From speech to AI agents, fifteen technologies follow the same sequence: a difficulty, the people who gained, the way money flowed, and the machinery and operators behind it. Read the sequence once and you can read the next one.
The table reads left to right like a story. Something was hard. Someone gained. A way to pay appeared. Infrastructure and devices made it real. Operators ran it.
Pick an era. Every term that has a DBpedia ID opens its entry in a new tab.
The patterns below are read from the table. Groupings and simplifications are ours, and marked as such.
Early technologies fought memory, copying and muscle. Mid-era technologies fought distance and audience size. The latest fight scarce expertise and the cost of coordinating action. The brief moves from matter to judgment.
Our grouping of constraints into five families.
Patronage paid the scribes. Sales paid the engine makers. Fees, advertising, commissions and metering followed. Agents may charge for outcomes. Each step prices something closer to what the customer actually wanted.
Our simplification: where each era's revenue is first anchored.
The telegraph needed power and lines. The internet needed telecoms. Generative AI needs data centers and electricity. Agents add identity and authorized access on top. A new layer arrives faster when the layers beneath it are already in place.
Our reading of the infrastructure column.
Scribes and temples held the pen. Printers held the press. Telegraph and telephone companies held the wire, broadcasters the airwaves. Platforms and app stores held the audience. Model providers and clouds hold the compute. Agent runtimes and data-space operators are the next gatekeepers.
Read left to right, from the operators column.
Every earlier difficulty was about doing something. The agent row is about letting it be done: integration, permission, reliability, accountability. The infrastructure it lists, identity and authorized access, is the product. The next advantage is likely to sit with whoever makes those checkable.
Interpretation of the AI agents row.
Ask them in order. The sixth answer is the next brief.
Each question is an entity in the knowledge graph. Select a question to open it there; use the arrow to read the answer.
Each technology removed one named constraint, and beneficiaries, a business model, infrastructure, devices and operators grew around it. The 15 rows run from spoken language to AI agents.
Constraint, beneficiaries, business model, infrastructure, hardware and operators. The table reads left to right in that order, and the sequence repeats in every era.
Equal distances show equal ratios of time before 2026, not equal numbers of years. A linear axis would squeeze the last 150 years, where most landmarks sit, into a sliver.
Television is placed at about 1935 for its 1930s regular service. Spoken language is undated and sits at the left edge. Every other dot uses one landmark year taken from the table.
Early technologies fought memory, copying and muscle. Mid-era ones fought distance and audience size. The latest fight scarce expertise and the cost of coordinating action. The five-family grouping is an editorial reading.
From patronage and institutional funding, through sales and leases, usage fees, advertising and licensing, platform commissions and metered use, toward possible outcome-based fees for agents. This simplifies where each era's revenue is first anchored.
The table lists power or electricity as an input for the telegraph, telephone, radio, television, computer and generative AI.
Cloud, data centers, power, networking, and training and inference data, running on GPUs, CPUs, servers and accelerators.
Operators move from scribes, schools and temples, through printers, industrial firms, network operators and broadcasters, to platforms, carriers, cloud and model providers, and now agent runtimes and data-space operators.
Their difficulty is not doing something but letting it be done: integration, permission, reliability and accountability. Their infrastructure adds identity and authorized access.
The table gives c. 3300-3200 BCE for the earliest known writing systems and cites the British Museum.
Linked terms use DBpedia resource IDs, each checked against the live DBpedia endpoint. API, gig economy, telecommunications company and web search engine were left unlinked because their DBpedia entries are redirects or disambiguation pages.
The five implications, the five constraint families and the revenue, infrastructure and operator groupings are editorial readings of the table, not claims from its sources.
Ask six questions in order: what is the constraint, who feels it most, who pays and for what unit, what must already exist beneath it, who will run it, and what new constraint it exposes.
Terms as used in this analysis. Each opens its entity in the knowledge graph.
The source of every claim above. Linked terms use DBpedia IDs.
| Time period | Technology | Difficulty | Beneficiaries | Business model | Infrastructure / inputs | Hardware / devices | Operators / enablers |
|---|---|---|---|---|---|---|---|
| Prehistory; origin undated | Spoken language | Reach ends where voice and memory end. | Communities, teaching, culture, exchange | None distinct; speech carries teaching, performance and trade. | Shared languages, customs, meeting places | Voice and ears | Families, teachers, storytellers, leaders |
| c. 3300–3200 BCE (British Museum) | Writing | Records are hard to keep, verify and carry. | Government, law, trade, schools, religion | Scribes and record-keeping, paid by institutions | Scripts, literacy, clay, papyrus, parchment, ink | Styluses, tablets, scrolls | Scribes, schools, courts, temples, officials |
| Manuscripts from the first writing; codex from the 1st century CE (Smithsonian) | Manuscripts and codices | Hand copying is slow and costly, so copies are few. | Scholars, clergy, lawyers, schools, rulers | Commissioned copies, patronage, libraries | Writing materials, skilled copyists, libraries | Scrolls, codices, pens | Scribes, monasteries, courts, universities, libraries |
| East Asia, 7th–8th centuries; Europe, c. 1450s (Library of Congress) | Printing | Hand copying limits speed, volume and consistency. | Schools, religion, science, commerce, public debate | Selling books and pamphlets | Presses, type, paper, ink, distribution | Presses, typesetting gear | Printers, publishers, booksellers, authors, distributors |
| Newcomen engine, 1712; Watt's condenser patent, 1769 (Science Museum) | Steam engine | Power is fixed in place and limited by muscle. | Mining, manufacturing, trade, travel | Engine sales and leases; fares and freight | Coal, iron, steel, machine shops, railways, ports | Stationary engines, locomotives, steamships | Engine makers, railroads, shipping lines, repairers |
| 1830s; first public Morse message, 1844 | Telegraphy | News moves no faster than its carrier. | Trade, government, press, railways, finance | Message fees, equipment, network services | Lines, poles, stations, power, trained operators | Keys, relays, receivers | Telegraph companies, news agencies, railways |
| 1860s; four-stroke engine, 1876 (National Diet Library of Japan) | Internal combustion engine | Travel and freight are bound to fixed routes and timetables. | Retail, logistics, leisure, commuting | Vehicle sales and finance; fuel, repair, insurance, dealers | Steel, oil, fuel, roads, parts | Cars, trucks, motorcycles | Automakers, garages, fuel suppliers, insurers, dealers |
| First call, 1876 | Telephone | No live voice at a distance. | Business, trade, homes, public services | Subscriptions, call charges, equipment rental | Cables, exchanges, power, rights of way | Handsets, exchanges | Telephone companies, network operators |
| Wireless telegraphy, 1890s; regular broadcasting, 1920 (Computer History Museum) | Radio | Sound cannot reach a mass audience at once. | News, entertainment, education, public information | Advertising, public funding, licences, sponsorship | Spectrum, transmitters, antennas, studios, power | Transmitters, receivers | Broadcasters, networks, regulators, advertisers |
| Experiments, late 1920s; regular service, 1930s; mass adoption after WWII | Television | Moving pictures and sound cannot reach mass audiences. | News, entertainment, education, advertising | Advertising, subscriptions, carriage fees, licensing | Towers, spectrum, satellites, cable, power | Sets, cameras, studio gear | Broadcasters, networks, cable and satellite operators |
| Electronic computing from the 1940s; ENIAC unveiled, 1946 | Computer | Processing and storing data is costly, specialist and hard to reach. | Manufacturing, finance, aerospace, government, research | Hardware sales and leases, software licences, maintenance, services | Semiconductors, power, operating systems, software | Mainframes, servers, PCs | IBM and other vendors, integrators, service firms |
| First ARPANET hosts, 1969; public and commercial growth, late 1980s–1990s (Internet Society) | Internet | Networks stand alone; sharing information is costly. | Search, e-commerce, publishing, education, communication, cloud | Access fees, advertising, transactions, cloud usage | Telecoms, routers, data centers, protocols | Routers, servers, PCs, modems | Internet providers, backbone and hosting operators, web platforms |
| Mobile web, 1990s; i-mode, 1999; smartphone boom from 2007 | Mobile internet | Access is tied to a fixed place. | Social media, navigation, messaging, gig work | Carrier plans, app sales, advertising, commissions, gig fees | Cell towers, spectrum, backhaul, chips | Smartphones, network gear | Carriers, app stores, iOS, Android |
| Transformer, 2017; public uptake of generative AI, 2022 (original paper) | Generative AI, including large language models | Writing, code and media once demanded scarce expertise and time. | Writing, programming, design, media, customer support | Metered API use, subscriptions, enterprise licences, hosted customization | Cloud, data centers, power, networking, training and inference data | GPUs, CPUs, servers, accelerators | Cloud providers, model developers, AI platforms |
| Long research history; LLM-enabled agents grew sharply in 2023 (2023 survey) | AI agents | Acting across systems raises problems of integration, permission, reliability and accountability. | Workflow automation, private or public data spaces, agentic commerce in retail, leisure and services | Subscriptions, usage-based orchestration, managed services, transaction commissions, possibly outcome-based fees | Cloud and edge computing, virtual machines, identity, authorized access, tools, workflows | GPUs, CPUs, servers, edge devices | Agent runtimes, model providers, workflow orchestration, data-space operators |
Dated claims follow the sources cited in the table: British Museum, Smithsonian, Library of Congress, Science Museum, National Diet Library of Japan, Computer History Museum, Internet Society, the original Transformer paper and a 2023 survey of LLM-enabled agents.