Everything Is a Target
A Plain-Language Guide to Personal Cybersecurity
- 8 chapters
- 51m
- Cybersecurity
- Free · no sign-up
The book explores supply chain attacks that can compromise even trusted software vendors, and discusses how artificial intelligence is changing cybersecurity threats. It examines the dark web where stolen data is sold, and explains why understanding these systems helps protect personal privacy. Each chapter builds on previous concepts to create a complete picture of modern digital defense.
Whether you're setting up your first home network or trying to understand AI-powered threats, this plain-language guide shows you exactly what to do. You'll learn how to backup important files, recognize when you've been breached, and make smart choices about security tools. This book is essential for anyone who uses the internet regularly.
Listen
-
Read this chapter
Overview
Computer security, also known as cybersecurity, digital security, or IT security, is about protecting software, systems, and networks from threats that could lead to stolen data, damaged hardware, or disrupted services. As people rely more on computers, the internet, and wireless technology, the need for strong computer security has grown. This is especially true with the rise of smart devices like smartphones and IoT-enabled appliances. The systems managing critical areas such as power grids, elections, and banking are now more vulnerable than ever. While much of this work involves digital tools like passwords and encryption, physical protections like locks are still used. IT security is a part of information security but doesn’t fully match all aspects of it.
Vulnerabilities and attacks
A vulnerability is a flaw in a computer system that can be exploited to compromise its security, and most known vulnerabilities are listed in the Common Vulnerabilities and Exposures database. An exploitable vulnerability has at least one working attack method, and a threat actor—often called a hacker—uses these to carry out threats against systems. In April 2023, the UK Department for Science, Innovation & Technology released findings from a survey of businesses, charities, and education institutions, showing that 32% of businesses and 24% of charities reported breaches in the past year. Medium and large businesses were hit more often, but small and midsize businesses are increasingly vulnerable because they lack advanced defense tools. These smaller entities are most affected by malware, ransomware, phishing, man-in-the-middle attacks, and DoS attacks. Domestic users face untargeted cyberattacks that use the Internet's openness to spread threats like phishing, ransomware, water holing, and scanning.
Backdoor
A backdoor is a hidden way into a computer system that skips normal security checks, and it can be built in from the start or created by bad configuration. These flaws are more of a worry for big companies and databases than for regular people, but they still pose a serious risk. Backdoors might be placed there by someone with permission to give access, or by hackers who want to cause harm. Criminals often use malware to plant these backdoors, which then let them take over a system from afar. Once inside, they can change files, steal personal data, install unwanted programs, or fully control the computer. Because backdoors are designed to stay hidden—sometimes buried in code or firmware—they’re tough to find and usually need deep knowledge of how the system works to spot them.
Denial-of-service attack
A denial-of-service attack (DoS) makes computers or networks stop working properly, preventing legitimate users from accessing them. Attackers lock users out with repeated wrong password attempts or flood systems with traffic until they crash. Some attacks come from single sources, making them easier to block with firewall rules, while others are distributed denial-of-service (DDoS) attacks originating from many places simultaneously—like computers taken over by a botnet. One DDoS type is distributed reflective denial-of-service (DRDoS), where innocent systems are tricked into sending traffic to targets. These attacks are powerful because they can be amplified, meaning attackers need little bandwidth themselves to cause major damage. For more on why people do this, see the section on attacker motivation.
Physical access attacks
A direct-access attack happens when someone gains physical access to a computer they're not supposed to, usually to steal data or copy information. They might change the system’s settings, install malware like keyloggers or worms, or even add hidden listening devices. Even if the computer has standard protections, attackers can get around them by booting from a CD or other external media. Tools like disk encryption and the Trusted Platform Module are meant to stop this kind of access. Direct service attacks are similar, letting an attacker reach into a computer’s memory through certain hardware components like graphics cards or network adapters that are designed to connect directly to the system's memory.
Eavesdropping
Eavesdropping happens when someone secretly listens in on computer communications, usually across an unsecured network. It’s hard to detect because it doesn’t slow down systems like other attacks do. Attackers can grab data sent between devices without needing to stay connected, sometimes even planting software on a machine and coming back later to collect information. Even closed systems aren't safe—electromagnetic signals from hardware can be monitored, a technique the NSA calls TEMPEST. The FBI and NSA have used programs like Carnivore and NarusInSight to spy on internet providers. Using a VPN, strong encryption, and HTTPS instead of HTTP helps protect against this kind of spying.
Malware
Malware is any software created with the intent to damage a computer or its users. Once it gets onto your system, it can steal private data like passwords, financial details, or business information. It might even let an attacker take over your device entirely, while also being capable of deleting or damaging your files for good.
Multi-vector, polymorphic attacks
In 2017, a new kind of cyber threat appeared that’s both multi-vectored and polymorphic. These attacks use more than one method to get in—like through the web, email, or apps—and they can move around inside a network once they're in. What makes them especially dangerous is that they change their form as they spread, making it hard for traditional security systems to catch them. They’re not just one type of attack but a mix, and they adapt constantly, using things like viruses, worms, or trojans that morph to avoid detection.
-
Read this chapter
Overview
Information security standards, also known as cybersecurity standards, are published guidelines meant to protect users and organizations from cyber threats. These standards cover the people, devices, networks, software, and data that make up a digital environment. They outline security concepts, technologies, policies, and best practices for handling incidents, along with training and implementation steps. Some standards also include ways to measure compliance, audit processes, and certification for those who follow them. The main goals are to reduce risks, keep development consistent, and set minimum requirements in vulnerable industries. These standards come from national and international groups to help improve overall cybersecurity and create a unified response.
History
Over the past several decades, people working in cybersecurity have come together in both domestic and international settings to build the policies and practices we rely on today. Much of this work began in the 1990s with efforts from the Stanford Consortium for Research on Information Security and Policy. In 2016, a study found that 70% of organizations in the U.S. were using the NIST Cybersecurity Framework as their main guide for IT security, though it can be costly to implement. At the same time, law enforcement agencies have faced challenges when trying to conduct cyber investigations across borders, especially when targeting criminals on the dark web. These cross-border operations raise difficult legal questions that still don’t have clear answers. As these tensions continue, they may help shape better global cybersecurity norms.
ISO/IEC 27000 Family of Standards
The ISO/IEC 27000 series is a set of international standards from the International Organization for Standardization and the International Electrotechnical Commission that helps organizations protect their information. At its core is ISO/IEC 27001, which sets requirements for an Information Security Management System using a risk-based approach and the Plan-Do-Check-Act cycle. Supporting it is ISO/IEC 27002, offering practical guidance on implementing controls across areas like network and physical security. For risk management specifically, ISO/IEC 27005 provides a framework tailored to identifying and treating information security risks. As cloud computing grew, standards like ISO/IEC 27017 and ISO/IEC 27018 were created to address cloud-specific concerns and protect personally identifiable information. Incident management is covered by ISO/IEC 27035, while ISO/IEC 27701 helps align information security with privacy regulations such as GDPR.
ISO/IEC 15408
ISO/IEC 15408, also known as the Common Criteria, is an international standard used to check how secure IT products and systems are. It gives a clear way to set security goals, put protections in place, and test whether those protections work. The standard has five parts: the first explains the basic ideas and model, the second lists security functions like access control and encryption, the third sets levels of assurance from EAL1 to EAL7, the fourth outlines how evaluations are done, and the fifth offers ready-made security packages for common products. Certificates under this system are recognized through the Common Criteria Recognition Arrangement, which helps companies save time and money when selling globally. The European Union has also built its own cybersecurity certification system based on this standard.
ISO/SAE 21434
ISO/SAE 21434 is a cybersecurity standard created by ISO and SAE working groups, published in August 2021, that outlines security practices for developing road vehicles. It’s part of a larger effort tied to European Union regulations being developed for vehicle cybersecurity. In coordination with the EU, the UNECE has introduced a Cyber Security Management System certification required for vehicle-type approval, defined under UN Regulation 155. ISO/SAE 21434 serves as a technical standard that helps show compliance with those rules. A related effort comes from UNECE WP29, which sets regulations for vehicle cybersecurity and software updates.
EN 18031
The EN 18031 standards come from the European Committee for Standardization, or CEN, working together with CENELEC, the European Committee for Electrotechnical Standardization. These rules apply to radio-based devices and systems, and they connect to the Radio Equipment Directive from 2014/53/EU. That directive has its own Delegated Act that helps make sure manufacturers meet certain requirements across Europe. The EN 18031 series sets out testing methods, performance expectations, and security guidelines so that different systems can work together smoothly and keep up with new industry demands.
NERC CIP
The North American Electric Reliability Corporation, or NERC, sets and enforces cybersecurity rules to keep the power grid safe across the U.S., Canada, and parts of Mexico. These rules focus on protecting key equipment, setting up digital security perimeters, training people, responding to incidents, and planning recovery efforts. The main standards are part of a series called Critical Infrastructure Protection, or CIP, with specific requirements numbered from CIP-002 through CIP-014. Anyone who operates or owns parts of the power system in NERC’s area must follow these rules, and the Federal Energy Regulatory Commission makes sure they do. If they don’t, they can face big fines.
NIST Cybersecurity Standards
The National Institute of Standards and Technology, or NIST, is a U.S. federal agency under the Department of Commerce that helps create and update cybersecurity standards, guidelines, and best practices. Originally formed to protect federal information systems, its work has grown to influence security programs worldwide. NIST uses a risk-based approach with five main functions: Identify, Protect, Detect, Respond, and Recover. These steps help organizations manage cybersecurity risks. While federal agencies must follow NIST standards, many private companies in finance, healthcare, manufacturing, and other fields adopt them voluntarily because they’re clear, flexible, and thorough.
-
Read this chapter
Overview
A supply chain attack happens when cybercriminals target weaker parts of a company’s network to sneak into the bigger system later. These attacks can happen in any industry, like finance, oil, or government. They can involve software or hardware — for example, by adding malware or spying devices during manufacturing or shipping. According to Symantec's 2019 Internet Security Threat Report, these attacks rose by seventy-eight percent in 2018. A supply chain is a complex system of businesses working together to get products from vendor to consumer. In cybersecurity, this kind of attack might involve tampering with electronics like computers or power systems to plant hidden code. It can also mean attacking a small software part that’s used by larger programs, so the malware spreads through the whole system.
Attack framework
Supply chain attacks start when hackers target the weakest link in a network, often smaller companies that are less protected. These advanced persistent threats don’t usually go after big targets directly, but instead focus on third-party software or products. In 2008, European law enforcement uncovered a credit-card fraud ring that used devices inserted into Chinese-made card readers to steal account details and make unauthorized transactions, leading to about $100 million in losses. According to a Verizon Enterprise investigation, 92% of cybersecurity incidents happened at small firms, showing how vulnerable interconnected supply chains can be.
Risks
Supply chain attacks are a major threat today, affecting not just tech companies but industries like oil, retail, and pharmaceuticals. The Information Security Forum points out that sharing information with suppliers is necessary for the supply chain to work, but it also creates risk—compromised data from outside can be just as damaging as from inside. Muhammad Ali Nasir from the National University of Computer and Emerging Sciences links this risk to globalization, noting that as supply chains spread across the globe and involve more players, there are more weak points. A cyberattack on one part can hurt many connected organizations at once due to the ripple effect. Poorly managed systems increase vulnerability, risking customer data, production delays, and reputational damage.
Compiler attacks
As of May 3, 2019, Wired identified a common link in recent software supply chain attacks. The infections are believed to have originated from pirated versions of widely used compilers found on illegal websites. These included corrupted builds of Apple's Xcode and Microsoft Visual Studio. In theory, using different compilers could help detect such attacks, since the compiler acts as the trusted foundation.
Target
In late 2013, Target was hit by a massive data breach that compromised the credit and debit card information of around 40 million customers. The attack occurred between November 27 and December 15, when malware infected the point-of-sale systems in over 1,800 stores. Even though Target had installed a $1.6 million cybersecurity system and employed security specialists, the breach still slipped through. Investigators believe hackers gained access via a third-party vendor—Fazio Mechanical Services, an HVAC provider based in Pennsylvania—using stolen credentials. The company’s profits dropped 46% in the final quarter of 2013, and it spent about $61 million dealing with the fallout. Customers filed nearly 100 lawsuits, accusing Target of negligence.
Stuxnet
Stuxnet was a computer worm widely believed to be a joint U.S.-Israeli operation, though neither government has confirmed that. It targeted industrial control systems, especially those managing factory machinery and nuclear equipment. The worm specifically worked on programmable logic controllers, giving false data to monitoring systems while secretly disrupting operations. Most infections happened in Iran, with analysts believing its main goal was the Natanz uranium facility. It likely entered through infected USB drives, requiring someone with physical access to plant networks—possibly engineers or maintenance workers, either knowingly or not. Once inside, it spread on its own, using multiple zero-day flaws in Windows systems to move across machines running Siemens software. Kevin Hogan, a senior security director at Symantec, said most infections occurred in Iran.
ATM malware
In recent years, malware like Suceful, Plotus, Tyupkin, and GreenDispenser has attacked ATMs around the world, especially in Russia and Ukraine. GreenDispenser lets attackers remove cash from infected machines, sometimes showing an “out of service” message, but those with access can drain the vault and delete the malware without a trace. Other versions usually steal magnetic stripe data and trigger cash withdrawals. The Tyupkin malware, active in March 2014 on over 50 ATMs in Eastern Europe, is thought to have spread to the U.S., India, and China. It targets ATMs made by major companies running 32-bit Windows systems and shows how much money each machine holds, allowing attackers to withdraw 40 notes from any cassette.
NotPetya / M.E.Doc
In June 2017, security researchers identified M.E.Doc, a Ukrainian financial software, as likely the starting point for NotPetya malware. Microsoft researchers said the infection may have come from a compromised update through M.E.Doc, though how exactly it was compromised wasn't clear. The software's creators initially denied the claim but later withdrew their statement and said they were cooperating with investigators. NotPetya acted like ransomware at first, encrypting files and demanding bitcoin, but the email for decryption keys was shut down, leaving victims stuck. Unlike WannaCry, it had no kill switch. The attack hit banks, an airport, Kyiv Metro, pharmaceutical companies, and Chernobyl's radiation systems in Ukraine, spreading globally to Russia, UK, India, and U.S. It used EternalBlue, a vulnerability from the NSA that had also powered WannaCry. NotPetya spread using SMB, PsExec, and WMI protocols. Ukrainian police said M.E.Doc employees could face criminal charges for ignoring warnings from antivirus firms about its weak cybersecurity. Colonel Serhiy Demydiuk of Ukraine's CyberPolice said the company "knew about it," and later reported that M.E.Doc cooperated with investigators.
-
Read this chapter
Overview
The most recent AI boom started in the 2020s, gaining speed and attention, with generative AI technologies like large language models, image, and video models leading the way. By 2026, major companies such as OpenAI, Anthropic, Google DeepMind, and Meta Superintelligence Labs had released frontier models, while Chinese firms including Alibaba Cloud, DeepSeek, and Moonshot AI also made significant contributions. Many of these models are open-source or open weights, meaning they can run locally on personal devices. Key developments that led to this boom include the transformer architecture from 2017, the GPT model in 2018, and reinforcement learning from human feedback. As the field advanced, techniques like reasoning models, mixture of experts, and multimodal learning emerged. Observers such as Ian Bremmer, Nicholas Thompson, and Kai-Fu Lee have noted an AI arms race or Cold War between the U.S. and China. The boom has also raised concerns about environmental impact, copyright issues, deepfakes, cybersecurity threats, and even existential risks like artificial general intelligence. In 2025, ChatGPT became the fourth most-visited website globally, behind only Google, YouTube, and Facebook.
History
In 1950, Alan Turing introduced the idea of machines that could think, proposing what became known as the Turing test to distinguish human-written text from machine-generated content. The term "artificial intelligence" was coined in 1956 by John McCarthy, who also helped organize the Dartmouth conference, marking the formal start of AI as an academic field. That same year, McCarthy created LISP, a programming language designed for list processing and widely used in AI development. In 1962, he founded the Stanford Artificial Intelligence Laboratory, or SAIL, and was also part of MIT's first AI lab, now called the MIT Computer Science and Artificial Intelligence Laboratory. Around 1966, Joseph Weizenbaum developed ELIZA, the first chatbot, intended as an emotional experiment. More recently, in 2022, DALL-E 2 and Midjourney launched as text-to-image models, followed by ChatGPT from OpenAI, which quickly became the fastest-growing software application, reaching over 100 million users within two months.
Biomedical
In 2020, a program called AlphaFold, created by DeepMind, made a huge leap in predicting how proteins fold. It scored over 90 on a global test called CASP's Global Distance Test. A structural biologist and Nobel Prize winner named Venki Ramakrishnan called the result "a stunning advance on the protein folding problem." Being able to predict protein structures from their amino acid sequences could speed up drug discovery and improve our understanding of diseases.
Images and videos
Generative AI has developed rapidly over the past decade. In 2015, Google introduced DeepDream, an AI that transforms existing images into surreal, hallucinogenic visuals. Then, in January 2021, OpenAI launched DALL-E, enabling users to create images from text descriptions. After that came models like Google’s Gemini. By 2024, tools such as OpenAI’s Sora made text-to-video generation widely used, especially in advertising for faster and cheaper production. These advances are happening faster than detection methods can keep up. As everyday people gain access to these tools, concerns about misuse grow. Misinformation has already spread online through fake videos, creating serious security risks.
Language
By August 2026, major AI companies like OpenAI, Anthropic, Google DeepMind, Meta Superintelligence Labs, SpaceXAI, and Nvidia had released large frontier models in the U.S., while in China Alibaba Cloud, DeepSeek, Moonshot AI, Z.ai, and ByteDance were also pushing forward. France’s Mistral AI stood out as Europe’s biggest player. At the same time, open-source options included smaller language models that could run on personal devices, such as those from Alibaba's Qwen, Google's Gemma, Nvidia's Nemotron, and Meta's Muse Glimmer. Among the well-known models, GPT-3 came out in 2020 from OpenAI, and then GPT-4 launched on March 14, 2023, powering Microsoft’s Bing search engine. Other notable models include PaLM and Gemini from Google, and LLaMA from Meta Platforms.
Music and voice
In 2016, Google’s DeepMind introduced WaveNet, a system that could generate raw audio of speech and piano, identifying speakers to create different voices—laying the groundwork for future AI models. By 2020, OpenAI released Jukebox, the first large-scale tool capable of producing songs in various genres and styles. In 2024, high-fidelity AI music became publicly available, leading to lawsuits from major record labels over copyright concerns. Also in 2024, services like Udio and Suno AI were targeted in these legal battles. In March 2020, 15.ai was founded, enabling audio deepfakes. Tools such as ElevenLabs allowed anyone to generate artificial vocals from existing clips, meaning any public voice—celebrities or politicians—could be used for fake robocalls, including attempts to influence elections.
Energy
The growing need for electricity to run AI systems has stretched power grids to their limits, causing fossil fuel plants to stay online longer than they would otherwise. Because renewable energy can’t provide steady, round-the-clock power, big tech companies are turning to nuclear energy. Microsoft, Google, and Amazon are all investing in nuclear projects. In September 2024, Microsoft struck a deal to buy electricity from a reactor at Three Mile Island that shut down in 2019 and is scheduled to restart in 2028 for its data centers. That reactor is located next to the unit involved in the worst nuclear accident in U.S. history, which occurred in 1979.
Cultural
People are split on AI, according to a Pew Research report. Most Americans worry about not having control over it and how it might hurt human creativity. A 2025 study by Nikolova and Angrisani found that people are especially unsure about AI being used in personal relationships, but they’re more okay with it in medicine—like helping create new antibiotics.
-
Read this chapter
Overview
The dark web is part of the internet that exists on special networks called darknets, which need specific tools or permissions to enter. Unlike regular websites, these hidden areas let people communicate and do business without revealing who they are or where they're located. The dark web is a small piece of something bigger called the deep web—content not found by normal search engines, though many confuse the two terms. Some darknets are private friend networks, while others like Tor, Hyphanet, I2P, and Riffle are widely used and run by groups or individuals. People who use the dark web often call regular internet "clearnet" because it's not encrypted. Tor, for example, uses a method called onion routing and works under the .onion domain.
Definition
The dark web is a small part of the deep web, which includes all web content not indexed by search engines. The term "dark web" first appeared in 2009, though the actual dark web likely existed earlier. Unlike the surface web, accessing the dark web requires special software like Tor, short for "The Onion Routing" project. These sites use the .onion domain and are only reachable through networks designed for anonymity. Tor encrypts users’ data by routing it through many servers, making it nearly impossible to trace a user’s location or identity. This layered encryption protects both users and website hosts from being tracked, allowing confidential communication, file sharing, and browsing without revealing who’s involved.
Content
In December 2014, Gareth Owen of the University of Portsmouth published a study showing that child pornography was the most common type of content on Tor, followed by black markets, with the busiest sites involved in botnet operations. Whistleblowing platforms and political discussion forums were also present, along with services related to Bitcoin, fraud, and mail-order operations. By December 2020, there were an estimated 76,300 active Tor sites, though only about 18,000 contained original material. In July 2017, Roger Dingledine, one of the three founders of the Tor Project, said that Facebook is the biggest hidden service. The dark web comprises only 3% of the traffic in the Tor network. A February 2016 study from King's College London offered another way to categorize content, emphasizing the illicit use of .onion services.
Ransomware
Ransomware groups use the dark web to run their operations, from planning to profit. Ransomware-as-a-Service models let criminals recruit partners through forums like RAMP and, before they were shut down after the 2021 Colonial Pipeline attack, Exploit and XSS. These sites offer toolkits, share profits—usually 60 to 80% of what victims pay—and vet new members. Groups like LockBit, ALPHV/BlackCat, and Cl0p also run data leak sites on the Tor network, a tactic that started with Maze in November 2019. That's where they publish stolen data if the victim refuses to pay. Many ransomware actors don't do everything themselves—they buy access from initial access brokers who compromise systems using phishing, weak passwords, or stolen credentials, then sell that access on underground forums. This creates a streamlined criminal supply chain that makes it easier for anyone to launch an attack.
Darknet markets
Commercial darknet markets are online spaces where people trade illegal goods like drugs, weapons, and stolen identity info, usually using Bitcoin for payment. The first big one, Silk Road, appeared in 2011 and was later shut down by authorities. Even after being closed, new markets quickly replaced them—by 2020, at least thirty-eight were active. These sites work like eBay or Craigslist, where users interact with sellers and leave reviews. One study looked at a popular market called Evolution from 2013 to 2015 and found that while some details about products seemed accurate, the actual quality of drugs was often different from what was listed. These markets also spread leaked credit card information for free.
Bitcoin services
Bitcoin's popularity in the digital underground stems from how it lets users obscure both their identity and actions. In dark web markets, people used services swapping bitcoin for virtual game currency like World of Warcraft gold, then converted that back into real money. Some tools like tumblers are found on networks such as Tor, while others like Grams connect directly to darknet sites. A study led by Jean-Loup Richet, a research fellow with the United Nations Office on Drugs and Crime, uncovered new money laundering methods using these tools, including escrow systems. As bitcoin grew in importance within digital spaces, it became a prime target for fraudsters. Since its introduction in 2014, cybercriminals have used it to launch more than 140 attacks on companies, spawning further criminal networks and the rise of cyber extortion.
Hacking groups and services
Hackers operate both solo and in organized networks, offering their skills on the dark web. Among the groups active in this space are xDedic, hackforum, Trojanforge, Mazafaka, dark0de, and the TheRealDeal darknet market. These entities have been involved in targeting individuals suspected of child exploitation, as well as providing tools for attacking financial institutions. Monitoring efforts come from government and private organizations, with research published in the Procedia Computer Science journal examining the methods used. The dark web has also facilitated large-scale DNS-based attacks, and many fake .onion sites exist that distribute malware disguised as useful tools. In 2023, around 100,000 ChatGPT user login details were sold, with researchers concluding most passwords were taken by the Raccoon data-stealing virus.
Financing and fraud
Scott Dueweke, the president and founder of Zebryx Consulting, says that Russian electronic currencies like WebMoney and Perfect Money are behind most illegal activity online. In April 2015, Flashpoint got a five-million-dollar investment to help clients gather intelligence from the deep and dark web. There are many carding forums, PayPal and bitcoin trading sites, as well as fraud and counterfeiting services. A lot of these sites are scams themselves. Phishing through cloned websites and other fake sites is common, with darknet markets often promoted using fraudulent URLs.
-
Read this chapter
Overview
A theory of everything is a single framework explaining all physical behavior in the universe. The idea began with Isaac Newton, who linked earthly gravity with planetary motion. Later, James Clerk Maxwell showed electricity and magnetism are connected. Albert Einstein's theory of relativity built on that, revealing how they relate to each other. By the 1930s, Paul Dirac helped merge relativity and quantum mechanics, creating quantum electrodynamics. As understanding of nuclear forces grew, the Standard Model was formed, combining three of the four fundamental forces, but not gravity. General relativity handles gravity well, yet it hasn't been unified with quantum mechanics—especially at the tiny scales of the Planck length. One promising idea is string theory, which says that all particles are made of vibrating strings and suggests extra dimensions beyond the usual four. It remains a candidate for a complete theory, though not without controversy or testable predictions.
Antiquity to 19th century
In the late 17th century, Isaac Newton described gravity in a way that suggested not all forces come from things touching each other. His Mathematical Principles of Natural Philosophy brought together Galileo's work on falling objects, Kepler's laws of planetary motion, and the tides under one law: universal gravitation. This was the first big unification in physics. Pierre-Simon Laplace later said that if an intellect knew all forces and positions in nature, it could predict everything using a single formula. But modern science shows that uncertainty is unavoidable, even without quantum mechanics—chaos theory alone makes long-term prediction impossible. In 1820, Hans Christian Ørsted discovered a link between electricity and magnetism, which James Clerk Maxwell completed in 1865 with his theory of electromagnetism, achieving the second great unification. Michael Faraday searched for a connection between gravity and electricity in the 1849–1850 experiments but found none.
Early 20th century
In the late 1920s, quantum mechanics revealed that chemical bonds result from electrical forces, supporting Dirac's claim that the laws of physics and chemistry are largely understood. Around 1915, Einstein introduced general relativity, sparking renewed interest in unifying gravity with electromagnetism. Though the strong and weak forces weren't yet known, Einstein was drawn to the idea that these two forces might be expressions of one deeper principle. He spent four decades searching for what became known as the unified field theory, a quest that set him apart from the mainstream physics community. During this time, he worked with collaborators like Nordström, Weyl, Eddington, Hilbert, Kaluza, and Klein. Einstein wrote in the early 1940s, "I have become a lonely old chap who is mainly known because he doesn't wear socks and who is exhibited as a curiosity on special occasions." He never succeeded in finding the theory he sought.
Late 20th century and the nuclear interactions
In the late 20th century, scientists searching for a theory of everything hit a wall when they discovered the strong and weak nuclear forces, which don't behave like gravity or electromagnetism. Einstein had hoped quantum mechanics would emerge from a deterministic unified theory, but that didn't happen. For years, gravity stayed outside the quantum framework, so physicists focused on the three forces described by quantum mechanics: electromagnetism, the weak force, and the strong force. In 1967–1968, Sheldon Glashow, Steven Weinberg, and Abdus Salam combined electromagnetism and the weak force into what's called the electroweak force. This unification shows a broken symmetry—electromagnetic and weak forces look different at low energies because the particles carrying the weak force have mass, while the photon does not. At higher energies, those differences disappear. Though the strong and electroweak forces coexist in the Standard Model, they remain separate, leaving the quest for a complete theory of everything unfinished.
Conventional sequence of theories
A theory of everything would unify all nature's basic forces—gravity, strong and weak nuclear interactions, and electromagnetism—with elementary particles. Scientific theory progression moves up levels: electroweak unification at 100 GeV, grand unification at 10¹⁶ GeV, then GUT force with gravity at Planck energy around 10¹⁹ GeV. Several Grand Unified Theories have been proposed, but simplest ones have been ruled out experimentally. Supersymmetric GUTs remain popular because they naturally produce dark matter and may connect to inflationary physics. However, these theories are incomplete—relying on renormalization and considered effective field theories, missing key phenomena at extremely high energies. The final step is reconciling quantum mechanics with gravity, which has not yet been achieved, though that's where a true theory of everything would need to arrive. Such a theory might also explain inflationary force, dark energy, and dark matter, though none have been definitively proven.
String theory and M-theory
Since the 1990s, physicists like Edward Witten have believed M-theory, describing 11-dimensional space and connecting to five string theories, might be the theory of everything, though broad agreement is lacking. A key string theory idea is that seven extra dimensions are needed for math to work, building on Kaluza–Klein theory which suggested a fifth dimension curled up so small we don't notice it. String theory introduces supersymmetry and extra dimensions to explain why gravity is much weaker than other forces. It has addressed quantum gravity questions like black hole entropy and produced breakthroughs through concepts like Gauge/String duality. But by the late 1990s, researchers realized there are possibly 10 to the 500th power different ways these extra dimensions could be shaped, creating the "string theory landscape." Some say this makes our universe's physical constants mere chance results rather than deep theory, leading critics to call it unscientific or philosophical, though others still see value in continuing the work.
Loop quantum gravity
Loop quantum gravity might help us build a theory of everything, but that’s not what it’s mainly about. It also suggests there's a smallest possible length scale. Recent work by Sundance Bilson-Thompson showed that the first generation of particles—leptons and quarks—can be modeled using braids of spacetime as building blocks, and these particles can survive quantum fluctuations. In this model, electric and color charge come from the topology of those braids. The original paper hinted that higher-generation particles might also be represented by more complex braiding, though no full construction was given. A 2008 paper by Bilson-Thompson, Hackett, Kauffman, and Smolin expanded the model to include infinite generations and weak force bosons, although photons and gluons were left out.
Present status
At present, there's no completed theory of everything that merges the Standard Model of particle physics with general relativity, nor one that allows calculation of values like the fine-structure constant or the electron’s mass. Many physicists believe that answers will come from ongoing experiments—such as those searching for new particles at large accelerators and the continued hunt for dark matter.
-
Read this chapter
Overview
When Everything starts up, it builds a list of every file and folder on NTFS and ReFS drives by checking their metadata, using the NTFS Master File Table for NTFS volumes. By default, it indexes all mounted drives, and keeps that list updated by watching for changes—on NTFS systems, it uses the change journal to track updates. Users can also add specific folders from other file systems to the index, though those will be slower to index. Searching works quickly through the index using either part of a filename or a regular expression, showing results as you type. Everything doesn't look inside files or index content, and on NTFS drives it only needs to update its index, so it uses minimal memory and processing power. The program also includes a command-line tool called "everything" that lets users access it from the console.
Security concerns
Everything is a search tool for Windows that needs special access to your computer's file system, which means it has to run with administrator rights. It can either be set up to work under an admin user account or as a Windows service. When it runs as a service, it lets other users search without giving them admin privileges. But here’s the key point: Everything doesn’t check whether a user should see certain files before showing results. So anyone who can use the search tool can see every filename on the drive, no matter what permissions they normally have.
Similar alternatives
Other search tools work the same way as Everything, directly tapping into the NTFS file system index to speed up searches. NTFS-Search, updated on July 5, 2017, and SwiftSearch, updated on July 6, 2019, are both open source options. There’s also UltraSearch from Jam Software, which offers a free version with limited features, while the full version is commercial software.
-
Read this chapter
Overview
Claude Mythos is Anthropic's most powerful Claude family line. The first model, Claude Mythos Preview, wasn't publicly released because it could find software flaws; instead, access was given to a few companies through Project Glasswing, which used the model to scan for security issues in important software. In June 2026, Anthropic launched Claude Fable 5, a general-use version with safety limits, plus a restricted version called Claude Mythos 5 with some safeguards removed. According to Anthropic, the two models are basically the same except for protections. If Fable 5's filters catch something related to cybersecurity, biology, chemistry, or model distillation, it gets passed to Claude Opus instead. In September 2026, Mythos 5.1 and Fable 5.1 were released. Industry estimates suggest that Mythos has around 8 trillion parameters, while Fable 5 has about 5 trillion.
Leak
A model called Claude Mythos became public knowledge on March 26, 2026, after leaked blog post drafts revealed its existence. Anthropic later confirmed the development of Mythos to Fortune magazine and said the model posed serious risks to cybersecurity. According to Axios, Anthropic had already warned government officials about Mythos's capabilities that same month.
Mythos Preview
Anthropic revealed its Claude Mythos cybersecurity model on April 7, 2026, saying it had no plans to share it publicly. Instead, the company launched Project Glasswing, giving access to over forty major tech firms including Microsoft, Apple, Google, and Amazon Web Services. That same day, unauthorized users reportedly accessed the system using information from the Mercor data breach. The NSA also used Mythos, even though the Department of Defense had blacklisted Anthropic after a disagreement. In June, access expanded to 150 organizations across more than fifteen countries. Anthropic said it expected to release “Mythos-class” models to all customers within weeks of announcing Claude Opus 4.8, pending further security improvements.
Mythos 5 and Fable 5
On June 9, Anthropic unveiled Claude Mythos 5 through Project Glasswing, releasing a variant called Fable 5 with extra safety measures. Three days later, the U.S. government contacted Anthropic, banning access to both models for anyone not a U.S. national, no matter their location, citing security risks. That order led Anthropic to cut off access for all users immediately. By June 26, some U.S. organizations were allowed to use Mythos again. Then, on June 30, Anthropic said the Department of Commerce had removed the restrictions, and access would return the following day. For a brief time, from July 1 to July 19, all subscribers got temporary access to Fable 5, extending an already-promotional window. On July 20, Fable 5 became part of Anthropic’s more expensive plans.
Specifications and capabilities
Claude Mythos is a large language model built to identify and fix software flaws, and it was tested by the UK AI Security Institute using a cyber range. In that test, Claude Mythos came out on top, beating out competitors like Claude Opus 4.6, GPT-5.4, and GPT-5.3 Codex. Industry reports estimate that Mythos has around 8 trillion parameters, while another model, Fable 5, has about 5 trillion. On July 19, 2026, Levent Alpöge, an Anthropic employee and mathematician, used Claude Fable 5 to solve a long-standing math problem known as the Jacobian conjecture in three-dimensional space—a challenge that had remained unsolved since 1939.
Reported vulnerabilities
Anthropic claimed Mythos found bugs in every major operating system and web browser, but an independent researcher questioned those claims, noting no outside verification. The researcher pointed out that a report from Anthropic admitted Claude Opus 4.6 actually discovered the bugs first before handing them to Mythos for exploitation. Some reported Firefox issues were not in Firefox itself, but in a test environment designed to mimic it with weaker security. When the two most exploitable bugs were removed, Mythos only succeeded in full code execution less than five percent of the time, though Anthropic said almost every successful run relied on those same two patched flaws. Another test showed that one of Mythos's headline vulnerabilities was also found by eight open-source models, including one with just 3.6 billion parameters and a cost of eleven cents per million tokens. Two weeks after its limited release, Mozilla announced it had fixed 271 security issues in Firefox using Mythos Preview. On May 14, 2026, Calif.io employees said they used Mythos to build an exploit targeting Apple M5 chips.
Financial response
Just hours after Anthropic revealed Mythos to the public, U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell brought together financial executives to discuss the risks. Several major banks, including JPMorgan Chase, Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley, began testing the model at the government’s direction. The Bank of Canada convened top lenders for a similar session the very next day. Mythos was also scheduled for review in upcoming meetings of the Bank of England’s Cross Market Operational Resilience Group and CMORG AI Taskforce. European Central Bank president Christine Lagarde praised Anthropic for limiting access to the system, but in response to European banks that hadn’t been granted entry, Mistral AI began developing its own model.
Governmental responses
In April 2026, as news of Claude Mythos spread, the U.S. Department of the Treasury reportedly sought access to the AI model. The White House and Anthropic met shortly after, and on May 13, 32 representatives urged the Office of the National Cyber Director to review federal cybersecurity policy. India’s finance minister held a meeting with banks and officials, while Japan’s Financial Services Agency formed a work-group in response. Canada’s artificial intelligence minister praised Anthropic for restricting access. Australia’s Prudential Regulation Authority also convened meetings with banks. Later that month, Anthropic said it had declined a request from a Chinese think tank at a Singapore conference, though the Chinese Embassy denied any government involvement.
Read
Free to download, keep and share. For general information only — not professional medical, legal or financial advice. Please consult a qualified professional.