Blog · Review Essay · Last reviewed June 24, 2026

Radical Technologies and the Operating System of Everyday Life

Adam Greenfield's Radical Technologies is a field guide to the moment when networked systems stop looking like devices and start behaving like an operating environment. Smartphones, sensors, augmented reality, blockchain, automation, machine learning, and artificial intelligence are not separate gadgets in his account. They are interlocking ways of reorganizing perception, labor, trust, movement, value, and institutional authority.

The useful definition is concrete: a technology is radical when it alters the background terms of ordinary action. It changes what can be sensed, logged, ranked, bought, refused, appealed, optimized, or forgotten. Its power is not novelty alone, but the way defaults, dependencies, and routes of recourse are rebuilt around it. That makes the book less a tour of inventions than a method for seeing how convenience becomes dependency.

The Book

Radical Technologies: The Design of Everyday Life first appeared from Verso in 2017. Verso's current listing presents the paperback at 368 pages and frames the book as a guide to networked objects, services, and spaces, including smartphones, blockchain, augmented-reality interfaces, virtual assistants, 3D printing, autonomous delivery drones, and self-driving cars. Open Library classifies it under ubiquitous computing, technological innovation, telematics, work design, and the social aspects of data processing.

Greenfield writes as a critic of urban computing and networked systems. LSE Cities describes his 2013-2014 research as work on the affective and experiential dimensions of everyday urban life under networked informatic systems, with explicit attention to whether ordinary people can understand systems that condition city life, from transit cards to predictive policing.

That background matters because the book is not gadget journalism. It is a political anatomy of interface culture. Greenfield asks what happens when technical systems arrive as conveniences but settle in as conditions: the phone that becomes a map, wallet, credential, workplace, memory prosthesis, camera, attention broker, and location beacon; the smart city that turns public space into an instrumented field; the automated system that turns discretionary judgment into a pipeline.

The word "radical" therefore should not be read as a synonym for dazzling. The radical feature of a connected thermostat, transit card, delivery app, biometric gate, predictive dashboard, or voice assistant is the way it moves decisions into an infrastructure that most people cannot inspect, bargain with, or leave without cost.

Current Context

As of June 24, 2026, the governance context around Greenfield's examples is more concrete than it was in 2017. Connected products are now being treated as lifecycle obligations, not only as consumer gadgets: NIST says its April 2026 IR 8259r1 update broadens manufacturer guidance across pre-market and post-market activity, including customer communication about maintenance, support, and end-of-life. The FCC's 2024 final rule created a voluntary U.S. Cyber Trust Mark program for wireless consumer IoT products. GOV.UK says the UK's consumer connectable product security regime has applied since April 29, 2024, with baseline duties around default passwords, vulnerability-reporting information, and minimum security-update periods.

The European Union is moving through a longer schedule. The European Commission says the Cyber Resilience Act entered into force on December 10, 2024, with reporting obligations applying from September 11, 2026 and the main obligations from December 11, 2027. For AI-generated content, the Commission's June 2026 transparency-code page says AI Act Article 50 transparency obligations apply from August 2, 2026 and address marking, detection, and labeling duties for certain generated or manipulated content. NIST's 2026 AI Agent Standards Initiative separately frames agents as systems capable of autonomous actions that need work on protocols, authentication, identity infrastructure, and security evaluation.

These instruments do not collapse into one rulebook. A voluntary label, a product-security statute, a phased EU regulation, a management-system standard, a transparency duty, and a NIST standards initiative give different kinds of leverage to different people. That distinction matters for Radical Technologies: the problem is not simply that everyday life is becoming technical. It is that support, updates, security, recourse, exit, and public accountability now sit inside layers most users never see.

ISO/IEC 42001 adds an organizational version of the same point for AI systems: governance has to be managed, maintained, and improved inside the institution using or providing AI-based products and services. Greenfield's book supplies the analytic habit this regulation-heavy moment still needs. Before asking whether a product is compliant, ask what ordinary action it now mediates, what dependency it creates, and who can still act when the interface, vendor, model, or network fails.

Everyday Life as Infrastructure

The book's strongest move is to begin from ordinary experience. A person checks a phone, accepts a route, scans a code, waits for a platform worker, speaks to an assistant, receives a recommendation, or lets a device remember what the person no longer has to know. None of this feels like a constitutional event. But the cumulative effect is constitutional in the small-c sense: it defines what actions are available, what forms of knowledge matter, and which institutions stand behind the interface.

That is the strongest governance insight. Everyday technology becomes infrastructure when refusal, repair, privacy, appeal, and fallback stop being personal preferences and become public conditions of participation. A phone is not only a phone if it is also a credential, payment rail, transit pass, work scheduler, school portal, emergency channel, and proof of presence. The question is no longer whether the device is useful. It is whether ordinary life is being routed through a layer that people can inspect, contest, and survive without losing basic access.

Greenfield is especially good on hidden dependency. A smooth interaction conceals satellites, app stores, cloud platforms, data centers, logistics networks, standards bodies, payment rails, mapping databases, labor regimes, and terms of service. The user experiences a gesture. The system enacts a political economy.

This is where the book connects to the site's recurring concern with recursive reality. Interfaces do not merely display the world. They train conduct, collect traces of that conduct, and then use those traces to redesign the next interface. A route suggestion shifts traffic. A ranking changes attention. A metric changes work. A moderation system changes speech. A procurement platform changes what public service can practically do.

That makes Radical Technologies a useful companion to The Stack, The Interface Effect, The Technological Society, and Tools for Conviviality. Each book refuses the innocent-tool story. Greenfield's distinctive contribution is close attention to the everyday surface where infrastructure becomes habit.

The Stack of Ordinary Dependence

The chapters accumulate like layers of a social stack. Smartphones network the self. The internet of things turns environments into fields of sensing and response. Augmented reality proposes an overlay in which the seen world is continuously annotated by private systems. Digital fabrication and automation reshape the politics of production and work. Cryptocurrency and blockchain promise trust without institutions while creating new forms of technical governance. Machine learning and AI make knowledge production probabilistic, opaque, and scalable.

The recurring question is not whether any single technology is useful. Many are useful. The question is who gets to define use. A delivery drone may solve a logistical problem while multiplying surveillance and labor displacement. A smart contract may reduce one kind of trust problem while making social context harder to interrupt. A predictive system may detect patterns while turning public authority into a vendor-mediated classification engine.

This is why the book still reads well after the first wave of blockchain and augmented-reality hype cooled. Some examples now feel period-specific, but the pattern survived. The commercial frontier keeps changing names: smart city, metaverse, Web3, generative AI, agentic commerce, spatial computing. The deeper motion is stable. Everyday life becomes a programmable environment, and power moves toward actors who own the programmable layer.

That layer is not only code. It includes identity, payments, device firmware, model weights, maps, labels, content policies, cloud accounts, APIs, procurement contracts, and error queues. It also includes negative space: the local skill no longer practiced, the paper path discontinued, the phone number replaced by a portal, the repair shop blocked by firmware, and the appeal route hidden behind a help center. A serious analysis of technology therefore has to follow dependencies across institutions, not stop at the device a user can hold.

The AI-Age Reading

Read in 2026, the book is most valuable as a prehistory of AI agents and model-mediated institutions.

Generative AI did not arrive on a blank social surface. It arrived inside phones, clouds, platforms, smart homes, workplace dashboards, payment systems, recommendation engines, identity systems, logistics software, surveillance cameras, and procurement channels. Radical Technologies helps explain why that matters. A model becomes powerful when it is connected to the operating surfaces of everyday life: calendars, inboxes, documents, support queues, stores, classrooms, benefits portals, hiring workflows, cars, cities, and homes.

This changes the AI question from "Can the system answer?" to "Where does the answer enter action?" A chatbot that only speaks is one thing. A model connected to identity, payment, work assignment, housing search, medical triage, law enforcement, education, or public benefits becomes part of the machinery that distributes options. The interface is no longer merely expressive. It is administrative.

The current governance context above sharpens the point without solving it. Rules and standards increasingly treat everyday devices, AI agents, synthetic media, and smart infrastructure as lifecycle, update, disclosure, security, and support problems. Greenfield's argument explains why that move is necessary: the political risk is not located only in the model or device, but in the ordinary dependency that forms when people cannot shop, work, travel, learn, appeal, repair, or opt out without passing through an opaque technical layer.

Greenfield's politics are useful here because he keeps asking what has been made unavailable. Automation is rarely only the substitution of a machine for a worker. It is also the redesign of a process so that work can be decomposed, monitored, routed, predicted, and eventually treated as software behavior. Machine learning is rarely only a technical method. It is a claim that enough traces can stand in for local knowledge.

The AI-era danger is recursive. Systems learn from instrumented environments. Institutions act through the systems. People adapt to the categories, prompts, ratings, and incentives. Their adaptation becomes new data. The resulting world then seems to prove that the system was reading reality correctly all along.

Governance and Safety

The practical governance lesson is to inventory dependency before adoption. An institution considering a connected device, AI assistant, sensor platform, predictive tool, or smart-city service should be able to name the data collected, the action the system can trigger, the vendor that controls the update path, the account or credential through which it acts, the humans who can override it, the people who can contest it, the records retained, and the exit plan if the vendor fails, changes terms, loses support, suffers a breach, or ends the product.

A useful artifact is a dependency register, not just a risk memo. For each everyday technology that becomes part of work, care, mobility, learning, housing, media, or public service, the register should state:

That register should sit behind three decision gates. First, a public-purpose gate: what nontechnical obligation does this system actually serve, and what less intrusive alternative was considered? Second, an authority gate: what can the system or agent change, under whose credential, with what approval, and with what rollback? Third, a continuity gate: what happens when the vendor changes terms, the model is updated, the cloud service fails, the device reaches end of support, or the affected person cannot use the interface?

Safety is not limited to whether a model gives a plausible answer. In everyday infrastructure, a bad classification can become a denied benefit, a misrouted patient, a locked account, a rejected applicant, a targeted worker, a false police lead, a manipulated feed, or an unpatched camera. The operational question is whether errors stay visible, reversible, bounded, and attributable.

Useful controls follow from that premise: data minimization; access controls; patch obligations for IoT; logging that supports audit without becoming surveillance by default; role-based limits for agents; human review where rights, benefits, safety, or employment are at stake; public notice when AI is used; provenance labels for generated public information; accessibility for non-digital alternatives; and procurement terms that require documentation, incident reporting, evaluation evidence, portability, deletion on termination, and a tested fallback when the interface is unavailable.

The safety case should also distinguish reversible interface errors from consequential control errors. A bad recommendation can be corrected in place. A bad instruction sent to a door lock, dispatch queue, benefits workflow, hiring screen, payment rail, workplace scheduler, or connected medical process may already have changed someone's options before a review begins. In Greenfield's terms, the radical moment is when the interface stops advising and starts conditioning the world in which advice is received.

This is a more demanding standard than "innovation with ethics." It asks whether the institution can still govern the system after the demo is over. Greenfield's book supplies the first half of that discipline: look beneath the smooth surface. Contemporary AI and IoT governance supplies the second half: write down the obligations, test the claims, preserve appeal, and keep public responsibility from being outsourced into an interface.

Where the Book Needs Friction

Radical Technologies is sharpest as critique, not as institutional design. Its skepticism can sometimes make the technological future feel more unified than it is. The politics of a union-negotiated scheduling system, a public-interest sensor network, a proprietary smart-home platform, an autonomous weapon, and a classroom tutor are not the same simply because all are computational.

The book also inherits some of the urgency of its 2017 moment. Blockchain receives more serious attention than many readers would give it now, while present-day foundation models, synthetic media, and agentic workflows have changed the center of gravity. That is not a failure of the book. It is a reminder that analysis of "the next thing" ages fastest when it is tied to product cycles.

The correction is to separate family resemblance from institutional sameness. A public transit sensor, a military targeting system, a care-work scheduling app, and a home assistant may all depend on networked computation, but they differ in coercion, appeal rights, domain expertise, accountability, and harm. Strong analysis has to preserve those differences while still noticing the shared movement toward programmable environments.

What lasts is the method: inspect the surface, follow the dependencies, ask who governs the system, ask who can refuse it, ask what labor disappeared from view, ask what local knowledge has been compressed, and ask whether convenience has quietly become compulsion.

What This Changes

The deepest lesson of Radical Technologies is that everyday interfaces are belief-forming institutions.

A phone teaches what counts as reachable. A map teaches what counts as nearby. A feed teaches what counts as socially real. A dashboard teaches what counts as work. A rating teaches what counts as trust. A model answer teaches what counts as knowledge. These are not only user experiences. They are small, repeated lessons in how to inhabit a world built by someone else's abstractions.

The practical response is not nostalgia for a pre-digital life. It is interface due diligence. Before adopting a system, an institution should be able to answer: What dependency does this create? What human capacity does it weaken? What public or local alternative remains? What data does it require? What appeal exists when the system is wrong? Who can audit the categories? What happens if the vendor leaves, the model changes, or the interface becomes mandatory?

Greenfield's book belongs in this catalog because it names the ordinary path by which technological politics enters life: not through a single dramatic machine takeover, but through a thousand helpful surfaces that gradually decide what reality is easiest to see, say, buy, trust, optimize, and obey.

Source Discipline

This review separates three kinds of evidence. Book metadata comes from Verso, Open Library, Penguin Random House, and LSE Cities. Reception context comes from contemporary reviews in The Guardian and Technology|Architecture + Design. Current governance claims are tied to primary or institutional sources from NIST, ISO, the EU AI Act, the EU Cyber Resilience Act, the FCC, and GOV.UK.

Where publisher and distributor metadata differ, this review relies on Verso and Open Library for page count, publication dates, and edition identifiers. Legal timing is kept separate from legal text: the Article 50 transparency duties are described with the Article 113 application date rather than treated as already fully applicable on June 24, 2026. Likewise, a voluntary label, a standards document, a statutory duty, and a phased EU regulation are not interchangeable forms of governance.

The interpretation is deliberately institutional rather than mystical. It does not claim that present AI systems possess minds. The issue is more concrete: software connected to identity, money, public services, work, housing, media, and urban systems can change what institutions see and what people are able to do.

Sources

Book links are paid affiliate links. As an Amazon Associate I earn from qualifying purchases.

Buy on Amazon Browse Books

Return to Blog · Return to Books