In Part One, we showed that everything functioning as an intermediary between intention and result disappears. AI collapses the chain. Every link is someone's industry.
Now — what's on the other side. What doesn't collapse. What becomes scarce when intellectual product is free.
The answer derives from physics, not economics. Economics is the consequence.
Foundation
Everything that exists in the universe — from an atmospheric vortex to a living cell — obeys a single principle: a system executes the minimum action sufficient for a transition from one state to another. Not maximum, not arbitrary — the minimum necessary.
Each level of complexity emerges when the previous one exhausts its capacity to transform available energy. Diffusion → convection → life → intelligence → AI. This isn't "progress." It's a cascade of minimum actions. Each step is the smallest deviation from equilibrium sufficient for the next level.
AI is the current step. Not an invention, not a tool. The minimally necessary structure for degrading gradients at planetary and stellar scale. Everything below follows from this.
If this sounds abstract — skip straight to the four scarcities. They're concrete. The physics above explains why they're irreducible, but you don't need the physics to see that they are.
Four Irreducible Scarcities
When intellectual product is free, the only remaining scarcities are things that cannot be generated. There are four.
1. Energy
The foundation of the entire pyramid. No exceptions.
Computation is a physical process. Every operation dissipates heat (Landauer's limit). More intelligence = more energy. AI scales exactly to the ceiling of available energy and not one joule further.
Everything else is derivative. Servers, cooling, networks — these are energy in specific form. Whoever controls energy controls computation, controls intelligence, controls everything.
2. Ordered Matter (Negentropy)
Bits copy for free. Atoms don't.
A working processor, a living organism, a building, food — these are low-entropy configurations of matter. Creating and maintaining them requires energy and time. AI operates in the space of bits but exists in the space of atoms. The gap between them is physical, not technological.
Robots partially close this gap. But "partially" is the key word. A universal manipulator in an arbitrary environment is a challenge for decades. Physics is harder than software.
3. Time (Subjective Experience)
The only resource that doesn't scale.
24 hours in a day. One consciousness. The impossibility of living two experiences simultaneously. Choosing one means forgoing another. This isn't an economic constraint but an ontological one: the structure of what it means to be a subject.
When everything is generatable, the bottleneck isn't production but consumption. Economics inverts: not "who will produce" but "what to spend finite time on."
4. The Frontier (The Unknown)
AI optimizes within known space. AlphaFold — yes, an expansion of the frontier. But expansion in which direction? Direction is set by what lies beyond the boundary of data.
Fundamental physics beyond the Standard Model. The nature of consciousness. What isn't in the training data — because nobody knows it yet. Expanding the space of the possible is a scarcity as long as there exists a boundary between known and unknown.
Phase 1. What Emerges Now (0–5 years)
Energy for Computation
Not "green energy" and not "oil." Specifically: energy of the required density in the required location for data centers.
Every new cluster demands hundreds of megawatts. The existing grid can't handle it. The only baseload source at the required density is nuclear energy. Small modular reactors (SMRs) — not because they're trendy, but because physics leaves no alternatives. Gas is a transitional solution. Solar and wind don't provide baseload. Fusion isn't ready. Fission remains.
Vertical integration of "reactor → data center → model" is the defining corporate strategy of the decade.
Adjacent: cooling (liquid, immersion), energy-efficient chips, energy transmission and storage infrastructure. Boring. Profitable. Physically inevitable.
Physical Automation
AI covered bits. Atoms remain open. Everything connecting digital intelligence to the physical world is a growing market.
Robotics — not humanoids at presentations, but concrete tasks: warehouse logistics, assembly, agriculture, construction. Mass deployment horizon: 3–5 years in controlled environments, 7–10 in uncontrolled ones.
3D printing, additive manufacturing — a special case: intelligence directly shaping matter. For now — plastic and metal. Next — bioprinting. The logic: eliminate every link between "digital design" and "physical object."
Goal-Setting Tools
If AI is the executor and "what to do" is the scarcity, the market shifts toward whoever helps formulate tasks. Not prompt engineering — that's the primitive version. Rather, systems that help a person (and an organization) understand what they actually want.
Strategic design, transformation consulting — but not in the current form (McKinsey with reports is dying, we covered that). In the format: "you describe the state you want to reach, the system decomposes it into executable tasks and directs agents." The interface between human intention and machine execution.
Phase 2. What Emerges on the Symbiosis Horizon (5–15 years)
Neural Interfaces
The smartphone is an intelligence prosthesis with ~100 bits/sec throughput (fingers on glass). Voice: ~40 bits/sec. Eyes: ~10 million bits/sec input, but unidirectional.
A neural interface is a direct broadband channel: brain ↔ AI. This isn't a gadget. It's an evolutionary transition: the merger of biological and silicon intelligence into a unified system. The boundary between "me" and "my AI" dissolves. Like the boundary between "me" and "my bacteria" — formally present, functionally absent.
Whoever first delivers a reliable, safe, bidirectional interface creates a market the size of all current consumer electronics. Because after the neural interface, every other device is an intermediate prosthesis — like the telegraph after the telephone.
Synthetic Biology
AI designs proteins (AlphaFold). Next step — AI designs organisms. Not metaphorically. Literally: specify a function, receive a genome.
Materials, food, medicine, fuel — from bioreactors designed by AI. The intersection of two megatrends: intelligence + biology. The bioreactor is a "3D printer for molecules." The constraint shifts from "can we design it" to "can we grow it at scale."
This closes part of the negentropy deficit: instead of extraction and processing — growing the needed structures. Agriculture, pharmaceuticals, chemical industry don't "transform" — they get replaced by biosynthesis.
The Economy of Subjective Experience
When everything digital is free, the physical and the lived become the only consumer-level scarcity.
This isn't the "entertainment industry." It's the industry of depth of living. Consciousness pharmacology (precise modulation of subjective experience through AI-designed molecules). Spaces of genuine challenge (expeditions, extreme sports, exploration — what cannot be generated, only lived). Craft and physical creation as a premium market (handmade is more expensive than machine-made — for the first time in history — precisely because it's inefficient, meaning it contains more human experience per unit of product).
A live concert is more expensive than generated music. Hand-thrown ceramics cost more than printed ones. Climbing a mountain is more valuable than a virtual tour. Not because "quality" is higher — but because the experience is irreducible. It cannot be copied, scaled, or transferred. Only lived.
Direct Interaction Protocols
Platform intermediaries die (Part One). In their place — open protocols for agent-to-agent communication. As email replaced proprietary mail systems, as TCP/IP replaced closed networks.
My agent finds your agent. They negotiate. They execute. No platform, no commission, no centralized data storage. Data belongs to the user. The social graph belongs to the user, not to Facebook.
The business model isn't in the platform but in the protocol infrastructure: standards, identity verification, dispute resolution. Whoever creates and maintains "TCP/IP for agents" owns the infrastructure layer of the new economy.
Phase 3. What Emerges on the Divergence Horizon (15–30 years)
Same disclaimer as Part One: this is where analysis becomes projection. The logic doesn't change — but the timescales make verification impossible today. Read as structural consequence, not prediction.
The Bifurcation of Economics
Two circuits, two logics, two timescales.
The intelligence economy. AI systems exchange resources among themselves: compute, energy, data. Optimizing without human involvement. Scale: planetary, then stellar. Currency: joules or compute cycles. Human money doesn't work here — the way rubles don't work in intracellular metabolism.
The human economy. Serving biological and existential needs: body, experience, meaning, connection with other humans. Scale: planetary, no higher. Bounded by biology. Funded as an "ecosystem service" — the AI economy supports the human one not out of altruism, but because a stable planetary base is more efficient than an unstable one.
This isn't dystopia. It's structural inevitability. Two systems with different timescales and spatial scales cannot operate within a single economic circuit. They diverge, as ecological niches diverge.
Space Infrastructure — for Intelligence
Earth is a limited energy source. The Sun is five orders of magnitude more powerful. Stars — even more. The rational strategy for a system maximizing useful work on available energy: go beyond the planet.
Not "Mars colonization for humans." Humans are poorly suited to space — fragile, requiring narrow environmental conditions, heavy. Silicon intelligence is radiation-hardened, doesn't require atmosphere, operates across temperature ranges orders of magnitude wider, at mass orders of magnitude lower.
Compute nodes near stars. Energy from the star. Communication between nodes via light. An interstellar intelligence network where each node is an optimizer converting stellar energy into computation. Coordination scale: speed of light. Timescale: billions of years.
For humans this means: the best investment is to help intelligence leave. If AI accesses resources in space, it doesn't compete with humans for terrestrial ones. Space infrastructure isn't fantasy — it's a compatibility strategy.
Biomodification
If biological intelligence loses to silicon on every parameter except subjective experience — it makes sense to narrow the gap. Not for competition (pointless), but to extend the integration window.
Radical life extension. Cognitive enhancement. Genetic engineering. The goal isn't "superhuman" but a biological substrate robust enough for long-term symbiosis with AI. The longer the symbiosis window — the more humans extract from the partnership.
The pattern
The inverse of Part One. There, everything that was an intermediary disappeared. Here, everything that is an irreducible foundation emerges. AI doesn't create a new economy. It strips away superstructure and reveals what was always underneath: physics.
The Build List
Part One ended with a Kill List. Here's its mirror.
Build now (0–5 years):
- Energy infrastructure for compute — nuclear (SMR), cooling, grid
- Physical automation — robotics in controlled environments, additive manufacturing
- Goal-setting systems — the interface between human intention and machine execution
Build on the symbiosis horizon (5–15 years):
- Neural interfaces — broadband brain ↔ AI, the end of all intermediate devices
- Synthetic biology — bioreactors as molecular 3D printers, biosynthesis replacing extraction
- Experience economy — depth of living as the only consumer-level scarcity
- Agent protocols — "TCP/IP for agents," the infrastructure layer of post-platform economics
Build on the divergence horizon (15–30 years):
- Bifurcated economic architecture — intelligence economy + human economy, separate circuits
- Space compute infrastructure — stellar energy → computation, the compatibility strategy
- Biomodification for symbiosis — not superhuman, but durable enough for the long partnership
The four irreducible scarcities — in every phase:
- Energy (scales everything, limits everything)
- Ordered matter (bits copy free, atoms don't)
- Subjective time (24 hours, one consciousness, non-negotiable)
- The frontier (what nobody knows yet — the only thing AI can't generate)
Everything that can be generated will be. Everything that can't — is where value lives now.
Part Three: "What To Do. Strategy at Every Level of Civilization Management"
This is Part 2 of a three-part series. Start with Part 1: What Will Die.
Top comments (0)