Intelligence is becoming infrastructure.
The first revolution stopped being a software story the moment its constraint moved to power, silicon and land. It is now underwritten on the timelines of heavy industry.
Two revolutions.One capital cycle.
Artificial intelligence is giving machines intelligence. Autonomous systems are giving them physical agency. The Dual Revolution researches what happens when those two systems begin to share one infrastructure stack.
- Systems modelled
- 20
- Dependencies mapped
- 34
- Capital layers
- 04
- Sectors framed
- 08
01 / Thesis
The standing position
Four statements. Everything else on this page exists to support, qualify or test them — and to make the reasoning visible enough to disagree with.
Autonomy moves the thesis into the physical economy.
Machines that reason are interesting. Machines that reason and act are an input cost. The second revolution is what converts capability into a line in someone's operating budget.
The two systems are entering one dependency chain.
Perception runs on the architectures built for language. Edge inference draws on the supply chain built for training. Both terminate in the same power constraint.
One dependency chain is one capital cycle.
That is the whole position. Everything published here is an attempt to establish where it holds, where it breaks, and what has already been priced.
Dependency first
Every position is traced to the constraint underneath it before it is held.
Qualitative where quantity would be false
Ordinal judgement is stated as ordinal judgement. Nothing here is measured.
Written to be disagreed with
Each framework states what would have to be true, and what would falsify it.
02 / Signals
Two systems, examined separately
The first revolution gives machines intelligence. The second gives them agency in the physical world. They developed independently for most of their history — and no longer do.
Signal A
Machine Intelligence
What machines can reason about
Capability is compounding against compute, data and capital rather than against algorithmic novelty. The first revolution is an infrastructure story wearing software clothing.
- Binding constraint
- Cost per useful task
Research observation
Select a system to read the desk's observation on it. The selection is carried through the rest of the document.
Signal B
Machine Autonomy
What machines can physically do
Robotics failed for forty years on perception and generalization, not on actuation. Both failures are being attacked by the first revolution rather than by mechanical engineering.
- Binding constraint
- Generalization outside structure
Research observation
Select a system to read the desk's observation on it. The selection is carried through the rest of the document.
03 / Convergence
The dependency model
A thesis framework, not a measurement. Every relationship below is a stated position about how the stack fits together. Select a system to trace what it rests on and what rests on it.
01Foundational Infrastructure
02Intelligence
03Physical Autonomy
04Economic Deployment
Select a system
Selecting a system dims everything it does not touch and traces its dependency path in both directions — what it rests on, and what rests on it.
- Rests on
- —
- Carries
- —
- Binding constraint
- —
04 / Capital
Four layers, in order
Each layer is underwritten on a different duration. The mismatch between them — fifteen-year assets carrying eighteen-month products — is where most of the argument lives.
The slowest layer in the system, and therefore the one that sets the ceiling for everything above it. Intelligence is now provisioned the way heavy industry is provisioned: by interconnect queue, by transformer lead time, by long-dated power contract.
- Binding constraint
- Lead time
- Asset duration
- 5–20 years
- Systems
- 05
Capability concentrates where compute, data and capital concentrate. The question worth researching is not who can build a frontier model but who is structurally able to keep building one, and what the rest of the stack does in response.
- Binding constraint
- Cost per task
- Asset duration
- 1–5 years
- Systems
- 05
This is where the first revolution crosses into the physical economy. Autonomy inherits the cost curve of semiconductors rather than the cost curve of industrial equipment, which is what makes the timing question different this cycle.
- Binding constraint
- Generalization
- Asset duration
- 3–10 years
- Systems
- 05
Adoption is gated by integration, liability and process redesign rather than by capability. The lag between what is demonstrable and what is deployed is the part of the thesis with the clearest capital consequence.
- Binding constraint
- Integration
- Asset duration
- 2–15 years
- Systems
- 05
05 / Matrix
Sector exposure framework
Ordinal judgement across eight sectors. Populated qualitatively on purpose: nothing here is measured and nothing is forecast. Select a sector for the research note.
| Sector | Intelligence exposure | Autonomy exposure | Infrastructure dependency | Capital intensity | Thesis horizon |
|---|---|---|---|---|---|
| Intelligence exposure: Structural | Autonomy exposure: Structural | Infrastructure dependency: High | Capital intensity: Structural | Current | |
The only sector with structural exposure to both revolutions at once. Every incremental unit of intelligence and every autonomous machine consumes the same constrained logic, memory and packaging capacity. This is the clearest single expression of the convergence thesis, and also the most crowded.
| |||||
| Intelligence exposure: Structural | Autonomy exposure: Moderate | Infrastructure dependency: Structural | Capital intensity: Structural | Current | |
An information-sector business with an industrial cost structure. Returns are decided by power procurement and interconnection rather than by tenant demand, which means the asset should be underwritten on utility timelines rather than technology timelines.
| |||||
| Intelligence exposure: High | Autonomy exposure: Moderate | Infrastructure dependency: Low | Capital intensity: Structural | Current | |
The first revolution converted electricity into a strategic input for a sector that had never treated it as one. The constraint is firmness and delivery, not generation in aggregate — which pushes value toward whoever controls interconnection and dispatchable capacity.
| |||||
| Intelligence exposure: Structural | Autonomy exposure: Low | Infrastructure dependency: Moderate | Capital intensity: Low | Emerging | |
The sector with the most capability exposure and the least capital intensity, which cuts both ways. Agentic systems move the product from tooling toward the work itself; that expands the addressable pool and simultaneously threatens the seat-based model that funded it.
| |||||
| Intelligence exposure: High | Autonomy exposure: Structural | Infrastructure dependency: Structural | Capital intensity: High | Developing | |
Forty years of disappointing returns came from perception and generalization, not actuation. Those are exactly the constraints the first revolution attacks, which is why this cycle deserves to be underwritten differently from the last three — while remaining the position most sensitive to being early.
| |||||
| Intelligence exposure: Moderate | Autonomy exposure: High | Infrastructure dependency: High | Capital intensity: Structural | Developing | |
The rare case where industrial policy and private capital point the same direction. Autonomy here is simultaneously a labour substitution and a reshoring instrument, so the demand signal is less dependent on the technology arriving on schedule.
| |||||
| Intelligence exposure: Moderate | Autonomy exposure: Structural | Infrastructure dependency: High | Capital intensity: High | Developing | |
The most legible autonomy market: movement is repetitive, already instrumented, and measured in units that convert directly into a return. Density is the constraint — the economics work network by network rather than nationally.
| |||||
| Intelligence exposure: High | Autonomy exposure: High | Infrastructure dependency: Moderate | Capital intensity: High | Emerging | |
Autonomy is procured as capability rather than as efficiency, which changes how it is priced and how long it is held. Contracts are durable and budgets are political; the binding constraint is industrial capacity rather than demand.
| |||||
06 / Horizons
Sequence, not schedule
Four postures, in the order the desk expects to hold them. No dates are attached to any of them, because none are known.
Observable now
Current
Capital is being committed to the physical substrate of intelligence faster than the applications above it have proven their economics. The interesting research question is not whether the buildout is large but which parts of it are durable.
01Infrastructure expansion
02Compute buildout
03Model proliferation
04Enterprise deployment
- Load-bearing systems
07 / Terminal
The desk view
Four projections of the same model. No market data, no feed, nothing streaming — every value on the panel is read from the research model published on this page.
Dual / Research Terminal
Thesis status
M01Two systems entering one dependency chain
Research horizon
M02Convergence
Thesis endpoint
Standing note
M03The two revolutions were independent for most of their history. They are no longer: perception runs on the same architectures as language, edge inference draws on the same supply chain as training, and both terminate in the same power constraint. When two systems share a dependency chain they stop being two theses and become one capital cycle.
Binding constraint
M04Shared silicon and shared power
Active systems
M05Open question
M06What is correctly priced as one cycle and what is still being priced as two?
Desk selection
M07No system selected. Choosing one anywhere in this document — the signal columns, the system map, the matrix, the horizon model — holds it here for the rest of the visit.
08 / Access
Private research
The public thesis is the whole of what is published today. The private research environment is in preparation, and opens to early readers first.
The public thesis
Everything on this page — the model, the matrix, the horizon framework.
Open
Thesis briefs
Long-form positions with the dependency work shown rather than summarised.
In preparation
Signal notes
Short releases when a constraint in the model moves, and why it matters.
In preparation
Capital maps
The system map extended into a named sector, traced to the balance sheet.
In preparation
Industry analysis
Where a sector's economics actually change, separated from where they are said to.
In preparation
Thesis revisions
What the desk changed its mind about, what caused it, and what it cost.
In preparation
System watchlists
The constraints under observation, and the conditions that would move them.
In preparation
Research archive
Every prior release, held against what subsequently happened.
In preparation
Access opens with the first private release.
Private releases will extend beyond the public thesis: the dependency work shown rather than summarised, sector maps traced to the balance sheet, and the revisions — what the desk changed its mind about, and what caused it. Requests open with the first of them.
Releases are infrequent and long. There is no newsletter cadence and no commentary between them.
One address, used for one purpose. This site sets no cookies and runs no analytics.
Research and commentary only. Nothing published constitutes investment advice.
Access
Requests are not being recorded yet.
Nothing on this page collects an address, sets a cookie, or runs analytics. When the first release is ready, this becomes a form.