Fourteen Journeys: How Value Chains Actually Evolve

Knowledge seems to have made a neat journey: from what a person remembered, to books and libraries, to search, to a machine that can answer back. Music can be made to look similar: live performance, records, radio, streaming, generation. Put the two beside each other and it is tempting to name three stages, give each one a colour, and send every other industry through them.

That temptation produces a clean table and a false history.

A map is not a song. A meal cannot be copied like a photograph. A robotaxi changes who acts, not merely how a ride is found. Wikipedia changed the production of an encyclopedia years before answer engines changed its presentation. In film, the evolution of watching and the evolution of making have separated into different branches. The more industries we add, the less honest a universal sequence becomes.

So this essay has a narrower purpose: to show the evolution itself. Fourteen industries, fourteen routes. The dates mark broad transitions rather than birthdays. Old forms do not necessarily disappear when a new one arrives; often they survive because they serve a different part of the human want.

The human starting point

Before following the chains, we need one proposition about the creature using them:

Humans seek effective influence over their internal state and external world. They prefer paths that reduce unwanted effort, delay, and uncertainty—but they value effort that produces agency, competence, identity, connection, or meaning.

That is an observation, not a destination every industry must approach. It explains why a person may want a machine to plan a commute but spend ten years learning the piano. The effort paradox is precisely that effort is both avoided as a cost and sought as a source of value. Effort can stand between us and what we want, or the effort can be part of what we want.

Music makes the distinction unusually visible because “listening to music” bundles together several human responses:

• Moving with it. Rhythm, bass and syncopation can solicit bodily movement. Work on groove and syncopation suggests that the pleasure is partly in the invitation to act, not only in receiving a sound.
• Watching it be made. The pianist’s hands, the orchestra coordinating, and the DJ reading a room can be compelling in something like the way skilled sport is compelling. Action-observation research finds that watching skilled movement engages some of the observer’s own motor system, especially when the skill is familiar.
• Listening privately. Music also regulates mood and arousal, helps construct identity, and creates a sense of social connection. These were the three broad functions that survived an empirical reduction of the many reasons people give for listening.

These wants may be related, or partly independent. We do not need to collapse them. The live performance could satisfy all three at once; the recording strongly served private listening; streaming made access to recorded music abundant; generative tools now alter making and editing. None is simply “the next version” of all the others.

That is how the atlas below should be read. A line records a succession of important forms. It does not claim that the newer form contains or replaces everything before it. The green final node is the frontier visible in July 2026, not a prediction of an inevitable destination.

The evolutionary atlas

How each chain reached its present frontier

Knowledge: from answers to delegated inquiry

The important breaks in knowledge were changes in where authority lived and what work the seeker still performed. Oral knowledge lived in people. Libraries gathered external memory. Print multiplied it. Search indexed other people’s pages but left the reader to compare them and build an answer. Conversational systems began returning the answer itself.

The 2026 frontier is a further change: a research agent can search, choose sources, synthesize them, use connected tools and revise its path while working. OpenAI’s Deep Research now supports connected sources, trusted-site restrictions and mid-course refinement. This is not simply conversation made faster. It moves part of inquiry from the user to the system. The unresolved history now concerns provenance: who checks the synthesis, whose knowledge was admitted, and what source deserves trust?

Music: the branches never collapsed

Recording separated sound from the body producing it. Radio separated access from possession. Streaming made an enormous recorded catalogue available on demand. Yet live performance, private listening, dancing and learning an instrument continued because they were not interchangeable uses of music.

Generation introduced another branch: first a finished song produced from a wish, then an increasingly editable production environment. By 2026 Suno’s tools include song editing, multitrack work, stem extraction and finer control rather than only rerolling a prompt; its release history makes that shift visible. At the same time, licensing and artist consent are becoming part of the technical form, as in the Warner Music Group–Suno agreement. Music therefore does not stand at one “stage.” Streaming, performance, instrumental mastery, generative composition and editable production coexist.

Maps: from representation to continuous guidance

A paper map represented a territory and left interpretation to the traveler. Route websites calculated a journey but still produced something like a printable map. Smartphone navigation moved the representation into the trip: location, traffic and missed turns could continuously change the instruction.

The latest extension is conversational and contextual. Google’s Ask Maps and Immersive Navigation combine map data with complex questions and real-time conditions. The chain has evolved from an object, to a query, to a service accompanying movement. The roads, businesses, public data and sensor networks beneath it remain a separately governed supply system; moving through a route is not the same as producing the world being mapped.

Software: from reusable code to delegated changes

Software evolved along several linked paths: programs became products; reusable libraries reduced repeated work; open source distributed authorship; software-as-a-service changed delivery; GitHub made the repository a shared operational place. Coding assistance initially suggested the next line inside an editor.

Repository agents now accept a task, inspect a larger codebase, plan changes, modify files, run checks and prepare work for review. GitHub Copilot CLI describes modes ranging from approval of individual actions to more autonomous execution. The frontier is therefore delegated engineering under tests and governance—not merely “say what you want and software appears.” Specification and verification have become more central as typing becomes less central.

Gaming: participation arrived before generation

Gaming’s most important transition was not from physical ownership to a digital store. It was from a finished authored game to a maintained social system. Online multiplayer, live-service updates, modding, Minecraft and Roblox all weakened the boundary between a game’s producer and its community, though in different ways. Some players only play; others build, host, modify or trade.

In 2026, AI mainly lowers the cost of work inside the production environment. Unity’s AI tools expose ask, plan and agent workflows in the editor, while its 2026 industry report describes cautious adoption concentrated in coding and other production work. Fully generated, persistent and coherent worlds are not the ordinary product. Moderation, discovery, interactive consistency and platform economics still shape what can survive.

Retail: intelligence moves ahead of the warehouse

Mail order enlarged the shop without moving it. E-commerce moved the catalogue and transaction online. Marketplaces gathered many sellers, while merchant platforms let more sellers operate their own storefronts. Recommendation systems then reorganized discovery around predicted preference.

The 2026 frontier is an agent acting across discovery, comparison, cart and checkout. Google’s Universal Commerce Protocol is an attempt to standardize those actions across agents and merchants; Amazon’s Alexa for Shopping can monitor prices, add items and automate some purchases. But this evolution is concentrated above the physical chain. Manufacturing, inventory, picking, delivery, returns and support remain, and the struggle is increasingly over which merchant or agent controls the buyer’s intent.

Video and film: watching and making split apart

The audience path ran from cinema to broadcast, home video, streaming libraries and recommendation feeds. The production path ran from expensive cameras and studios to digital editing, cheap distribution and creator platforms. YouTube and TikTok transformed who could publish without making the viewer’s activity equivalent to production.

Generative video is now changing the production branch again. Runway’s Agent moves from a conversational direction to multi-shot video while retaining a timeline for intervention; Adobe’s 2026 Firefly and Premiere updates combine generation with editing workflows. Meanwhile audiences still mostly select or receive completed sequences. Any history that gives video one line obscures this split.

Photography: capture and synthesis now coexist

Roll film lowered the expertise needed to capture a photograph. Instant film shortened the wait. Digital cameras removed film and chemical development. Phones joined capture to computation, editing and distribution. Stock-photo search evolved alongside this route but was never the successor to taking a photograph; it served a different need.

Synthetic images created another branch, initially dominated by prompting. The 2026 movement is toward exact, reversible control: marking regions, preserving chosen elements, refining and transforming rather than accepting a sealed result. Adobe’s Precision Flow and AI Markup illustrate that direction. Photography now includes evidence-producing capture, computational interpretation, synthetic construction and hybrid editing, making authenticity and provenance part of the chain itself.

Personal transport: availability and operation diverged

Private car ownership joined access to a vehicle with responsibility for driving, storing and maintaining it. Taxis separated ownership from the journey. Ride-hailing improved dispatch and availability without changing the basic operating model: a human still drove each trip.

Robotaxis change operation. Waymo reported more than 20 million fully autonomous trips and service expansion across more than eleven cities in its May 2026 update. Yet this is not the replacement of all transport by one new form. Private cars, public transport, cycling, taxis, ride-hailing and geographically bounded autonomous fleets coexist. The frontier is a capital-intensive, regulated fleet service whose expansion depends on safety evidence and local operational coverage.

Encyclopedias: a human source beneath machine answers

Printed encyclopedias concentrated expert editing into a durable household reference. CD-ROM reduced the physical bulk and made search easier. Wikipedia then made a deeper production change: a continuously revised commons written and governed by volunteers. Search engines often presented fragments of that commons without changing how the source was made.

AI answer systems now add another presentation layer above it. Wikipedia’s own evolution is different: Wikimedia’s AI strategy uses AI to assist translation, moderation and routine work while preserving human editorial judgment. The present chain therefore has a human-curated source, automated maintenance assistance, and machine-generated answer interfaces—three layers with different incentives and different standards of accountability.

Food: planning digitized; matter did not

Food never followed a single line from cooking to delivery. Households, restaurants, industrial producers, supermarkets, takeout, meal kits and delivery platforms serve different combinations of nourishment, convenience, craft and social ritual. Recipes became searchable and shareable information, but a recipe is not a meal. Every meal still consumes ingredients, time, equipment and physical work somewhere in the chain.

The latest change compresses planning and procurement. Instacart’s 2026 assistant can turn preferences, photographs or ideas into a meal plan and ready-to-buy cart. It does not eliminate substitution, perishability, picking, delivery or cooking. The evolution has reached from intent into the grocery basket; it has not converted food into an informational good.

Money and payments: coexistence rather than succession

Cash, deposits, cards, bank transfers, wallets and instant-payment systems did not simply replace one another. They altered different layers: the claim people hold, the interface they use, the message sent, and the mechanism by which institutions settle. Cryptoassets and stablecoins introduced additional issuers and rails rather than one universally accepted next form of money.

In 2026, tokenized deposits, stablecoins, programmable settlement and agent-initiated payments are developing alongside conventional systems. The BIS Annual Economic Report 2026 argues that stablecoins reveal some potential of programmable payments while failing important properties expected of money, and places tokenization within a trusted two-tier monetary system. The evolution here is institutional competition over trust, par value, interoperability and regulation—not a consumer journey from owning money to participating in it.

Travel and lodging: planning advances faster than accountability

Travel agencies assembled fragmented information and booked through reservation networks. Online travel agencies exposed comparison and booking to the traveler. Peer marketplaces added new lodging supply, but also new platform governance. Search, booking, payment and help during disruption remained distributed across different actors.

AI is changing inspiration, comparison and itinerary construction first. Expedia’s company-commissioned 2026 survey found far greater comfort using AI for planning than for booking. Airbnb’s 2026 release similarly expands services and AI-supported comparison around its marketplace. The present frontier is therefore an AI planning and support layer attached to platforms that remain accountable for inventory, payment, refunds and failures.

Virtual places: generation before persistence

Virtual places began as text spaces whose worlds were maintained in language and social convention. Graphical multiplayer worlds added continuous visual space. Second Life, Minecraft and Roblox developed different forms of persistence and user construction; social VR added embodiment and presence. These were not merely better maps because the place was produced by software and social activity rather than surveyed from an external territory.

World models now generate interactive environments in response to user action. DeepMind’s Genie 3 can produce navigable environments in real time, but only for limited durations and action spaces. Persistent multi-user consistency, durable ownership, interoperability and creator rights remain unresolved. The 2026 frontier is an early generated simulation—not yet a stable world in which a society can continue living after the prompt ends.

A living atlas

The point of setting these histories beside one another is not to make them end in the same place. It is to see each one accurately enough that its next change can be recognized without forcing it into somebody else’s vocabulary.

Knowledge is moving from answer retrieval toward delegated inquiry. Maps are becoming continuous contextual guidance. Software is delegating repository work. Retail and food are automating parts of intent and procurement while their physical operations persist. Music, video and photography have acquired generative production branches. Gaming already had a participatory branch before those tools arrived. Transport is automating operation. Encyclopedias retain a human source beneath AI presentation. Money remains competing institutions and rails. Travel separates convenient planning from trusted booking. Virtual places can now be generated before they can reliably persist.

Those are the evolutions as they stand in July 2026. The last node of each line should move when the industry moves; the lines do not owe us a shared destination.