The story of the internet is one of those technology arcs where the people who built it had no idea what they were building. ARPA was funding a packet-switching experiment to make military communications survive a nuclear strike. Graduate students at UCLA and Stanford were running cables between PDP-10s and writing protocol software in assembly. CERN was solving an internal documentation problem. NSFNET was funded as a research network for American universities. None of it was supposed to become the substrate of global commerce, communication, and conflict, and yet here we are. This is how it actually happened, more or less in order, with the people involved named where they’re known.
History#
Phase 1: Early networking and ARPANET#
In the early 1960s, the US Department of Defense’s Advanced Research Projects Agency (ARPA) sought to create a communication network resilient enough to survive a nuclear attack. Computer systems at the time were centralized, with a single point of failure that could disrupt communication. That problem drove the development of decentralized networking concepts.
The intellectual foundations came from several people working independently. J.C.R. Licklider at MIT wrote “Man-Computer Symbiosis” in 1960, imagining interactive computing as a partnership rather than a batch-job submission. Leonard Kleinrock at MIT (later UCLA) wrote his 1962 dissertation on queueing theory in packet networks. Paul Baran at RAND published his “On Distributed Communications” papers in 1964, designing a network architecture that could route around damaged nodes. Baran’s papers are often credited as the conceptual origin of packet switching, though his specific design was never built. Donald Davies at the UK National Physical Laboratory independently arrived at similar ideas in 1965 and coined the term “packet.”
Larry Roberts at ARPA pulled these threads into the ARPANET design. BBN (Bolt, Beranek, and Newman) in Cambridge, Massachusetts won the contract in 1968 to build the Interface Message Processors, the dedicated minicomputers (Honeywell DDP-516s) that handled the packet routing.
The first ARPANET node went live at UCLA in September 1969. On October 29, 1969, Charley Kline at UCLA tried to log into a remote system at Stanford Research Institute. He got as far as typing “L-O” before the SRI host crashed. The first message on what would become the internet was, accidentally, the word “lo.” The system was restarted about an hour later and the full “LOGIN” went through.
By the end of 1969, ARPANET had four nodes: UCLA, SRI, UC Santa Barbara, and the University of Utah. Host counts grew through the 1970s as more universities and research institutions joined. The applications people built on top of it defined what networked computing would be. In 1971, Ray Tomlinson at BBN wrote the first email program and chose the @ symbol to separate user from host because it was the only punctuation on his keyboard that wasn’t already used in usernames. Telnet (1971) allowed remote login. FTP (1971) allowed file transfer.
The networking protocol question took longer to settle. ARPANET originally ran NCP (Network Control Protocol), which assumed a reliable underlying network. Vint Cerf (Stanford) and Bob Kahn (DARPA) published the foundational TCP paper in 1974 (RFC 675; the protocol was just called TCP at this point and split into separate TCP and IP layers later in the 1970s). TCP/IP was designed to be more robust than NCP, especially for connecting networks of different types.
The transition was a hard cutover. On January 1, 1983, the ARPANET “flag day,” every host on the network was required to switch from NCP to TCP/IP. The cutover was disruptive enough that operators wore T-shirts that read “I survived the TCP/IP transition.” That was also the moment ARPANET stopped being a single research network and started being a network of networks, which is the literal definition of “internet.”
Before its decommissioning in 1990, ARPANET saw the creation of other foundational technologies, including FTP and Usenet, a distributed discussion system that was a forerunner to modern forums and social media. Most of the internet protocols an operator works with in 2026 trace directly back to ARPANET-era design decisions.
Phase 2: Bulletin Board Systems and the emergence of online communities#
In the mid-1980s, a new phase of online activity emerged as Bulletin Board Systems (BBS) became more popular. A BBS was usually one computer with a modem and a phone line, run by a hobbyist who let others dial in to read messages, download files, and play text adventures. Long-distance phone charges meant most BBSs served their immediate area code, and the culture was deeply local: each BBS had its own SysOp, its own community, its own rules. The closest equivalent in current internet culture is probably a Discord server, except you paid by the minute and had to wait for someone else to hang up before you could connect.
BBSs were often focused on specific interests: computer programming, science fiction, politics, music, fandom. They were customizable and personal in a way the corporate-platform internet of 2026 mostly isn’t. Many BBSs were havens for people who felt marginalized or isolated offline, and they provided a space to connect with others who shared their interests. Beyond discussion forums, BBSs offered file sharing, online games (door games like trivia and turn-based strategy), and chat rooms.
The most famous BBS was probably The WELL (Whole Earth ‘Lectronic Link), founded in 1985 in San Francisco out of the Whole Earth Catalog community. The WELL was text-only and intensely social, a gathering point for people interested in technology, social issues, and counterculture. Howard Rheingold’s 1993 book The Virtual Community documented what the WELL felt like in that era, and many of the early online-culture norms that survived into the modern web came out of conversations there.
As the internet became publicly accessible in the 1990s, BBSs gradually faded, but their legacy is visible in every modern online community. The experimentation, the counterculture, the user-defined norms, all of it carried forward into Usenet, then the web, then the platforms.
Usenet: The precursor to modern social media#
Another significant development of the late 1970s and 1980s was Usenet, a distributed network of discussion forums that allowed users to exchange messages on a wide variety of topics. Usenet was first created in 1979 by two Duke University graduate students, Tom Truscott and Jim Ellis, originally as a way to share information between Duke and the University of North Carolina.
Usenet was based on a distributed store-and-forward architecture: messages were copied between servers around the world rather than living on a single central host. That made Usenet one of the first truly global online communities and meant it could survive the failure of any single node.
Usenet quickly became popular among computer enthusiasts and academics, who used the network to share information and collaborate on research. It wasn’t just a tool for serious discussion; it was also a place for socializing, humor, and an emerging fandom culture. Discussion groups (newsgroups) were organized into a hierarchy by topic, covering everything from science and technology to sports, entertainment, and the long tail of science fiction and fantasy fan groups that helped shape what online community feels like.
Usenet users formed close-knit communities within these newsgroups, sharing jokes, memes (the term itself predates the modern usage but the practice is older), and personal stories. The culture rewarded humor and satire, and many users enjoyed crafting elaborate pranks and hoaxes to share with the community.
The Great Renaming of 1986-1987 was the first attempt to bring structure to Usenet’s sprawl. Rick Adams, Gene Spafford, and a small group of other Usenet administrators reorganized the entire newsgroup hierarchy into the “Big Seven”: comp.*, news.*, sci.*, rec.*, soc.*, talk.*, and misc.*. (humanities.* and other categories were added later, with alt.* deliberately left outside the hierarchy as the place where unmoderated groups could live.) Despite some later retellings that describe it as a prank, the Great Renaming was a deliberate administrative effort.
Another important aspect of Usenet culture was its commitment to free speech and open discussion. Usenet was one of the first online communities to embrace netiquette, a set of guidelines for appropriate online behavior, and many users saw it as a platform for free expression and open debate.
As Usenet grew, it also began to attract spammers, trolls, and other disruptive users. Some discussion groups became dominated by flame wars and personal attacks, which led to calls for moderation and control. The cultural shift accelerated with Eternal September: AOL added Usenet access for its subscribers in September 1993, breaking the traditional pattern in which each September a wave of new college freshmen would arrive on Usenet, learn the norms, and gradually be absorbed. Dave Fischer’s 1994 post coined the term, lamenting that the September of 1993 never ended. From that point Usenet’s culture as it had been was effectively over.
Despite these challenges, Usenet remained active throughout the 1980s and 1990s, and it played a major role in shaping the culture of the early internet. The discussion-forum patterns Usenet established paved the way for everything from web forums to Reddit to the platforms that followed.
Online gaming: from Spacewar to modern esports#
Online gaming has its own thread in the internet story, and it gets less attention than it deserves given how much modern online culture comes from it. The earliest networked game is generally credited to Spacewar!, written in 1962 by Steve Russell and others at MIT on a PDP-1. Spacewar was local-only (multiple players on the same machine via shared controllers), not networked across hosts, but the multiplayer-game-on-a-computer concept it established became the template everything later built on.
In the BBS era, the most popular games were text-based role-playing games where players created characters and explored virtual worlds through text-based descriptions. Multi-User Dungeons (MUDs) were the canonical version: text-based multiplayer worlds where users interacted in real time, primarily played by college students and academics with university network access. MUDs were one of the first instances of virtual worlds with user-generated content, since players created their own areas, objects, and characters. The first MUD (Roy Trubshaw and Richard Bartle’s MUD1 at the University of Essex in 1978) is the direct ancestor of every modern MMORPG, from EverQuest and World of Warcraft through Final Fantasy XIV.
Another popular type of BBS game was the door game, which was a separate program the BBS would launch when a user selected it from the menu (the user “went through the door” into the game). Door games were typically text-based: trivia, word puzzles, strategy games like Trade Wars 2002 and Legend of the Red Dragon. Highly competitive, with bragging rights and high scores driving most of the engagement.
As internet connections became faster in the 1990s, online gaming evolved rapidly. Doom (id Software, December 1993) wasn’t initially designed for internet play (it shipped with serial-cable and IPX LAN support), but the Doom community quickly built tools to tunnel Doom traffic over the early internet, and the deathmatch culture exploded. Quake (id Software, June 1996) was designed from the ground up for online play with native TCP/IP networking, and Quake introduced dedicated servers so players could connect to central hosts and compete in organized tournaments. Quake is often credited as the start of competitive esports as a recognizable activity.
The culture of early online gaming was tight-knit and highly specialized. Players formed close bonds through shared interest, and the communities were small enough that regulars knew each other by handle. Competitive rivalry coexisted with the social side: many lasting friendships and relationships started in 1990s game lobbies and IRC channels.
Looking back, the early years of online gaming were experimental, technically constrained, and overwhelmingly text-based. The gameplay was primitive by current standards, but the social and competitive patterns those early games established are essentially the same patterns running in 2026’s MMOs, MOBAs, and esports leagues. The Twitch streamer competing in a multimillion-dollar tournament is doing a recognizable descendant of what Quake players were doing in 1996.
Phase 3: The World Wide Web and online services#
The 1990s saw the birth of the World Wide Web and the rise of commercial online services like CompuServe and GEnie. Tim Berners-Lee, a British computer scientist at CERN (the European Organization for Nuclear Research), invented the Web starting in 1989.
Tim Berners-Lee and the creation of the World Wide Web#
The Web was solving a problem CERN had: the lab needed a way for physicists to share documentation across the many different operating systems they used, and the existing options (Gopher, FTP) were either too rigid or too sparse. Berners-Lee wrote a proposal in March 1989, his manager Mike Sendall wrote “Vague, but exciting…” on the cover, and Berners-Lee got the time to build it.
At the heart of the Web is a set of protocols Berners-Lee designed: HTTP (Hypertext Transfer Protocol) for moving documents over the network, HTML (HyperText Markup Language, originally a small subset of SGML) for the document format, and URLs for a uniform syntax to name resources. Together they let users create web pages linked through hyperlinks.
Berners-Lee also wrote the first browser, WorldWideWeb, in Objective-C on a NeXT workstation. It was both a browser and an editor, released to the public in 1991. The browser displayed text and inline images, allowed users to follow blue underlined hyperlinks, and rendered HTML in a standardized way that worked across the operating systems CERN’s researchers used. The first web server, CERN HTTPd, also ran on a NeXT machine at CERN; the hostname info.cern.ch is still online and serves the original 1991 content.
The first web server outside Europe was SLAC (Stanford Linear Accelerator Center), set up by Paul Kunz on December 12, 1991. SLAC’s existing physics-papers database became the first non-trivial reason for non-CERN physicists to use a web browser. NCSA Mosaic (Marc Andreessen and Eric Bina at the National Center for Supercomputing Applications, University of Illinois) shipped its 1.0 release in April 1993 and made the Web visible to a much wider audience: it was easy to install, ran on Windows and Mac as well as Unix, and rendered inline images, which Berners-Lee’s WorldWideWeb didn’t. Andreessen left Illinois in 1993 to co-found what became Netscape Communications, and Netscape Navigator’s 1994 release and 1995 IPO turned the Web from research curiosity into commercial inevitability. (See the Netscape post for the longer arc of that company, from 80% market share to bankruptcy in about twelve years.)
Berners-Lee’s vision was a platform that would let people share information and collaborate across geographic and organizational boundaries. The Web has gone a lot of places he didn’t necessarily intend (the centralized platform economy, surveillance advertising), but the foundational protocols are still doing the job he designed them for in 2026.
CompuServe, Prodigy, GEnie, and AOL: the commercial online services#
Before everyone was on the open web, the consumer model was dial-up to a proprietary online service. These services provided email, chat rooms, forums, news, and other online tools accessible to anyone with a modem and a credit card. Users dialed in to a central network and paid by the hour for their connection time.
CompuServe was the oldest and the most enterprise-flavored. Founded in 1969 as a computer time-sharing service for businesses, it gradually evolved into a consumer-oriented online service through the 1980s. CompuServe offered email, chat rooms, forums, and online games. Its forums (CompuServe called them “SIGs,” Special Interest Groups) covered everything from technical topics to hobbies and were popular among business professionals and serious computer enthusiasts.
GEnie (General Electric Network for Information Exchange) launched in 1985 as a consumer-oriented online service from the start, run by GE. It offered email, chat rooms, forums, a real-time news service, and a shopping service that let users purchase products from online retailers. GEnie was popular with hobbyists and gamers, with a strong gaming community that helped pioneer the online RPG genre.
Prodigy launched in 1984 as a joint venture between IBM, Sears, and CBS (CBS dropped out before launch). Prodigy offered email, forums, chat rooms, and a range of news and information services. The Prodigy interface was unusually graphical for the era, with banner-style ads at the bottom of the screen that subsidized lower subscriber fees.
AOL (America Online) had the longest run with the broadest reach. Founded in 1985 as Quantum Computer Services (initially serving Commodore 64 users), renamed America Online in 1991, and aggressively marketed through the famous CD-ROM mailings that arrived at almost every US household for over a decade. AOL was popular with families and casual users, known for its easy-to-use interface, “You’ve got mail” notification, instant messaging (AIM), and chatroom culture. AOL also made one of the most important decisions in early internet culture: in September 1993, AOL added Usenet access for its subscribers, kicking off the Eternal September that permanently changed the open internet’s culture.
All of these services used proprietary software and networks to connect users. They were limited in scope compared to the open internet that eventually replaced them, but they were an essential precursor to the modern online world. CompuServe and GEnie folded into the broader internet through the late 1990s. Prodigy was absorbed by SBC and shut down by 2001. AOL bought Time Warner in January 2000 in what was for years considered the worst merger in corporate history, then watched the broadband transition strand its dial-up subscriber base. AOL the company still exists in 2026 as a Yahoo subsidiary, but the era of the walled-garden online service ended with the broadband transition.
The first online sale#
Among the milestones of the early commercial web, the first online sale stands out. On August 11, 1994, Phil Brandenberger of Philadelphia bought a Sting CD (Ten Summoner’s Tales) from NetMarket for $12.48 plus shipping. The transaction used PGP-encrypted credit card data, which made it the first commercial sale on the web that used proper end-to-end encryption. The New York Times covered the event the next day. NetMarket was a small startup that doesn’t exist anymore; the Sting CD does.
Phase 4: Government interconnection and the globalization of the internet#
As the internet grew in popularity and importance, governments around the world began to invest in its infrastructure. In the United States, the National Science Foundation (NSF) created NSFNET, a high-speed network that connected research facilities and academic institutions across the country.
The National Science Foundation and NSFNET#
The NSF funded NSFNET starting in 1985, originally as a 56 Kbps backbone connecting NSF-funded supercomputing centers. The point was to give academic researchers access to the few supercomputers that existed at the time, but NSFNET ended up serving as the de facto backbone for the broader academic internet. As more institutions connected, the original 56 Kbps capacity strained quickly.
To address this, the NSF launched the High-Performance Computing and Communications (HPCC) program in 1987 (formalized by the High-Performance Computing Act of 1991). The goal was to develop technology for faster, more reliable internet communication. NSFNET was upgraded to T1 (1.544 Mbps) in 1988 and to T3 (45 Mbps) in 1991.
One outcome of the HPCC program was the National Research and Education Network (NREN), a high-speed network for research and education institutions. NREN was designed to be faster and more scalable than NSFNET, using newer technologies like Asynchronous Transfer Mode (ATM) and Fiber Distributed Data Interface (FDDI) for more efficient data transfer.
By the mid-1990s, the combination of NSFNET and NREN had become the backbone of the US academic internet, enabling new applications and services to be developed and deployed at scale.
The NSF also played an indirect role in the development of the Web. In 1989, Tim Berners-Lee proposed his hypertext system at CERN, and as the proposal turned into working software in 1991, NSF-funded networks were where most of the early adoption happened. The NSF provided funding to establish early US web servers (the first non-European web server, set up by Paul Kunz at SLAC on December 12, 1991, ran on NSF-connected infrastructure) and funded the development of NCSA Mosaic at the University of Illinois, the browser that made the web visible to the general public.
NSFNET was decommissioned in April 1995, when commercial backbones (Sprint, MCI, AT&T, ANS) had grown enough to take over the traffic. The NSF’s role had been to bootstrap the infrastructure; once the commercial sector could carry it, the federal subsidy ended. The transition is roughly the moment the internet stopped being an academic experiment with a government-funded backbone and became commercial infrastructure.
The emergence of search engines: from Archie to Google#
During the mid-1990s, the number of websites on the internet grew rapidly, making it harder for users to find the information they were looking for. Search engines emerged in response, offering a way for users to look up specific content on the web.
One of the first true search engines was Archie, developed in 1990 by Alan Emtage, a student at McGill University in Montreal. Archie indexed files stored on FTP servers (it predated the web), letting users search for specific files by keyword. Limited in scope by modern standards, but it demonstrated the potential of indexing and searching large amounts of information on the internet.
Gopher (1991, Mark McCahill and team at the University of Minnesota) was a hierarchical document-distribution protocol that competed with the early web. Gopher had its own indexing and search add-on called Veronica, plus a related tool called Jughead. Gopher quickly became popular among early internet users, with over a million users by 1995. The protocol declined sharply after the University of Minnesota started charging licensing fees in 1993, which sent most users to the unrestricted web. Strictly speaking, Gopher was a document protocol rather than a search engine, but Veronica running against the Gopherspace was many people’s first taste of an indexed network-wide search.
The first true web search engines arrived in 1993-1994: W3Catalog (September 1993), Aliweb (October 1993, manually submitted), JumpStation (December 1993), and WebCrawler (Brian Pinkerton at the University of Washington, released April 20, 1994). Yahoo (originally “Jerry and David’s Guide to the World Wide Web”) started as a directory in early 1994 and was renamed Yahoo! later that year. AltaVista launched in 1995 with full-text search of the whole web, and Lycos and Excite followed.
In 1996, two graduate students at Stanford University, Larry Page and Sergey Brin, started a research project called BackRub (they met in 1995). BackRub used a new algorithm, PageRank, that ranked web pages based on the number and quality of links pointing to them. The project was renamed Google in 1997 when they registered the google.com domain, and incorporated as a company in September 1998.
Google’s success came in large part from PageRank. The algorithm produced meaningfully better results than what competitors were doing, because backlink analysis turned out to be a stronger signal of relevance than the keyword frequency and meta-tag analysis other engines were using. Google remained the dominant search engine through the 2000s and beyond, and the PageRank insight is still recognizable in modern search-ranking systems even after two decades of evolution.
Peer-to-peer file sharing and the Napster era#
In the late 1990s, a new technology arrived that changed the way people shared digital files: peer-to-peer (P2P) file sharing. Rather than relying on centralized servers to host files, P2P networks let users share files directly with each other.
Napster, created by college student Shawn Fanning in 1999, was the first P2P program to reach mass adoption. Napster used a centralized index of files shared by its users, allowing others to search for and download files directly from whoever was hosting them. The architecture was technically hybrid (centralized index, decentralized transfer), but the user experience was revolutionary: type a song name, get a download. Napster grew to roughly 80 million users at peak, and it made it incredibly easy for people to share music files.
Napster’s success came with immediate legal blowback. The RIAA sued Napster for copyright infringement in December 1999, and Metallica’s Lars Ulrich filed a separate suit in April 2000 that became the symbolic centerpiece of the controversy. The Ninth Circuit ruled against Napster in A&M Records, Inc. v. Napster, Inc. (2001), and Napster was forced to shut down its file-sharing service in July 2001. It limped on as a paid music service through various ownership changes (and still exists in 2026 as a streaming brand owned by Algorand, of all things).
Despite Napster’s shutdown, P2P file sharing continued to evolve. Gnutella (released March 2000) used a fully decentralized architecture with no central index, making it harder to shut down. Kazaa (2001, FastTrack protocol) introduced supernodes that distributed the indexing load across well-connected peers. Both Gnutella clients and Kazaa faced their own legal challenges (and Kazaa became notorious for shipping with bundled spyware), and many of them were forced to shut down or pivot to legitimate business models.
The most significant evolution came with BitTorrent, released by Bram Cohen in July 2001. BitTorrent used a fully decentralized approach where files were broken into pieces and downloaded from multiple sources simultaneously, with each downloader also uploading completed pieces to others. The protocol made file sharing faster and more efficient (large files actually got faster to download as more people joined the swarm), and it made it much harder for authorities to shut down individual networks because there was no central server to seize. BitTorrent remains widely used in 2026 for both legitimate large-file distribution (Linux ISOs, game patches, scientific datasets) and continued copyright-infringement use cases.
The legal and cultural war over P2P shaped a generation. The music industry argued (with some justification) that file sharing decimated CD sales; the file-sharing community argued (with some justification) that the labels had been overcharging for albums full of filler and that legitimate digital distribution barely existed until P2P forced the issue. The eventual resolution came through Apple’s iTunes Store (2003) and later streaming services (Spotify, Apple Music, YouTube Music), which proved that people would pay for digital music if the legitimate service was actually better than the pirate version. P2P file sharing remains active in 2026, mostly for content that streaming services don’t carry, but the original Napster-era mass-piracy moment ended once Spotify made paying for music more convenient than stealing it.
The internet goes global: interconnecting the world’s networks#
By the early 2000s, the internet had become a global phenomenon connecting people across most of the inhabited world. One major factor driving that growth was the increasing availability of high-speed connections (DSL and cable broadband replacing dial-up), which let users access more content and services online than slow dial-up could realistically support.
Another driver was the continued expansion of physical infrastructure. In the late 1990s and early 2000s, companies built their own undersea fiber optic cables to connect different regions of the world. The submarine cable map of 2026 has hundreds of cables running across every ocean basin, many of them now owned outright by Google, Meta, Microsoft, and Amazon rather than by traditional telecoms. International internet exchange points (IXPs) in Frankfurt, Amsterdam, London, Ashburn, and São Paulo became the places where traffic from different networks gets routed across the world.
Global expansion brought its own challenges. Internet censorship became a serious concern as more governments sought to control access to online content; China’s Great Firewall (launched as the Golden Shield Project in 2003) is the most prominent example, but Iran, Russia, and several other countries built similar systems. The defensive technologies that emerged in response include VPNs (originally a corporate remote-access tool, repurposed for circumvention), Tor (released 2002, originally funded by the US Naval Research Laboratory), and more recently Shadowsocks and V2Ray for environments where standard VPN protocols are detected and blocked.
The other big challenge was cybersecurity. With more sensitive data shared online and more attacks driven by criminal monetization rather than research curiosity, the security posture of internet-connected systems had to mature quickly. TLS replaced plaintext HTTP for sensitive transactions starting in the late 1990s, then for all web traffic by the late 2010s (Let’s Encrypt launching in 2016 made HTTPS effectively free and removed the last cost barrier). New protocols (DNSSEC, RPKI), new threat models (nation-state actors, ransomware ecosystems), and new defensive architectures (zero trust, BeyondCorp) became part of the standard infrastructure picture.
Despite these challenges, the internet continued to grow. By the early 2000s, it had become an essential part of daily life for billions of people around the world, connecting them to information, entertainment, and each other on a scale the ARPANET designers couldn’t have imagined.
Phase 5: Mobile devices and the Internet of Things#
The 2010s saw the rise of mobile devices and the Internet of Things, with billions of devices connecting and communicating: smartphones, tablets, smart homes, wearables, and a long tail of embedded systems.
The emergence of mobile devices and wireless networks#
The arrival of smartphones in the early 2000s and 2010s shifted how people accessed and used the internet. Smartphones let users go online from anywhere, and for many people they became the primary device. Mobile internet usage grew faster than anyone projected, and demand for wireless networks (3G, then LTE, then 5G) followed.
The first smartphones were essentially handheld computers with phone capabilities: the Palm Treo (2002 onwards) and various BlackBerry devices were the dominant pre-iPhone form factor, used primarily for email, messaging, and basic web browsing. Multimedia and rich apps were limited; Java ME and BREW were the platforms for what passed for mobile applications.
The iPhone (introduced January 9, 2007, released June 29, 2007) was the inflection point. The iPhone shipped with a real WebKit-based browser, multimedia support, and a touch interface that worked. The App Store launched July 10, 2008, creating the modern app ecosystem and a new distribution model. Android (Google bought Android Inc. in 2005, the HTC Dream / T-Mobile G1 launched October 2008) followed quickly and became the dominant mobile OS by global market share, with iOS leading in revenue.
Smartphones reshaped how people consume and share information. Social media platforms like Facebook (founded 2004), Twitter (2006), and Instagram (2010, acquired by Facebook in 2012) became hugely popular on mobile, letting users stay connected, share photos and videos, and discover new content. Streaming services (Netflix’s streaming launch in 2007, Spotify’s US launch in 2011) offered instant access to libraries of movies, TV, and music on the move. Mobile-first design became standard practice around 2014, when global mobile internet usage exceeded desktop for the first time.
Smart homes, wearable technology, and the Internet of Things#
Through the 2010s, the internet expanded beyond computers and mobile devices to a sprawling category of “smart” devices grouped under the Internet of Things (IoT) banner. Smart thermostats (Nest, founded 2010, acquired by Google in 2014), security cameras (Ring, Wyze, the various Chinese-origin brands), voice assistants (Amazon Echo with Alexa in 2014, Google Home in 2016), and wearable technology like fitness trackers (Fitbit) and smartwatches (Apple Watch in 2015) all became part of the connected device population.
One advantage of IoT devices is remote control through a mobile app or web interface: adjust the thermostat, turn off the lights, lock the front door from anywhere with an internet connection. Another is interoperability between devices: a smart home system can automatically turn off lights and adjust the temperature when a user leaves, then restore everything when they return. Matter (the Connectivity Standards Alliance protocol launched 2022) finally gave the industry a common standard that competing ecosystems could speak.
The downside is security and privacy. Many IoT devices have been demonstrated to be vulnerable, often catastrophically. The Mirai botnet (September 2016) infected hundreds of thousands of default-credentialed IoT devices (mostly DVRs and IP cameras) and used them to launch DDoS attacks of unprecedented scale, including the October 2016 attack on Dyn DNS that took down large parts of the US east-coast internet. IoT device firmware is rarely updated, default credentials are rarely changed, and the typical consumer-grade device has roughly the security posture of a Linksys router from 2005.
Despite the concerns, IoT continues to grow. Industry analysts estimate over 25 billion connected IoT devices in 2026, with the count expected to keep increasing as embedded networking becomes default in everything from cars to appliances to industrial sensors. The security side of this remains unresolved.
Cloud computing and the rise of big data#
Cloud computing also became a major trend in the late 2000s and 2010s, allowing businesses and individuals to store and access data from anywhere with an internet connection. The shift enabled new forms of collaboration and made it easier than ever to work with others and share information.
Cloud computing means delivering computing services (storage, processing power, software) over the internet rather than hosting them on a physical computer or server. Users access these resources through a network of remote servers. The major providers as of 2026 are Amazon Web Services (launched 2006, the original and still the largest), Microsoft Azure (2010), and Google Cloud Platform (2008 as App Engine, expanded into full IaaS over the next decade). Together they hold the substantial majority of the public cloud market.
The advantages that drove cloud adoption are concrete: access from anywhere with an internet connection, on-demand scalability (no need to over-provision hardware for peak load), and pay-as-you-go billing that converts capex into opex. The combination let smaller companies access infrastructure that previously required a dedicated data center, which is one of the structural reasons the 2010s saw such a wave of startups.
Cloud also enabled big data: the storage and processing of datasets too large for a single machine to handle. Distributed computing frameworks (Hadoop, then Spark), columnar databases (BigQuery, Redshift, Snowflake), and data lake architectures became standard for organizations dealing with multi-terabyte working sets. Machine learning models trained on those datasets (recommender systems, fraud detection, then computer vision and natural language) drove a lot of the application-level innovation through the mid-2010s. The 2022 LLM moment is downstream of all of this: GPT-3 and what followed required cloud-scale compute that simply didn’t exist a decade earlier.
The transformation goes well beyond technology. Cloud and big data have reshaped industries from healthcare and finance to manufacturing and transportation. They’ve also concentrated computing infrastructure into a handful of hyperscalers, which is its own architectural and political question.
The future of the internet: AI, blockchain, and beyond#
Three technology threads are reshaping the internet right now. AI is the most consequential: LLM-powered assistants (ChatGPT November 2022, Claude 2023, Gemini 2024) have moved from research curiosity to mass consumer products with hundreds of millions of users in under three years, search engines are being rebuilt around AI summarization, coding assistants are part of the standard developer workflow, and the infrastructure side is being retooled for inference at scale. Blockchain went through the 2021-2022 cryptocurrency boom and the November 2022 FTX collapse; the speculative bubble deflated but the technically interesting applications (stablecoins, on-chain settlement, zero-knowledge proof systems) continue to advance. Quantum computing has moved from theoretical to commercially relevant, with IBM, Google, and IonQ shipping systems in the hundreds of qubits as of 2026; the security-relevant consequence is post-quantum cryptography, since a sufficient quantum computer would break RSA and elliptic-curve crypto via Shor’s algorithm, and NIST finalized replacement standards (CRYSTALS-Kyber for key exchange, CRYSTALS-Dilithium for signatures) in August 2024.
Privacy and security regulation continues to evolve in parallel. GDPR (2018) and the California, Brazilian, and other equivalents have changed how platforms handle data; whether they’ve changed user outcomes is a question security practitioners argue both sides of. The internet that emerges from the next decade will be shaped by these threads and others nobody is predicting yet, in ways the people who built the original protocols never anticipated.
Cybersecurity: the dark side of the web#
Every era of the internet has had its security story running underneath, and the historical phases above each have one. ARPANET produced the first worm. The BBS and Usenet years produced the first scam culture and the first generation of script kiddies. The web era produced the first commercial-scale credit-card theft and the cryptographic transitions that came with it. The mobile era produced spyware-as-a-service, Pegasus, and stalkerware. The cloud era produced supply-chain compromise at unprecedented scale. The security history isn’t a separate story from the internet’s history; it’s the same story told from the defender’s seat.
Three episodes in particular keep showing up in the work of modern red and blue teams, because the protocols and patterns they exposed are still central to how the internet works in 2026.
Cybersecurity in 2026 is no longer just about viruses. It covers nation-state espionage, critical infrastructure protection, ransomware-driven cybercrime, and the integrity of the routing fabric itself. Most of what operators work on today traces back to design decisions made in the trust-rich early internet, when “everyone on the network is a researcher at a friendly institution” was a reasonable working assumption.
The Morris Worm (1988)#
The first major internet security event. Robert Tappan Morris, a graduate student at Cornell University, released a self-replicating program on November 2, 1988 that exploited vulnerabilities in Unix Sendmail (the debug command), Fingerd (a stack buffer overflow), and rsh/rexec (weak or absent authentication), plus a dictionary attack against weak passwords. Morris intended the worm as an experiment to measure the size of the internet, but his propagation rate was too aggressive: the worm reinfected hosts faster than they could handle, and roughly 6,000 of the internet’s ~60,000 hosts (about 10%) were rendered unusable.
The response was the formation of CERT/CC at Carnegie Mellon in November 1988, with DARPA funding. CERT was the first computer emergency response team and remains operational, now under SEI. The Morris Worm prosecution was also the first conviction under the 1986 Computer Fraud and Abuse Act; Morris got three years probation and 400 hours of community service. He later co-founded Y Combinator with Paul Graham.
The inherent trust of BGP#
One of the most consequential vulnerabilities in the internet’s infrastructure is the Border Gateway Protocol (BGP). Designed in the 1980s, BGP operates on trust: if an autonomous system announces “I know the route to YouTube,” other ASes believe it.
In BGP hijacking, an attacker (or a misconfigured router) announces a more specific route for an IP block they don’t own. Global traffic for that IP is then routed through them. The two canonical incidents:
- February 24, 2008: Pakistan Telecom tried to block YouTube domestically by announcing a more-specific route for YouTube’s address block to its own customers. The announcement leaked outside Pakistan via PCCW (Pakistan Telecom’s upstream), and YouTube was effectively unreachable globally for about two hours.
- April 24, 2018: An attacker hijacked Amazon’s Route 53 DNS announcements, redirecting MyEtherWallet.com traffic to a phishing site that drained roughly $150,000 in Ethereum. The attack lasted about two hours before legitimate routes propagated again.
The mitigation effort that emerged is RPKI (Resource Public Key Infrastructure) and Route Origin Validation: cryptographically signing which networks are authorized to announce which prefixes. Adoption has been gradual; as of 2026, most of the major transit networks validate, but the long tail of smaller ISPs still doesn’t.
DNS cache poisoning#
The Domain Name System (DNS) translates google.com to an IP address. In July 2008, researcher Dan Kaminsky disclosed a flaw in the DNS protocol that allowed attackers to inject false DNS records into a caching server with high reliability. The result: an attacker could redirect a user typing bank.com to a malicious clone without the user (or their browser) noticing anything wrong.
Kaminsky’s coordinated disclosure was the largest the security community had ever organized. He worked privately with the major DNS vendors (BIND, Microsoft, Cisco, Nominum) starting in March 2008. A coordinated multi-vendor patch released on July 8, 2008 mitigated the vulnerability across all major implementations. Kaminsky disclosed the full technical details on July 21, by which time most production DNS servers had been patched. The episode established the playbook for coordinated multi-vendor security disclosures.
Modern threats#
The current threat picture includes:
- Ransomware: encrypted-files-for-payment evolved from the WannaCry / NotPetya era into the RaaS ecosystem of 2024-2026 (see the Ransomware Analysis post for the full picture, including the LockBit takedown and BlackCat exit scam).
- DDoS amplification: attackers send small spoofed requests to misconfigured services (DNS open resolvers, memcached, NTP) that return much larger responses to the victim’s IP. The 1.35 Tbps attack on GitHub in February 2018 was a memcached amplification.
- Supply chain attacks: the 2020 SolarWinds Orion compromise, the 2021 Codecov breach, the ongoing npm/PyPI supply chain incidents. All inherit the trust model of “the package my CI system fetched is what the maintainer intended to publish.”
- BGP hijacking: still landing in 2026 against networks that haven’t adopted RPKI ROV.
The architectural response is Zero Trust, where “inside the firewall” no longer means “safe.” Every connection requires authentication, every device is treated as potentially compromised, and the network perimeter stops being a primary defensive layer. Zero Trust adoption has been slow because it requires retrofitting trust into a network that was designed without any.
Trivia#
Thirty pieces of trivia about the birth of the internet:
- The first domain name ever registered was Symbolics.com, registered on March 15, 1985, by the Symbolics computer company. The domain still resolves, currently to a museum-style page about the company’s history.
- The first YouTube video ever uploaded was “Me at the zoo”, uploaded by co-founder Jawed Karim on April 23, 2005.
- The world’s first website, info.cern.ch, was created in 1991 by Tim Berners-Lee and is still online today , restored from a backup by CERN in 2013.
- The first instant messaging program, Talkomatic, was developed in 1973 on the PLATO System at the University of Illinois.
- The first webcam, set up in 1991 by a team at the University of Cambridge Computer Lab, pointed at the shared coffee pot so people could see whether there was coffee available without walking down the hall. It moved online in 1993 when the lab got web access.
- The
@symbol in email addresses was chosen by Ray Tomlinson in 1971 because it was the only punctuation character on his keyboard that wasn’t already used as part of a username. - The first commercial spam email was sent on May 3, 1978, by Gary Thuerk, a DEC marketing rep, to about 400 users on ARPANET. Thuerk later claimed it generated $13 million in DEC sales; that figure is his own retrospective and is widely disputed by people who were on ARPANET at the time.
- The first emoticon
:-)was created by computer scientist Scott Fahlman at Carnegie Mellon on September 19, 1982, at 11:44 AM, on a department bboard. The original post is preserved. - The first self-replicating program described as a “virus” was Creeper, written in 1971 by Bob Thomas at BBN. It moved between PDP-10s on ARPANET and displayed
I'M THE CREEPER: CATCH ME IF YOU CAN. Ray Tomlinson (the email guy) later wrote Reaper to chase and delete Creeper, arguably the first antivirus. - The first website to sell goods online was a bookstore called Book Stacks Unlimited (Books.com), founded by Charles Stack in 1992 and later acquired by Barnes & Noble.
- The term “surfing the internet” was coined by librarian Jean Armour Polly in 1992.
- Amazon was originally named Cadabra. Bezos founded Cadabra Inc. on July 5, 1994, then renamed it Amazon later in 1994 after a lawyer misheard “Cadabra” as “cadaver” on a phone call.
- Google was originally named BackRub, because Page and Brin’s PageRank algorithm worked by analyzing backlinks. They renamed it in 1997 when they registered the google.com domain.
- The “404” error code was not named after a room at CERN. That’s a myth. It follows the standard HTTP status code schema (4xx for client errors).
- The first banner ad appeared on HotWired.com on October 27, 1994. It was for AT&T and read “Have you ever clicked your mouse right HERE? YOU WILL.” The click-through rate was 44%, a number no banner ad has come close to since.
- Before the World Wide Web, the internet was almost entirely text-based.
- The first item sold on eBay (then called AuctionWeb) was a broken laser pointer for $14.83, in 1995. Founder Pierre Omidyar emailed the buyer to confirm they understood it was broken; the buyer replied that they collected broken laser pointers.
- “Wi-Fi” doesn’t stand for anything. It was a marketing term picked by the Wi-Fi Alliance via a branding agency in 1999. The “Wireless Fidelity” backronym is folklore, and Phil Belanger of the Alliance has confirmed it doesn’t stand for anything.
- China’s Great Firewall (formally the Golden Shield Project, launched 1998, operational from 2003) is the most extensive national internet filtering system ever built. It blocks Google, Facebook, Twitter, Wikipedia in some configurations, most Western news outlets, and any service that won’t host its data inside Chinese jurisdiction. It actively inspects encrypted traffic, blocks VPN protocols on detection, and is staffed by an estimated 2 million people working in monitoring and enforcement roles. The internet of 2026 is meaningfully bifurcated between the open web most readers of this post use and the parallel network roughly 1.4 billion people live inside.
- There are seven keys to the internet. ICANN holds quarterly Root Key Signing Ceremonies where seven Trusted Community Representatives, each holding a physical key, witness the signing of the DNS root key. It’s a real ceremony, though the popular framing of “they could restart the internet” overstates what the key material does. It controls DNSSEC root trust, not internet connectivity.
- Email is older than the World Wide Web. Email started in the late 1960s and 1970s on time-sharing systems and then ARPANET; the Web started in the early 1990s.
- The most expensive domain name ever sold was Voice.com, purchased by Block.one for $30 million in 2019.
- The first tweet was sent by Jack Dorsey on March 21, 2006: “just setting up my twttr”.
- Approximately 500 hours of video are uploaded to YouTube every minute as of 2024 (up from ~300 hours per minute in 2014). The exact rate keeps climbing.
- The internet consumes roughly 1-2% of global electricity by most current estimates, and data centers in particular are projected to keep growing as a share of total demand through the late 2020s, partly driven by AI workloads.
- The
HTTP://part of a URL is a protocol identifier. In the early days you had to type it; today browsers assume it (or HTTPS). - The first firewall was a physical router configured to block traffic, not software. Bill Cheswick and Steve Bellovin at AT&T Bell Labs are credited with building one of the earliest examples in the late 1980s, and Cheswick and Bellovin’s 1994 book Firewalls and Internet Security is the foundational text.
- “Spam” got its name from a 1970 Monty Python sketch in which a group of Vikings drown out conversation by chanting “spam, spam, spam, spam.” Early Usenet users started calling repetitive unwanted messages “spam” by analogy.
- The internet has its own Hall of Fame at internethalloffame.org , inducted classes including Berners-Lee, Cerf, Kahn, Kleinrock, Postel, and many others.
- More than half of all web traffic comes from automated bots, not humans. Imperva’s 2024 Bad Bot Report put the figure at about 49.6% bot traffic, with bad bots specifically accounting for ~32% of all internet traffic.
What survives#
The internet didn’t get built by anyone with the authority to design it. It got built by a few thousand researchers, hobbyists, and entrepreneurs who happened to make compatible choices over five decades, with periodic standardization moments (TCP/IP, HTTP, DNSSEC, RPKI) where the community agreed on what to do next. That decentralized origin story explains both the parts that work surprisingly well and the parts that fail predictably.
Most of what red and blue teams work on in 2026 traces back to design decisions made in the trust-rich early internet. BGP still operates on trust, DNS resolution still has to be hardened against poisoning, email is still federated and largely unauthenticated, and the legacy of “everyone on the network is a friendly researcher” is visible in every protocol decision the community is now retrofitting security into. Knowing how the technology got here makes the work of securing it easier, because most of the live attack surface is downstream of decisions made decades ago.
For security and infrastructure professionals, staying informed about both the history and the current threat picture is part of the job. The protocols and patterns aren’t going to change as fast as the threats against them.
