A conversation on the subject of this page starts with a message to the scientific director of the institute.
Write to the scientific directorThe sixth technological order: what changes when programs begin to act
Economists call a technological order (a techno-economic paradigm) the set of technologies of one level that shapes production, trade and management for several decades. Five such orders have followed one another since the late eighteenth century; the sixth is taking shape now, and its defining trait is already visible in 2026: a program no longer only computes and advises, it acts: it makes payments, edits code and changes records in databases. This page explains where the notion comes from, how the new order differs from the previous one and what it means for a company, a state and a person.
Forecasts on this page are marked by the discipline of truth: each names a branch and an external benchmark by which it can be checked. Figures carry a date and a source.
What a technological order is
The notion grew out of an observation by Nikolai Kondratiev: in “Long waves of the business cycle” (1925) he showed that an economy develops in long waves of forty to sixty years, each tied to a renewal of technology. Sergei Glazyev, in “The theory of long-term techno-economic development” (1993), named the carrier of such a wave a technological order: interlinked industries of one technical level, from the core of key technologies to the branches that serve them. Carlota Perez, in “Technological Revolutions and Financial Capital” (2002), described the same course as a succession of techno-economic paradigms and showed the role financial capital plays in each: first a bubble around the new technology, then its quiet deployment.
| Order | Core technologies | What changed in the economy |
|---|---|---|
| First, from the 1770s | Textile machinery, water power | The factory instead of the workshop |
| Second, from the 1830s | The steam engine, railways, coal | National markets joined by rail |
| Third, from the 1880s | Steel, electricity, heavy engineering | The large corporation and mass production |
| Fourth, from the 1930s | Oil, the automobile, organic chemistry | The assembly line and mass consumption |
| Fifth, from the 1970s | Microelectronics, computing, communication networks | Information as a commodity, platforms and global supply chains |
| Sixth, taking shape | In Glazyev’s periodization, nano-, bio-, information and cognitive technologies; in the institute’s observation, programs that act on their own | Execution moves to machines; value moves to rules, checking and responsibility |
Authors place the boundaries of the orders a decade apart; the table gives the accepted sequence rather than exact dates. The argument over which technologies form the core of the sixth order is not settled, and this page does not try to settle it.
Why the trait of the new order is a program that acts
Three observations from 2026, each with a source, show that what has changed is not one product but the way work is done.
- Computation became an order of magnitude cheaper per year. The price of a call to a language model at a fixed quality of answer falls roughly tenfold a year (estimates by a16z of 12 November 2024 and by Epoch AI of 12 March 2025). The horizon of tasks the best model completes without a person has grown to about twelve hours, as measured by Model Evaluation and Threat Research (METR) on 20 February 2026. The details, with benchmarks to 2035, are on the page Horizon 2027–2035.
- A program received write access. Agents, programs built on neural networks that call tools themselves, make payments, edit code, delete and change data. The cases of 2025–2026 in which such access ended in the loss of a production database are analyzed from primary sources on the page Incidents. The first live payment in Europe executed by an agent was carried out by Santander and Mastercard on 2 March 2026 (The agent economy).
- Scarcity of attention stops being the intermediary’s footing. The fifth order built its platforms on the fact that a buyer cannot compare thousands of suppliers and goes to one storefront. The buyer’s own agent compares them in seconds and closes the deal by the rules it was given. Hence the institute’s position, marked as a forecast of branch B (the trend continues): the platform remains the place of first contact, while the transaction moves to direct protocols between buyer and seller under the protection of action checks and receipts. The benchmark is the share of autonomous agent payments in turnover, which the register of forecasts tracks (in Russian).
The general law here is W. Ross Ashby’s law of requisite variety: a controller must be no simpler than the environment it controls. A platform that simplifies its environment for the sake of control, by blocking direct settlements and keeping participants inside white lists, loses to external complexity as soon as the participants have an executor of their own. Why this is so is set out on the page Ashby’s law and Beer’s model.
What it means
| For whom | What changes | What stays valuable |
|---|---|---|
| A company | Execution, such as correspondence, documents, payments and code, moves to programs; platforms and intermediaries stop being the only way to the customer. | A perimeter of its own: the customer base, memory, keys and the rules of transactions stay with the company, not with a platform; every action of a program runs under the organization’s rule. |
| A state | Rules written for a person and for a deterministic program meet a third participant: a program that acts with granted rights. The law on artificial intelligence, Order No. 117 and the requirements for critical information infrastructure already answer this (section Rules and liability). | Provability: an action that leaves a verifiable trace, and responsibility that stays with the organization. |
| A person | Hiring shows it first: employment of specialists aged 22 to 25 in occupations exposed to artificial intelligence fell by 19 percent (Stanford Digital Economy Lab, 12 August 2026). | Setting the task, signing the rule, judging the result and bearing responsibility, which a program is not given. More on the page The future with agents. |
What to do now
- Separate the storefront from the core: platforms for first contact, while the customer base, memory, keys and the rules of transactions stay in a perimeter of one’s own.
- Understand what a software agent is and how it differs from a chatbot: the page What an agent is and the eleven-step learning path.
- Write the rules for programs’ actions so that they can be checked before execution, and keep a log of decisions that cannot be rewritten after the fact; how this is done is shown on the page Formal verification.
- Check forecasts against benchmarks once a year instead of trusting one expert: Truth marking.
Where to go next
For twenty minutes: The agent economy and Horizon 2027–2035. For hours: The doctrine of a managed transition, The program and the book Proven Action, where the map of 2027–2035 is laid out along five axes. If the task is practical: Custom automation.
Autonomous Non-Profit Organization (ANO) “Research Institute of System Synthesis”, edition of 26 October 2026. Forecasts are marked with a branch and a benchmark; primary sources are named in the text.