports. 1.4GHz quad-core 8th‑generation Intel Core i5 with 8GB of RAM and 256GB of storage, purchased in August of 2020. Physically, it is in mint cosmetic condition without any dents or scratches.

Probing the seller as to the origins of the befallen, I got this morsel of knowledge:

Got this laptop from a friend. As for the screen, the ribbon connector failed within 2 weeks of them buying it, but got it fixed under warranty. Then it broke again a few years ago, and was given to me. The laptop also might have some motherboard issues, as it keeps crashing-restarting every ~30 mins, and saying it had a kernel panic. It’s in almost perfect cosmetic condition, though.

If the ribbon connector has failed, that could be the root of every problem. If it was shorting or otherwise failing to communicate, that would explain the random crashes.

The first port of call was seeing how it worked. I plugged it in, and sure enough it sprung to life. Neither the primary screen nor the Touch Bar displayed anything, but everything else did work. It was running macOS Catalina version 10.15.7. Funnily enough, I ran it for a few hours and didn’t encounter any panics. I was prepared to deal with one every half-hour, so the absence was very welcome.

Apple has dropped support for the two Thunderbolt version of the 2020 MacBook Pro, so the latest officially supported macOS version is Sequoia, which I proceeded to install. The fans whined through the entire install process, and it presented some incredibly pretty colours in the form of graphics corruption. Some of it reminded me of old TV static, though that might have been my ageing monitor’s analogue input support. When the Apple loading screen began to arbitrarily cycle through different colours, I found it most mesmerising.

I can’t say I’m too enamoured with Liquid Glass from a general design and usability perspective (though I do find the refractive glass technically stunning), but I’ve admired from afar the design language introduced in Big Sur and refined through Monterey, Ventura, Sonoma, and finally Sequoia. A lovely balance of usability and class.

Fixing It

Despite not encountering the every-half-hour-crash issue myself, the logs did present themself when I first turned the MacBook on, so I inspected them. The relevant sections were:

panic(cpu 0 caller 0xfffffff0220736d4): userspace watchdog timeout: no successful checkins from com.apple.dfrd since load
service: com.apple.dfrd, no successful checkins since load (180 seconds ago)
service: com.apple.remoted, total successful checkins since load (180 seconds ago): 18, last successful checkin: 0 seconds ago
service: com.apple.bridgeaudiod, total successful checkins since load (180 seconds ago): 19, last successful checkin: 0 seconds ago
service: com.apple.logd, total successful checkins since load (180 seconds ago): 19, last successful checkin: 0 seconds ago

CORE 0 is the one that panicked. Check the full backtrace for details.
CORE 1: PC=0x00000001ae095c44, LR=0x00000001a961b558, FP=0x000000016d834490
Total cpu_usage: 13758820
Thread task pri cpu_usage
0xffffffe19a7ac600 watchdogd 97 0
0xffffffe19a919a00 bridgeaudiod 37 0
0xffffffe19a8b8200 watchdogd 31 0
0xffffffe19a12bc00 kernel_task 0 3808968
0xffffffe199ed8600 kernel_task 0 6486281

Panicked task 0xffffffe19a798000: 202 pages, 3 threads: pid 36: watchdogd

This is, believe it or not, good news. Good in that it can be dealt with. The dfrd mentioned stands for Device Function Relay Daemon, which is the software service that manages the Touch Bar. The failing checkins are being handled by Apple’s infamous T2 Security Chip, which is failing to make its connection. This could be three things:

1. Ribbon cables are making a poor connection or are otherwise damaged.
2. The T2 Security Chip is borked in some way, causing problems communicating.
3. The Touch Bar and screen themselves are physically broken.

The easiest first approach was to reset the NVRAM (⌥ Option + ⌘ Command + P + R) and the System Management Controller (SMC) (^ Control + ⌥ Option + Shift + Power). This could have helped the components sort themselves out. It didn’t help, but it was worth a try.

Inspecting the display panel itself under a bright light, I could see no damage. The Touch Bar also flickers intermittently, indicating it might be alright from a hardware perspective and that its unfunctionality might just be a software issue. Experimenting further, the Touch Bar seemed to successfully and reliably trigger input when touched, despite not displaying anything. My working theory was that it was a firmware issue and that it’d be worth attempting to revive the Mac firmware. However, this is my only Mac, and a second is required to facilitate the ‘revival’ process.

I figured running Apple Diagnostics was worth a shot (D during start-up), and that gave me the note ‘There may be an issue with the display. Reference Code: VFD001’. Thanks Sherlock. Alas, such an error almost guarantees it is a hardware issue and killed my hopes of it being a software fix.

Given the effort to fix it, I’ve chosen to leave it in its current state. In the future, I’ll consider repairing it. Before buying the Mac, I pre-emptively purchased a set of the asinine screwdrivers Apple mandates to gain entry into their devices on account of hating consumer repairs. It was a reasonable $15 for a set including all the bits I needed and some extras that’ll prove useful.

Using the Touch Bar

Given that the MacBook’s Touch Bar is functional with the exception of failing to display anything, I refused to write it off entirely. That’s valuable keyboard real estate, and I happen to be one of the half-dozen people on this planet who actually rather likes the Touch Bar. To make it at all usable, I need to see what is on it.

Here are the methods I’m aware of and have used to display the Touch Bar’s display upon the screen, ordered from most to least preferential:

1. Zoom
   Enabled in System Settings > Accessibility > Zoom > ‘Touch Bar zoom’, it provides a zoomed-in view of the Touch Bar that spans the bottom of the display whenever the Touch Bar is touched. When not being touched, it fades completely from view. It also shows a cursor (or cursors) to indicate where it is being touched. You cannot interact with the Zoom display with your cursor.
2. Xcode
   Accessed by opening Xcode, then Window > Touch Bar. This lets you simulate either the 1^st-generation Touch Bar with Esc or the 2^nd-generation Touch Bar without it. This works pretty nicely if you’ve always got Xcode open, or you’re happy with the overhead of leaving it open in the background. You can fully interact with the cursor.
3. Switch Control
   Enabled in System Settings > Accessibility > Switch Control. Toggle on Switch Control, then select ‘System’ and ‘Toggle Touch Bar’. The Touch Bar then appears at the bottom of the screen, and you can interact with it via the cursor. You can drag the Switch Control panel to the very edge of your screen to hide it, and you can also fiddle with the settings to lower the panel’s opacity like it isn’t even there. This approach is my least favourite, as one must deal with the Switch Control panel, and the panel displaying the Touch Bar is immovable and always anchored to the bottom of the display.

Unfortunately, and rather annoyingly, a lot of older solutions for simulating or displaying the Touch Bar are non-functional in newer macOS versions (roughly Catalina onwards) due to private hooks being deleted.

On my Touch Bar, I’ve set up the AirPlay button to the rightmost location, next to the Touch ID/power button, so I know that I can always press that location and then press the relevant display on the left-hand side, just slightly in from the edge where the cancel button is. It isn’t optimal, but it is low risk and allows me to connect the Mac to a wireless display blind with a bit of luck and without too much faffing with custom software.

Purpose

I needed this Mac mainly as a Safari testing machine. It isn’t an Apple Silicon model, so its days are already numbered, but it’ll be useful for compiling some software for macOS for a brief little while. It will also run some simulators, let me test VoiceOver, allow me to open, edit, and inspect documents from the iWork suite, and facilitate the creation of Liquid Glass icons in Icon Composer.

It is a thoroughly imperfect machine. Once it has run its course, I’ll consider chucking Linux on it, keeping it around to deal with old Apple devices, or selling it for parts. No, it isn’t a great computer, but for just $25 I can’t argue with it. That is cheaper than a single month of a subscription to most browser testing services.

What was a deeply technical community with generally agreed-upon terminology was largely burst open with the wide release of ChatGPT. Suddenly, a lot more people were exposed to modern AI, and marketing-led commentary thrust a deeply technical topic haphazardly into simple terms. Now we’re stuck with a linguistic debt.

I’d suggest part of this is due to resources. With most major systems, progress is slowly made with resources coming along with it as well, and then a takeoff happens. With AI, the takeoff happened concurrently with the creation of resources aimed at the general public, so natural selection didn’t have time to occur before adoption.

Te ultimate purpose of any jargon is to simplify communication. Rather than saying something super complex and informative, such as ‘Large-Scale Transformer-Based Natural Language Pattern Recognition and Generative Model’, we say ‘large language model’, and rather than ‘large language model’, we say ‘artificial intelligence’, and rather than ‘artificial intelligence’, we say ‘AI’.

The International Standards Organisation has made an effort to standardise terminology in ISO/IEC 22989, but things are still lacking, and confusion is common.

AI

AI in general is poor as a term. It is pointlessly broad. The term ‘AI’ is used to refer to the non-player opponents in Super Mario Kart and ChatGPT with little distinction. Some people go to the lengths of always specifying ‘generative AI’, but even that is poor. Most AI is arguably generating something.

AI is a buzzterm that companies want to see, so it is applied everywhere without discretion. What does it even mean for a washing machine to have AI? At a certain point, a word is so broadly used that it doesn’t mean anything.

LLM

‘Large Language Model’ is a specific term but is used generally. When people are discussing LLMs, they often mistakenly refer to the interface itself. For example, ChatGPT, Claude, or Gemini. These are offerings which house multiple models of differing types. They’re presented as a single interface, but behind the scenes there are multiple parts at play.

Commonly, people are not referring to LLMs in isolation but instead wrapped up with other technologies, tools, or models. This is messy, because there is Anthropic’s Claude, the interface, and Anthropic’s Claude, the set of AI models including Claude Opus, Claude Sonnet, and Claude Haiku. Likewise, there is Google Gemini, the interface, and Google Gemini, the set of AI models, including LLMs, vision models, and image generation models.

To use ChatGPT as an example, you’re interacting with an interface. ChatGPT itself is not a ‘model’ per se but instead a wrapper to interact with models¹. When you make a request in ChatGPT, your message is fed to a model router, which determines which model it should be forwarded to. GPT-5 is architected to contain multiple different levels of ‘thinking’. There is also a step where the result is evaluated and moderated.

It can also call on ‘tools’, such as tools to query the internet, interact with external applications, or generate images. When it generates images, it takes the user-provided prompt, runs it through an LLM to understand how to handle it, then feeds it through the actual image generation model.

The term ‘LLM’ doesn’t adequately describe the multi-modality of most AI systems in use today. Commonly, one model can handle not only text but also other media inputs. Different terms have been proposed, such as LMM (Large Multimodal Model) or Foundation Models, but their usage is uncommon.

It also doesn’t really specify the size of the model. ‘Large language model’ is the commonly used term for language models of all sizes. Small Language Models (SLMs) are models that are smaller in scope and scale than LLMs but are often regarded under the LLM banner for ease of description. Just ‘language model’ would be the most fitting title but has become a bit generic. A very basic Markov chain is arguably a language model, but no reasonable person would consider them anywhere comparable to even a small language model. The issue is only exacerbated by the reality that the size of models is ever-increasing, so what used to be an LLM could be argued to be a SLM now. There is no defined cutoff.

AGI & ASI

‘Artificial General Intelligence’ and ‘Artificial Superintelligence’ are popular terms for describing advanced AI systems, but everyone seems to have different criteria.

The AI effect is alive and well. If you showed someone in the 1950s our current technology and an LLM, they’d probably take it as full computer intelligence. It’d likely be difficult to convince them otherwise, even when presenting them with details. Does this make our current AI AGI/ASI?

Maybe? There is no real widely agreed upon definition. One person’s AGI is another person’s GPT-3. ISO/IEC 22989 defines it as a ‘type of AI system that addresses a broad range of tasks with a satisfactory level of performance’, following it up with the note that ‘AGI is often used in a stronger sense, meaning systems that not only can perform a wide variety of tasks, but all tasks that a human can perform.’.

I’d suggest that AI systems have been able to address a ‘broad range of tasks’ to a ‘satisfactory level of performance’ for a while now. The addressing of a potential interpretation that AGI can perform ‘all tasks that a human can perform’ opens further questions. Humans are physical, and many tasks we do involve physical interactions. I interpret this as the definition meaning that AGI would require a physical manifestation.

The October 2025 paper, A Definition of AGI, attempted to come to a definition. They put forward the definition: ‘AGI is an AI that can match or exceed the cognitive versatility and proficiency of a well-educated adult.’ I take issue with this. How does one measure cognitive versatility and proficiency? Humans have tried to quantify intelligence with measures such as IQ, but it is an impossible task. IQ is a (rather flawed) measure, and having a high IQ does not necessarily indicate being well-educated. What is a well-educated adult? There is so much variation in people.

A highly educated person is rarely highly educated in many fields and certainly not in all fields. AI models can fail simple spatial and logic puzzles, yet solve complex mathematics problems that humans cannot manage. A person can be poor with computers but extremely skilled at writing. ‘Well-educated’ is an extremely subjective measure to the point of being useless as a checkpoint or definition.

Vibe Coding

Andrej Karpathy coined the term ‘vibe coding’ in early 2025:

There’s a new kind of coding I call “vibe coding”, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. […] I ask for the dumbest things like “decrease the padding on the sidebar by half” because I’m too lazy to find it. I “Accept All” always, I don’t read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it. The code grows beyond my usual comprehension, I’d have to really read through it for a while. Sometimes the LLMs can’t fix a bug so I just work around it or ask for random changes until it goes away. It’s not too bad for throwaway weekend projects, but still quite amusing. I’m building a project or webapp, but it’s not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.

As the coinage gained popularity, it transitioned into being used by some people as a catch-all term for any LLM-assisted development. This is a case of semantic broadening, which is typical in language, though this was a particularly swift instance.

Many people use ‘vibe coding’ in its original definition, many people use it just for LLM-assisted development in general while being aware of the original definition, and many people are unaware of the original definition at all. Like much of the rest of AI terminology, this makes the term a mess in usage, with different people interpreting it wildly differently. Those who use AI for assistance in writing code are often branded ‘vibe coders’ just the same as those who slop out all their code with AI without checking it.

Agent

‘Agent’ and ‘agentic’ have been the follow-up buzz terms in tech circles following the saturation of ‘AI’ as a term. Generally, an agent is presented as an AI system which takes action like a human using human mediums, but as with seemingly all AI-related terminology, the specifics are fuzzy.

Simon Willison describes agents by saying: ‘An LLM agent runs tools in a loop to achieve a goal.’ I think this is the best definition to date, but there are still many other interpretations out there, and people can opt to apply it in other contexts.

‘AI Agent’ is the buzz phrase that one-ups ‘AI’, so marketing sees that it is applied widely and broadly. Much like the term ‘AI’ itself, ‘Agent’ has been applied so liberally that it has lost meaning. ‘Agentic Toilet’ is meaningless technobabble marketing speak.

Open-Source & Open-Weights

There are some open-weight AI models, which can be run and used locally on one’s own compute. This is often confused with a model being open-source. Some people state that open-weight models are open-source, mistakenly arguing that open-weight is the closest thing.

Open-source models do exist, though are fairly uncommon. They include the training data, the training code, the full weights, and a permissive licence. If you wanted to, you could dive in and modify an open-source model at any step during its creation. Open-weight models are much more common but only provide the weights. Your only options for modification are upon the final result, and you’re more limited regarding what you’re permitted to do.

It is the difference between receiving a pre-baked cake you can decorate (fine-tune) and eat (host) as you wish, but whose creation remains a mystery, versus receiving the recipe, the raw ingredients, the kitchen utensils, and the total freedom to bake it yourself from scratch.

The Open Source Initiative has tried to clarify the difference by publishing The Open Source AI Definition and clarification on open-weights to middling success in swaying the zeitgeist.

Anthropomorphising

With a technology that is built to replicate (or replace) and mimic human output and intelligence, it is no surprise that people treat it like a human.

Read any mainstream media coverage, and you’ll see AI offerings treated as humans. In 2025, when Grok was called out for generating non-consensual sexual imagery en masse, including of minors, news outlets turned directly to the AI model for comment. It may speak like a human, but an AI model is not one. To quote the 1979 IBM Training Manual: ‘A computer can never be held accountable. Therefore a computer must never make a management decision.’

If we’re taking issue with anthropomorphism, then machine learning is a misnomer as well, as machines cannot ‘learn’. ‘Training’ fails for the same reason. Thinking and reasoning models cannot ‘think’ nor ‘reason’ in the human sense. They cannot ‘hallucinate’ – ‘confabulate’ would be a more apt term but somewhat falls victim to the same anthropomorphism.

We use biological metaphors (learning, neurones) for mathematical processes (gradient descent, weighted matrices), which invites the public to misinterpret maths as magic. We reach for biological metaphors because they are what we experience ourselves, are easy for the layperson to understand, and have applicable terminology, even if not entirely accurate.

Companies have much to benefit from AI being anthropomorphised. Models are intentionally designed to drive user engagement and build emotional connections. Not only are having ‘emotions’ and ‘opinions’ beneficial to human interaction with AI models from a utility perspective, but they also keep users engaged. Many people consider themselves to be in relationships with AI and are distraught when their ‘personalities’ are altered. They are locked in, because to leave means to ‘kill’ their partner and their bond. Even people not in that deep will still speak of their favourite model with affection. Claude (which is a human name) is often spoken about like a development friend rather than a development tool.

Even if we were to collectively act against anthropomorphising machines, then what terminology do we use instead? It is extremely difficult to succinctly describe these words we use for human experiences, given that all our language is designed specifically around the experiences of living beings. Do we create new language entirely to address these points? I think we’re so far along now that even if new terms were devised, it would be near impossible to transition people to them.

Terminology is messy. There isn’t a way around that. Language is a flowing, moving, undulating mass of entangled ideas and concepts. To attempt to straighten it and force it to conform to a design is a fool’s errand. However, if we can’t come up with specific definitions, then it makes regulation and policy difficult. It spurs miscommunication. It makes every explanation longer and requiring of a definition lest confusion arise. Even in technology circles, terms are used fast and loose. What hope is there for people outside the tech sphere to understand?

I think the only way this would realistically be cleared up is if major AI companies collectively agreed upon given technology, but that seems extremely unlikely. You certainly aren’t going to convince people to stop anthropomorphising AI.

This site, along with many other sites I’ve built since becoming familiar with the project in late 2024, is built with Web Origami. It has quickly become my favourite way to build static websites and work with and transform data.

The project itself is a tad difficult to simply explain succinctly. I could try to describe it based on what it can do. I could call it a static site generator, or a data transformer, or one of a number of other things, but that doesn’t really explain it satisfactorily. Origami is extremely versatile and powerful. It uses a content/transformation model that allows massive potential yet remains simple and easy to start using.

The project is split into a few distinct parts, or ‘aspects’, as the creator Jan Miksovsky describes them. The aspects I value and use most as a person who largely uses Origami to create statically generated websites are the Origami dialect of JavaScript and the built-in functions.

Origami is described as a dialect of JavaScript, rather than a language in and of itself, because it is, at its core, a super convenient JavaScript expression system with paths. It is extremely concise to write without being difficult to read and feels a bit magical. For someone who knows JavaScript, it feels intuitive and is exactly what you’d expect from the language, and for someone reasonably tech-savvy but unfamiliar with JavaScript, it is intuitive enough to understand and start writing quickly.

Origami is thankfully not magic in the sense that it abstracts things away and does some unknowns behind the scenes that you just have to trust. Instead, it is magic in the sense that it just seems to work. The exact thing you’re trying to do is the intuitive thing, and Origami has a clear and obvious way of achieving any goal you can throw at it. How Origami interprets and manages your code behind the scenes is as clear as glass, and the documentation goes as far as to explain not just what to do to achieve your desired outcome but how and why the given approach or method works.

Simple Structure

The thing one must understand about Web Origami is that almost everything is a tree. So much so that the project was previously called TreeOrigami.

Let’s say that we have a few markdown files in a folder, each with some YAML frontmatter as is typical for a developer blog. We’ll assume the contents of each file looks like this:

---
title: Blog Post 1
date: 02/08
---

This is the contents of a most informative and wonderful blog post.

We can get the title of each post with Tree.map(blog-posts, (post) => post.title). That is descending into the folder and markdown files within it to pull out the value of the YAML title key in the frontmatter. It’d return this:

post-1.md: Blog Post 1
post-2.md: Blog Post 2
post-3.md: Blog Post 3

When everything is a tree, it is easy to work with content. Directory full of markdown? That’s a tree to work with. YAML file? That’s a tree to work with. Want a YAML value from the frontmatter of one of those markdown files from that directory? Just descend down the tree. There is no complication because you’ve got two different types of data that you’re working with – they’re all interoperable trees.

Origami is designed to handle data, and you’re given the tools to work with that data by the simplest means possible. There are no arbitrary limitations or opinionated architectural decisions one must fight with. You’re making your project, and your approach and decisions are what matters. Origami is very open and flexible in how you use it. You’re not given a template to fit your project into. You build a template for your project which Origami fits into.

If an Origami-based approach is too complex for your needs, or you need to do something really bespoke, you can write your own JavaScript functions and call them directly. Via this method you can develop arbitrary reusable data transformations or whatever else you might need. I’ve used functions to count the number of words in an article, to transform dates into different formats, and to turn a YAML file defining events into annually occurring displays on a larger page, just to name a few of my uses.

Builtins

Origami also comes with a huge number of built-in functions that you can call. Separated into four top-level namespaces: Dev, Origami, Protocol, and Tree, there are so many useful actions.

Builtins in the ‘Dev’ namespace are for developing and debugging Origami projects. You’ve got commands including Dev.changes to compare differences and Dev.serve to start a local web server. In the ‘Protocol’ namespace you’ve got tons of handlers for various URL schemes such as Protocol.https and Protocol.files, allowing you to fetch data from remote locations. In the Origami namespace, you’ve got lots of general utilities useful for building things, like Origami.mdHtml for converting markdown to HTML and Origami.slug for converting input to a safe string suitable for URLs. Then, you have the Tree namespace, which has so many utilities for working with trees: simple commands like Tree.first for getting the first value and more complex commands like Tree.calendar which returns a structure for years/months/days.

I’ve only touched on a scarce few builtins there. Hundreds exist, each with useful functionality. Builtins are easy to use, too. If you have, for example, a folder called ‘posts’ with a bunch of items you want to randomise the order of, you could do so with this Origami code:

posts
-> Tree.shuffle

Building a site with Origami is just stringing together these simple transformations into something larger. Origami is designed such that there is no limit to the potential complexity at a macro level, but at a micro level everything forever remains small, modular, and composable. The ceiling is high, but the floor remains low.

Not Just Sites

Origami is extremely good as a static site generator, and that is all most people use it for, but that isn’t its only purpose or function. Origami is applicable in many other contexts, including within the terminal.

Want to convert a PNG into an AVIF? ori "Origami.image.format(image.png, 'avif')". Want a graphical representation of the structure and contents of a directory? ori "Dev.svg(src)" > folder-visual.svg. Have a YAML file that you’ve just realised has all the data sorted the wrong way around? ori Tree.reverse films.yaml.

I’ve always hated faffing around with jq, yq, and other such tools for manipulating data, and often I’d end up resorting to writing a Python script for what really should be just a simple data transformation. Origami allows keeping everything as simple as simple can be.

The async-tree library that is the key to Origami’s capabilities is also completely open for anyone to use, independent of the language and other parts of the project. Of course, this also means that if you’re really pushing boundaries and making complex systems atop or as part of an Origami project, you can import the async-tree library to make your work much easier. This also makes it possible to create complex extensions to broaden Origami’s capabilities.

Extensions

Beyond all the functionality explicitly built-in, Origami has a collection of extensions for integrating with external tools, systems, and services.

I’ve been pretty vocal in my love for the search tool Pagefind and use it on many of my sites, including Vale.Rocks. There is a fantastic Pagefind extension for Origami that makes it trivial to implement.

Some folks have taken to using extensions for achieving integrations with cloud storage managers such as Dropbox and Google Drive so that they can have a polished authoring flow.

Origami is great. I love it, and I’ve used it on many sites. It is worth noting, however, that it isn’t a massive project, at least at time of writing. It is largely developed by a single person, though there are some external contributions and suggestions are always listened to. The community is also pretty small, meaning you might not find information on the web about every single possible pattern and approach. The community which does exist is fairly active, though, and the documentation for the project is fantastic and very comprehensive.

If you’re interested, Origami is worth trying. There isn’t any risk associated with trying it, and I do think there is utility in it for a lot of cases. The potential has barely been tapped.

However, China takes localisation to a whole different level entirely. The country’s regulation and censorship have created what feels at a glance, and even at further inspection, like an alternate universe. So much is similar to what is known by the rest of the world, and yet so much is different.

Regulation

Gaming in China is heavily regulated and has been subject to much government oversight since inception. Given the pervasiveness of online games, crucial to understanding the form China’s gaming industry has taken is understanding China’s almost self-contained national web. Within the country, there are a few companies that handle major games. Most notably Tencent and NetEase.

Tencent is one of the world’s largest companies by revenue and notably operates QQ and WeChat. That is just the very tip of the iceberg, however, as Tencent is among the largest video game vendors. They wholly own Riot Games, Turtle Rock Studios, Sumo Group, and more. They also have majority stakes in Supercell, Grinding Gear Games, and more. They also have minority but sizeable stakes in a plethora of major players, including Epic Games, Ubisoft, Remedy Entertainment, Fatshark, Roblox Corp, and FromSoftware. Further, they’ve got multiple in-house studios.

NetEase is also massive, though not quite to the scale of Tencent. It develops games including Marvel Rivals and has investments in Bungie, Devolver Digital, and others. Especially during the 2020s, many other Chinese brands have gained a footing in the market and begun to see massive growth, such as miHoYo, which publishes globally via Cognosphere trading as HoYoverse.

Regarding hardware, in 2000 a ban was enacted on imported game consoles, leading to variants and partnerships such as Nintendo consoles branded under ‘iQue’². This ban caused damage to the home console market in China and a flourishing of the PC and later mobile gaming markets. Though the ban was lifted in 2015, consoles still remain relatively unpopular, never having gotten a chance for a solid foothold during the market’s formative years. Following the iQue, Nintendo partnered with Tencent to distribute the Nintendo Switch in China, and while handheld consoles like the Switch have seen some popularity, they still don’t rival PC and mobile.

Elsewhere in the world, the vast majority of games released are released to a global audience in more or less the same state that they are in their country of origin. China, however, has extremely different versions of many games. There are periodic freezes on game approvals, preventing them from being published in the country. Notably from 2018 to 2019 and a significant slowdown through the early 2020s, though approvals have been more common since.

Games are expected to adhere to specific values and avoid allowing some immoral or unethical actions, which leads to them sometimes requiring alteration for distribution in China. The National Press and Publication Administration (国家新闻出版署Guójiā Xīnwén Chūbǎnshǔ) is currently the primary force behind this and has a few stipulations against content which:

• Spreads superstition or is deemed too horrific or cruel. This particularly applies to corpses, skeletons and blood, which are often removed, replaced, or recoloured. Cult or occult content that conflicts with state ideology is also restricted. In some cases developers go further than what is necessitated, just to ensure there are no complications due to found infractions.

• Falls outside of traditional gender roles and depictions of non-heterosexual marriage. Depictions of effeminate men are specifically targeted, and relationships are often recontextualised as friendships or removed entirely to ensure compliance.

• Depicts historical revisionism, such as games which depict alternate timelines where China is divided or Axis powers win during World War II. Maps which fail to align with China’s official borders are also noted.

• Includes gambling mechanics, which are very much regulated. Exact drop rates and likelihoods were made required to be published in 2017. This has had especially significant effects on games like Counter-Strike which have lootboxes.

Due to the heavy region-locking and the language barrier, many Chinese variants of games are scarcely accessible, or even known of, outside of China. Chinese variants of games commonly can’t be legitimately accessed at all outside of mainland China. You often need to provide state-issued ID to gain access, which allows enforcement of the aforementioned restrictions.

This requirement of state-issued ID is largely relevant to age gating. Throughout the late 2010s and 2020s, heavier restrictions began to be imposed, primarily to avoid gaming addiction. Strict playtime limits were enacted for minors, with them only being able to play them at certain times on certain days for limited periods. A black market of rental adult accounts sprung up as a result. In 2021, to ensure compliance, Tencent launched ‘Midnight Patrol’, a system which uses facial detection and algorithmic processing of player behaviour to detect underage players using other people’s accounts, such as their parents, to circumvent playtime restrictions.

Chinese Variants

Many popular games have significantly altered China-specific variants, including:

• Minecraft’s Chinese variant is My World: China Edition (我wǒ的de世shì界jiè). It is free-to-play with many micro-transactions and subscriptions. In addition to all that one would typically expect from Minecraft, it also has full additional gamemodes and takes cues from role-playing games. The mobile version even included a virtual house builder for a time, as you might see in a mobile match-three puzzle game.

• Plants vs Zombies 2 was first developed by PopCap’s Shanghai studio and then later by Talkweb Games. It is notable for having a huge amount of exclusive content, including extra worlds, plants, and game systems.

• Game for Peace/Peacekeeper Elite is Tencent’s version of PUBG. There are reductions in violence such as players not dying when eliminated, instead just waving goodbye.

• LuoBuLeSi was China’s version of Roblox that only operated for a few months in 2021, launching in July and closing in December, likely due to the open-ended nature, with user-generated content being deemed too difficult to moderate. The playtime limits for minors also likely had severe impact on Roblox’s ability to reach its primary demographic. At the time of publication, it has been confirmed to be returning in 2026.

This is just a small selection, for these are just some of the cases of games having Chinese variants that I find more interesting. Almost every game that is made available within China sees some degree of change to comply with regulations, even if that is just tweaking a few icons.

There is also extraterritorial censorship which impacts global titles. Marvel Rivals proved itself popular but is developed by NetEase. As such, certain phrases have been censored from the in-game chat. Genshin Impact, which is developed by miHoYo, also faced significant backlash from Western players when it was discovered various terms were censored in chat. Ubisoft made an attempt to censor the global version of Rainbow Six Siege via graphical changes, including removing slot machines from maps and redesigning icons to remove skulls. Following player backlash, they reverted the changes. When Black Myth: Wukong launched, influential streamers covering the game were given a list of topics to avoid. There are many more cases.

Many cases exist of games being banned or restricted due to their display and portrayal of subjects sensitive to the government. The Ministry of Culture banned Hearts of Iron for ‘distorting history and damaging China’s sovereignty and territorial integrity’. The game’s depiction of Manchuria, West Xinjiang, and Tibet as sovereign nations likely also had impact on this decision. Battlefield 4 was also banned in China due primarily to an expansion pack titled ‘China Rising’. The launch of Diablo Immortal was delayed after the game’s Sina Weibo account published a post widely interpreted as a jab at Xí Jìnpíng.

Mobile Games

There is also a huge market for mobile games. Within China, many mobile games are not distributed as standard, self-contained applications installed on the operating system but are instead distributed within applications like WeChat. For example, the aforementioned Chinese version of Plants vs Zombies 2 is supported by a great number of Mini Programs.

Due to the distribution method, games also commonly have significant social integrations. It is easy to consider a game’s integration with WeChat the same as a game might integrate with Facebook, but they are different beasts entirely. Games are often deeply integrated into social identities, with tie-ins to leaderboards and in-app purchases directly through those services.

The growing capabilities of mobile devices have also had the same effect as they’ve had globally, with many more ambitious and technically demanding games reaching phones. Honor of Kings and other more complex mobile games see great successes.

Grey Market

Regardless of regulation, much global gaming content and many services that aren’t strictly forbidden, but also aren’t strictly allowed, exist within the country. They operate in an odd state, possibly by virtue of the government turning a blind eye to it so the industry can strengthen.

The global version of Valve’s game distribution powerhouse Steam operates to a degree within China. It is blocked intermittently, sometimes just in part, but is often available. The exact state of the service’s accessibility from within China varies frequently, with especially the Steam Community portion of the site being blocked. In 2021, Steam China launched, which adheres to China’s requirements but has a much smaller catalogue of games and lacks many features present in the global version.

Many people use VPNs to be able to access external gaming content, and a lot of people use game boosters (加速器jiāshùqì), which are popular for improving performance of connections to overseas servers. The Great Firewall and poor international routing make them borderline mandatory for interacting with a global player base.

Internet Cafés

In China, and indeed throughout much of Asia, internet cafés have retained a degree of relevancy. There has been a decline similar to the one seen globally during the 2010s with the increased popularity of smartphones and reductions of mobile data costs, but not nearly as drastic. Wangba, as they’re known, are still extremely popular. So popular that they’re actively growing.

They serve as cost-effective ways to play games conveniently on powerful computers which people may otherwise be unable to afford. Internet cafés also play directly into China’s massive e-sports culture. e-sports is much bigger than it is in the West. Titles including League of Legends and CrossFire (a Korean FPS massive in China) are extremely popular in part because of internet cafés. It is a symbiotic relationship of sorts.

Despite all the restrictions, China’s gaming industry is the world’s largest video gaming market by sales and is only positioned to get bigger. In 2024, China’s first AAA game, Black Myth: Wukong, made its release to a solid critical and public reception. There is even suggestion that the game’s success might trigger further investment in the space. Part of this is also China’s opportunity for cultural export, with games being seen as an opportunity to distribute the aforementioned censorship and content regulations to overseas audiences. Video games are proving a valuable export with clear economic incentive, and with government support China’s video game industry seems positioned to see huge investments and expand a great deal further on a global scale.

History

Mao Zedong was the founder of the People’s Republic of China and led the country from its establishment in 1949 until his death in 1976. He developed Maoism, a variation of Marxism–Leninism. Following the beginning of the Chinese Civil War, he aided in the establishment of the Chinese Red Army and then proceeded to head the Land Reform Movement (土tǔ改gǎi), industrialisation via five-year plans, and launch the Campaign to Suppress Counterrevolutionaries (镇zhèn压yā反fǎn革gé命mìng运yùn动dòng). He led intervention in the Korean War and oversaw the Great Leap Forward (大Dà跃yuè进jìn) campaign from 1958 to 1962.

In 1966 he launched the Cultural Revolution, aiming to purge capitalist and traditional elements from Chinese society. During this period, there was mass destruction of artefacts, a violent class struggle, and, notably relevant within the context of this article, his cult of personality.

Mao Zedong’s cult of personality was a propaganda campaign designed to elevate Mao’s status as a beacon of communist China. His image was widely distributed in portraits and badges, and every Chinese citizen was given ‘Quotations from Chairman Mao Tse-tung’ (毛Máo主Zhǔ席xí语yǔ录lù), otherwise known as the Little Red Book, which they were expected to carry at all times and read from frequently.

Following Mao’s death, the Boluan Fanzheng program was launched, which translates in English to ‘Eliminating chaos and returning to normal’. As part of this program and subsequent efforts by Deng Xiaoping, Hu Yaobang, and others, China began to repeal some of the Cultural Revolution’s changes.

However, some felt that this occurred too abruptly and moved too swiftly, giving people too much freedom too suddenly and causing them to act in pursuit of further liberties. This caused disputes which directly led to the 1989 Tiananmen Square protests. China, which had before the protests seen greater (but far from full) press freedom, fell suddenly back under intense censorship. After a few years, restrictions were relaxed somewhat – particularly after Deng Xiaoping’s 1992 southern tour. However, media remained very much restricted.

China’s first interactions with the internet were as early as 1987, but it wasn’t until 1989 that it started to properly gain a footing, starting with inter-university contact, as most countries’ internet communications did. In 1994, China began to embrace the internet, becoming properly linked to the internet in April. The next year, citing economic opportunity, the basis of infrastructure and a national network were established.

China, being an extremely large country, took a while to gain proper country-wide infrastructure, with it taking until the early 2010s for proper connections to reach rural areas. For a brief while in the 2000s, there was a surprisingly wild and free web, with minimal oversight. Unfortunately, as the internet became established, and the web began to form, censorship shaped it. China began to block many international sites, preventing penetration into the Chinese market. Under Xí Jìnpíng, censorship has been far stricter.

Unlike much of the rest of the world, where the global interconnectivity of the internet is emphasised, China emphasises ‘Internet Sovereignty’ (网wǎng络luò主zhǔ权quán) – the idea that the government should have complete control of the internet as it operates and is available within the country.

The Great Firewall

This sovereignty is achieved via what is widely referred to as the Great Firewall (防Fáng火huǒ长cháng城chéng), a reference to the Great Wall of China. It is the combined product of the various legislation and unofficial degrees paired with technological enforcement.

It is a complex and oppressive measure. It does not merely block offending URLs but goes to great lengths by resetting connections, performing DNS poisoning to direct to incorrect IP addresses, scanning the actual data in transit via deep packet inspection, enacting full IP blocks, performing man-in-the-middle certificate takeover attacks, and more. It is one of the most complex filtering systems deployed at scale. Some sites are let through the filter but in extremely degraded states in an effort to avoid the backlash of full blocks while still functionally preventing access.

All of China’s major telecommunications companies – China Telecom, China Unicom, and China Mobile – are state-owned, meaning that their services are inherently under government jurisdiction, making circumvention extremely difficult. They’re choke points of internet traffic.

Many of these restrictions can be bypassed by means of VPNs which are legal, albeit with restrictions. VPN providers must obtain state approval and face imprisonment if they fail to do so. As such, much information can be accessed, but it is made difficult and consuming of money and time. Unfortunately Tor functions in a degraded state, requiring workarounds, and many more standard proxies are thwarted by the firewall. The friction is only growing as time progresses.

The Great Firewall is a large part of this, and its prevention of access to international services has determined China’s development during the Fourth Industrial Revolution by giving preference to domestic organisations. Not only is this positive for China’s local economy, but it is also beneficial for the government as it has greater control over domestic companies than international ones. The Great Firewall acts as both a protective trade barrier and an instrument of ideological control.

Restricted Topics

There are many factors of China that are heavily suppressed. Information about the Tiananmen Square protests and massacre is one of the more famous examples of censorship, but there is far more. In their 2025 world press index, Reporters Without Borders placed the People’s Republic at 178 out of 180, higher only than North Korea and Eritrea. Some major topics subject to censorship include:

• Taiwan

  Specifically, discussion of the country Taiwan as the independent country which it is, rather than as a part of China.

• Human Rights Abuse

  There are ongoing mass human rights abuses against Uyghurs and other Turkic Muslim minorities who are held in concentration camps. Starting in 2014 under the guise of the Strike Hard Campaign Against Violent Terrorism and escalating in 2017 with the mass unlawful incarceration of some one million Uyghurs, it is widely regarded as the largest mass internment of ethnic and religious minority groups since the Second World War.

  There are also great and horrific human rights abuses in Tibet which are censored so heavily that the availability of reliable information is even impacted globally. Discussions of the human rights abuses or Tibetan independence are very heavily censored.

• Government Dissent

  Much criticism of the government is silenced, particularly anything which is related to or could lead to activism. It is clearly outlined in Communiqué on the Current State of the Ideological Sphere (关Guān于yú当dāng前qián意yì识shí形xíng态tài领lǐng域yù情qíng况kuàng的de通tōng报bào), a document leaked by journalist Gao Yu. The ‘Seven Noteworthy Problems’ outlined in the document are:

  1. Promoting Western Constitutional Democracy: An attempt to undermine the current leadership and the socialism with Chinese characteristics system of governance.
  2. Promoting ‘universal values’ in an attempt to weaken the theoretical foundations of the Party’s leadership.
  3. Promoting civil society in an attempt to dismantle the ruling party’s social foundation.
  4. Promoting Neoliberalism, attempting to change China’s Basic Economic System.
  5. Promoting the West’s idea of journalism, challenging China’s principle that the media and publishing system should be subject to Party discipline.
  6. Promoting historical nihilism, trying to undermine the history of the CCP and of New China.
  7. Questioning Reform and Opening and the socialist nature of socialism with Chinese characteristics.

• COVID-19

  Less so now given that the peak of the pandemic is past, but China heavily restricted discussion of the coronavirus. It also restricted discussion about other outbreaks, such as SARS during the 2002–2004 SARS outbreak, which was also first identified in China.

This censorship has gradually gained a tighter grip around Hong Kong, which previously enjoyed a largely unrestricted internet, with the passing of the 2020 Hong Kong National Security Law (香Xiāng港gǎng国guó家jiā安ān全quán法fǎ) presenting a significant alteration to the digital landscape.

Different Services

Given that the internet within China is aggressively moderated and filtered, internet services external to the country which cannot be controlled and don’t align with regulatory requirements are often disrupted or blocked wholesale. Some major services that are blocked include Google, YouTube, Facebook, Wikipedia, Reddit, Instagram, WhatsApp, Twitch, Twitter, Pinterest, Signal, Discord, the Internet Archive, TikTok, and ChatGPT. This is by absolutely no means a comprehensive list, and there are so many more sites and services blocked. News sites especially are aggressively blocked due to China’s minimal press freedom.

With a population in excess of a billion people (roughly 20% of all web users), China has a size such that it can support a more or less self-contained web, with its own systems and alternatives to global platforms. The Great Firewall’s disruption of global competitors has created a massive domestic tech ecosystem within the vacuum.

Many services have received direct Chinese versions, such as LinkedIn, which had a Chinese version titled 领lǐng英yīng from 2014 until 2021 and then a stripped-down jobs-only version called InCareer until 2023. In other cases, there are equivalents or services with similar functionality that take their place. Commonly, the functions of many apps are combined into a singular.

The ‘web’ as it exists elsewhere, with lots of interconnected separate sites, is not the same web that exists in China. Like a more aggressive version of the consolidation of the global web into a few social media sites that many have been critical of. There are many sites, but they’re all scrutinised.

Another defining feature of this domestic ecosystem is the absolute dominance of live commerce. Unlike the West, where e-commerce and social media are generally distinct, platforms like Douyin, Kuaishou, and Taobao Live merge short-form entertainment with instant purchasing. This has birthed a substantial industry of livestream anchors and a highly accelerated consumer web experience.

One of the most famous Chinese apps is Weixin (微Wēi信xìn), which operates globally as WeChat. Developed by juggernaut Tencent, Weixin/WeChat is arguably an operating system in and of itself. It is used by almost everyone in China, as well as in other countries like Malaysia, to the point that it is difficult to get by without it. It encompasses messaging, voice/video calls, payments, location sharing, social media functionality, mini-programs, delivery, transport, games, appointments, and just about everything else one could think of. It is well and truly an ‘everything app’.

Weixin is integrated with China’s mass surveillance systems and has very restrictive terms of service which align with government policies. All data is kept within the country. However, outside China, WeChat data is largely stored elsewhere, and different platform policies apply (unless interacting with someone within China). The convenience and difficulty of avoiding Weixin and similar ‘everything apps’ creates a walled garden that keeps users deeply entrenched in platforms the state can easily monitor.

Systematic Oversight

This monitoring is not a side effect of the technology but a core feature of its design. There are a number of restrictions and influences from China’s uniquely structured internet.

In addition to mass censorship, there is also mass surveillance. In late 2012, the real name system came about, mandating online accounts be linked to people’s identities. This was furthered in 2025, when the National Online Identity Authentication (国Guó家jiā网wǎng络luò身shēn份fèn认rèn证zhèng) system launched.

Often in the West you’ll hear jokes about China’s social credit system based on mistranslated details presented in poorly researched, sensationalist articles. There is no single, national numeric ‘social credit’ score assigned to people like a financial credit rating. Companies and individuals who ensure compliance are treated more favourably, and those that fail to or frequently violate regulations may be restricted from certain benefits. On a smaller level, companies and local governments sometimes award perks to people who take positive action, but this has minimal impact on daily life. Credit systems do exist, but they’re fragmented and not a nationwide value.

However, it is worth noting that judicial blocklists do carry severe real-world consequences, such as bans on purchasing high-speed rail or airline tickets, which often fuels the Western confusion. It is a shame that China’s social credit system as a whole is so pervasive a myth, given how ubiquitous surveillance and censorship are within the country by other means.

In 2025 a new regulation was passed which has made it so that influencers without verified credentials are banned from speaking on certain topics, including finance, law, medicine, and education, in an effort to reduce misinformation. Further legislation, such as the Cybersecurity Law, mandates data localisation and network operator compliance, legally binding technology companies to state surveillance apparatuses.

There are also a whole host of restrictions and regulations which have completely reshaped China’s gaming industry and culture.

Online Culture

Like much of the world, China’s online culture is a combination of both local and global trends and culture. However, the language barrier and isolation of China from the rest of the globe’s internet do lead to a more separate digital zeitgeist. Much that does make it into China is interpreted differently or associated with different meanings, much like information that is known globally about goings-on within China is distorted. This even leads to humour, seen in instances like the aforementioned Western misconceptions about the social credit system which have become the basis of many jokes and online discussions.

Within China, internet users are commonly referred to as netizens (网wǎng民mín). There are many global memes which have translated over to China, and there are also China-specific versions, interpretations, and equivalents of popular memes. Memes reflect culture and what people are thinking about. There are some interesting cases, though, such as the Doge meme maintaining relevance thanks to commonly being available as an emoji and used in messages to indicate sarcasm or irony.

For example, Baozou comics (暴bào走zǒu漫màn画huà) take the place of Western rage comics. Where in the West someone might reply they’re ‘eating popcorn’ when drama occurs, the Chinese equivalent is ‘eating melon’ (吃Chī瓜guā). There are many equivalents to many English terms. GOAT, meaning Greatest of All Time, is roughly equivalent to YYDS, which stands for Yǒngyuǎn de shén and means ‘Forever God’. You also have 润Rùn学xué. Rùn (润) means ‘moist’ or ‘profitable’ and is spelt in pinyin like the English word ‘run’. As such, it is used to refer to leaving China and is featured in discussions of how to get overseas visas.

As is typical for memes, they are not just empty humour but also reflect society at large and the general thoughts of the populace. In this vein, comparing Xí Jìnpíng to Winnie-the-Pooh has become quite the meme, though one that is heavily censored against. Automated filter systems easily detect direct references to disallowed topics, so anything sensitive is referred to with evasive manoeuvres such as homophones, historical allegories, and deliberate typos.

Artificial Intelligence

Chinese large language models will refuse to discuss censored events or criticise the government, instead following Chinese Communist Party narratives. Trying to reference censored events such as Tiananmen Square is restricted, as is drawing comparisons between Winnie-the-Pooh and Xí Jìnpíng:

It is worth noting that this exchange was conducted in English. The language used to converse with AI models can influence their responses and determine the contents of their messages. It isn’t just DeepSeek which is censored, for heavy restrictions are present in other major models. Alibaba’s Qwen, Moonshot’s Kimi, Zhipu AI’s GLM, Xiaomi’s MiMo, and many more are censored.

In the face of more and more people relying on AI for information, this is critical. More critical, however, is the use of language models for more advanced, nuanced censorship. Tricky turns of phrase, or vague allusions to a topic, are increasingly unviable. Moderation passes with language models can detect negative government sentiment or anything pertaining to a restricted topic to an almost unavoidable degree. This emphasises the importance of local AI models to reduce the ability of governments and companies to censor and control, keeping information and technology free and democratised.

Unfortunately, there is the further issue of Chinese-managed models being used to sway sentiment online, presenting as normal users and astroturfing discussions to influence people and sow discord.

China’s digital landscape is a version of the web that is not World Wide but is instead fractured along the lines of ideology and nationalism. It has evolved not with the web as it is generally known globally, but alongside it, building its own identity, rules, and customs. It is restricted and only becoming more restricted in time, keeping information from entering and a population from exiting. It is a digital silo, where the tools designed to facilitate the sharing of knowledge and information have instead been repurposed to define the boundaries of a sovereign, state-sanctioned truth.