1. The Forces

If you understand which forces are accelerating, where they reinforce each other, and which feedback loops dominate, you can reason about what comes next without relying on anyone's headline forecast.

The forces do not arrive one at a time. They arrive together, and the interactions matter more than the individual curves. A robotics breakthrough in isolation is interesting. A robotics breakthrough combined with near-zero-cost intelligence, collapsing energy costs, and atomically precise manufacturing is a different kind of event. The map must be read as a system, not a list.

1. The intelligence curve

The dominant force. Not because artificial intelligence is the most important technology in some abstract ranking, but because intelligence is the input to every other curve on this map. Faster intelligence accelerates materials discovery, drug design, chip architecture, energy system optimisation, weapons development, governance capacity, and its own improvement. This is the master lever, and it is being pulled harder and faster than almost anyone anticipated even twelve months ago.

1.1 Foundation models: the capability ramp

In early 2025, the frontier models could hold a conversation, write passable code, and perform well on standardised tests. By April 2026, the picture has changed in ways that matter structurally, not just incrementally. Anthropic's Claude Opus 4.6 (released 5 February 2026) tops benchmarks in agentic coding, computer use, and tool use. Twelve days later Sonnet 4.6 delivered a one-million-token context window in beta. OpenAI's GPT-5.4 (5 March 2026) scored 83% on the GDPval knowledge-work benchmark and set records on computer-use tests. DeepSeek V4 arrived with a trillion parameters and native multimodality.

The important shift is not the benchmark numbers. It is what the models can now do in practice: operate a computer end to end, maintain context across a million tokens of working memory, and chain together multi-step professional tasks - legal analysis, financial modelling, software architecture, experimental design - with declining need for human correction at each step. The distance between "impressive demo" and "deployed replacement for a junior professional" closed faster in Q1 2026 than in the preceding two years combined.

1.2 Agents: from chat to autonomous execution

The shift from conversational AI to agentic AI is the most consequential change in the deployment surface. An agent is not a chatbot that remembers your name. It is a system that receives a goal, decomposes it into sub-tasks, uses tools, recovers from errors, and delivers a result - often without human intervention between the goal and the output.

In January 2026, Anthropic launched Claude Cowork as a persistent agent workflow environment, now generally available on macOS and Windows. By March, the "Dispatch" feature allowed Claude to use your computer autonomously while you are away from the desk. Anthropic's Model Context Protocol crossed 97 million installs in March and was taken under Linux Foundation open governance - meaning the standard by which agents connect to external tools, APIs, and data sources is now shared infrastructure, not a proprietary advantage. Close to 75% of businesses plan to deploy AI agents by end of 2026 (Deloitte). The Agentic List 2026 identified 120 companies building enterprise-grade autonomous systems.

The economic logic is stark. An agent that can complete a four-hour analyst task in twelve minutes does not merely save time. It changes the cost structure of every firm that employs analysts. When the agent can also hand off its output to another agent - one that formats the deliverable, another that checks compliance, a third that schedules the client meeting - you get what the industry is calling "digital assembly lines." These are not hypothetical. They are in production at scale in legal, financial, and consulting firms in Q1 2026.

1.3 Recursive improvement and the intelligence feedback loop

The most important feature of the intelligence curve is that it bends upward on itself. AI systems are now the primary tool used to design, train, evaluate, and optimise the next generation of AI systems. Google's TurboQuant algorithm (presented at ICLR 2026) dramatically reduces the memory overhead of the KV cache, one of the main bottlenecks in scaling context windows - an efficiency breakthrough that was itself produced by AI-assisted research. The chip designs that will power 2027's models are being laid out with the help of 2026's models.

This is not yet full recursive self-improvement in the sense that alarms some safety researchers - where a model autonomously rewrites its own architecture to become smarter without human oversight. But the gap between "AI-assisted AI research" and "AI-led AI research" is narrowing. Each generation of model makes the next generation arrive faster and perform better, which in turn makes the generation after that arrive faster still. The curve is not linear. It is not even simply exponential. It is an exponential whose exponent is itself growing.

2. The embodiment curve

Intelligence without a body can reshape information work. Intelligence with a body can reshape the physical world. The embodiment curve tracks the convergence of robotics hardware, sensor systems, AI-driven motor control, and manufacturing cost reduction that is turning robots from expensive industrial tools into general-purpose physical agents.

2.1 Humanoid robotics: the deployment threshold

At CES 2026, Boston Dynamics unveiled Electric Atlas, an enterprise-grade humanoid designed for material handling and order fulfilment, integrated with Google DeepMind's Gemini Robotics AI for reasoning through complex instructions in unstructured environments. Tesla announced 50,000 Optimus Gen 2 units for 2026 at $20,000-30,000 per unit. 1X's NEO humanoid began home deliveries. Unitree shipped over 5,500 humanoid robots in 2025, and manufacturing costs dropped 40% between 2023 and 2024.

The numbers matter because they cross a threshold. At $25,000 a unit, a humanoid robot costs less than one year's salary for most human workers in most OECD countries. At $16,000 (Unitree's G1 price point), it costs less than a year's salary for most workers on Earth. These units do not take breaks, do not get injured, do not need health insurance, and operate around the clock. They are not yet as versatile as a human, but the gap between "can do specific warehouse tasks" and "can navigate an unstructured home" is closing in quarters, not decades.

2.2 Beyond humanoids: swarms, soft robots, and micro-scale

Humanoids get the press coverage, but the embodiment curve extends well beyond bipedal form factors. Drone swarms are already operational in military contexts and are being adapted for agriculture, infrastructure inspection, and logistics. Soft robotics - flexible, compliant machines made from polymers and other deformable materials - are entering surgical suites and food processing lines where rigid robots cannot safely operate. At the micro and nano scale, xenobots (living robots made from biological cells) and synthetic micro-machines are being developed for targeted drug delivery, environmental remediation, and in-body diagnostics.

The common thread is that intelligence is being given physical agency across every scale, from warehouse floors to human bloodstreams. Each scale creates its own set of economic, medical, military, and ethical implications.

2.3 The intelligence-embodiment feedback loop

NVIDIA's Isaac GR00T open models now enable robots to understand natural language instructions and perform complex multi-step tasks using vision-language-action reasoning. This means the intelligence curve and the embodiment curve are no longer developing in parallel - they are feeding each other directly. Better AI makes robots more capable. More capable robots generate more training data. More training data makes AI better at controlling robots. The loop is closed, and it is accelerating.

3. The biology curve

Biological systems - human bodies, ecosystems, agricultural crops, pathogens - were until recently treated as given. Things that evolved, that you worked with or around, that you treated when they broke but did not redesign from scratch. That assumption is now obsolete.

3.1 Gene editing: from repair to redesign

CRISPR was the opening act. In 2026, the toolkit has expanded well beyond simple double-strand breaks. A January 2026 breakthrough demonstrated turning genes on without cutting DNA at all, by removing chemical tags that silence gene expression. Prime editing is entering broader human validation across multiple therapeutic areas. Retron-based editing showed improved precision and efficiency, correcting multiple disease-causing mutations simultaneously in mammalian cells. Base editing is now in expanded Phase 2 clinical trials (Verve Therapeutics' VERVE-102 for permanent LDL cholesterol reduction, following Eli Lilly's acquisition).

The trajectory is from single-gene repair to multi-gene editing to programmable biological redesign. Each step makes the previous one look primitive. And critically, each step is being accelerated by the intelligence curve: AI systems design the guide RNAs, predict off-target effects, model protein folding, and optimise delivery vectors faster than any human team.

3.2 Computational biology and AI-driven drug discovery

Eli Lilly inaugurated LillyPod - the pharmaceutical industry's most powerful AI supercomputer - built on 1,016 Blackwell Ultra GPUs delivering over 9,000 petaflops. Where traditional wet labs test roughly 2,000 molecular hypotheses per year, LillyPod can simulate billions in parallel. Lilly aims to cut the typical ten-year drug development timeline in half. MIT's protein-based drug design model predicts how synthetic proteins fold and interact, reducing pharmaceutical R&D costs by billions. Insilico Medicine's first AI-discovered drug reached Phase 2 trials.

This is not incremental improvement to the pharmaceutical pipeline. It is a structural transformation of the discovery process itself. When you can test billions of molecular hypotheses computationally before touching a pipette, the bottleneck shifts from discovery to clinical validation and regulatory approval - and there is already pressure to accelerate those stages too.

3.3 Longevity and the optional-aging frontier

Altos Labs initiated human clinical trials in 2026 for cellular rejuvenation using Yamanaka factors, targeting neurodegenerative and immune-related aging disorders. The longevity field is moving from isolated interventions to systems-level thinking: aging understood not as a defect but as a progressive loss of coordination between biological systems, and therapies aimed at restoring that coordination rather than patching individual symptoms.

If even a fraction of this programme delivers - if biological aging becomes meaningfully decelerateable within the next decade - the downstream implications cascade through every domain in this document. Pension systems, retirement economics, generational wealth transfer, career planning, political succession, ecological footprint, population dynamics, the very concept of a "life stage" - all of it bends. A student entering a 5.5-year medical programme in 2026 would be training for a field whose foundational assumptions about what bodies do and when they fail may shift under her feet before she finishes.

3.4 Synthetic biology and the engineered biosphere

Researchers are engineering microbes to produce complex medicines, programming cell therapies, and designing organisms that do not exist in nature. Colossal Biosciences created "woolly mice" - rodents with mammoth-like traits - using synthetic biology techniques, a proof-of-concept for genetic de-extinction that also demonstrates the ability to radically rewrite the phenotype of a living organism. Gene drives - genetic modifications designed to spread through wild populations - remain in development for applications ranging from malaria vector control to invasive species management.

The deeper implication is that the biosphere is becoming a design space. Biological evolution, the process that produced every living thing on Earth including us, is being supplemented, redirected, and in places replaced by intentional engineering. This is not a future possibility. It is happening now, in labs that are themselves being accelerated by the intelligence curve.

4. The energy curve

Energy is the other master input, alongside intelligence. Nearly everything humans do requires energy, and the cost and availability of energy set hard constraints on what is economically feasible. Those constraints are loosening faster than most forecasts anticipated.

4.1 Solar plus storage: the deflationary engine

The cost of solar photovoltaic electricity has fallen roughly 90% since 2010 and continues to decline. Battery storage costs are on a parallel trajectory. In the sunniest regions, solar-plus-storage is already the cheapest source of new electricity generation by a wide margin. The implication is not just cheaper electricity. It is a structural shift in the political economy of energy: from a world where energy is scarce, concentrated, and controlled by whoever sits on the fuel deposits, to one where energy is abundant, distributed, and approaching near-zero marginal cost in favourable geographies.

Near-zero marginal-cost energy, combined with near-zero marginal-cost intelligence, is a combination that has never existed in human history. It removes the two primary inputs that have constrained economic activity since the industrial revolution. The downstream effects are not a gentle acceleration of existing trends. They are a phase change in what is economically possible.

4.2 Fusion: from "thirty years away" to "under construction"

Fusion energy is no longer a joke about perpetual delay. The US Department of Energy announced $352 million in frontier science funding in March 2026, including substantial fusion and plasma physics allocations. Multiple private companies are now building pilot reactors rather than writing proposals. The timeline to commercial fusion has compressed from "decades" to "this decade, possibly" - not because the physics changed, but because AI-driven plasma modelling, advanced materials designed by computational methods, and vastly cheaper high-temperature superconducting magnets have attacked the engineering bottlenecks from multiple directions simultaneously.

Commercial fusion would not just add to the energy supply. It would remove the last serious argument that energy constraints prevent indefinite economic growth. Combined with the intelligence curve, it would mean that both the cognitive and physical inputs to production become effectively unlimited. Whether that produces abundance or catastrophe depends on distribution, governance, and timing - questions this project addresses in Parts 2 through 4.

4.3 The energy-intelligence feedback loop

AI systems consume enormous amounts of energy. The first orbital data centre nodes launched to low-Earth orbit on 11 January 2026. NVIDIA's Space-1 Vera Rubin Module (announced 16 March) delivers data-centre-class AI performance in orbit. SpaceX filed for an orbital data centre constellation of up to one million satellites. Starcloud raised $170 million to fund orbital deployment.

The logic is not mad. In orbit, solar energy is available 24 hours a day, cooling is free, and the limiting factor - ground-based power grid and real estate constraints - vanishes. If the energy and launch cost curves continue their current trajectories, orbital computation may become cheaper than terrestrial computation within the decade. This is the energy-intelligence feedback loop running at planetary scale: cheaper energy enables more intelligence, more intelligence designs better energy systems, cheaper energy enables even more intelligence.

5. The manufacturing and materials curve

Intelligence and energy are inputs. What they produce depends on what you can build with them. The manufacturing curve is being reshaped by three converging forces: AI-driven materials discovery, advanced fabrication techniques approaching atomic precision, and the collapse of traditional supply chain assumptions.

5.1 Computational materials discovery

When an AI system can simulate the properties of millions of candidate materials before any of them are synthesised, the design-test-iterate cycle that historically took years compresses to weeks. This is already happening. New battery chemistries, high-temperature superconductors, lighter and stronger structural materials, better catalysts for chemical processes - each of these feeds back into other curves on this map. A better battery chemistry accelerates the energy curve. A better catalyst accelerates the biology curve. A better superconductor accelerates the intelligence curve by enabling more efficient computing hardware.

5.2 Distributed and atomically precise fabrication

3D printing and additive manufacturing have been improving steadily for years, but the convergence with AI design tools is producing a qualitative shift. When an AI system can design an optimal component and a fabrication system can produce it locally without tooling, the traditional model of centralised mass production and global supply chains faces a structural challenge. Not an immediate collapse - the incumbent systems have enormous inertia - but a growing number of products and components where distributed fabrication is cheaper, faster, and more adaptable.

At the far end of this trajectory sits atomically precise manufacturing: the ability to arrange atoms exactly where you want them. This remains more aspiration than reality in April 2026, but the gap between aspiration and laboratory demonstration is closing, and when it does close, it represents another phase change - from a world where manufacturing is constrained by available materials and processes to one where anything physically possible can be built.

6. The surveillance, persuasion, and synthetic media curve

Every force described above has a shadow application in surveillance, persuasion, and information manipulation. This is not a separate domain. It is the same technologies applied to the problem of knowing what people think, shaping what they believe, and fabricating evidence that is indistinguishable from reality.

6.1 Ambient sensing and total awareness

The combination of cheap sensors, ubiquitous connectivity, AI-driven analysis, and declining storage costs means that comprehensive surveillance of physical spaces, digital communications, financial transactions, biological signals, and social networks is now technically feasible and economically affordable for any state and many corporations. The question is not whether this capability exists. It is who has it, who is constrained by law or norms from using it, and how quickly those constraints erode under pressure.

6.2 Synthetic media and the collapse of evidence

Generative models can now produce text, images, audio, and video that are indistinguishable from authentic content by most humans and many automated systems. The implications go well beyond "deepfakes of politicians." When any piece of evidence - a photograph, a recording, a document, a video - can be fabricated at negligible cost and with high fidelity, the evidentiary foundations of journalism, law, science, diplomacy, and personal trust are all compromised simultaneously. This is not a future risk. It is a present condition, and the defences (watermarking, provenance tracking, detection algorithms) are losing the arms race.

6.3 Personalised persuasion at scale

An agent that can hold a million-token context window, remember your preferences across conversations, and operate your computer autonomously is also an agent that could, in principle, be directed to persuade you of things that are not in your interest. The same capability that makes AI assistants useful - deep understanding of individual context, persistent memory, ability to act - makes AI persuaders dangerous. When every individual can be targeted with arguments tailored to their specific psychology, belief structure, social context, and emotional state, the concept of informed consent in democratic societies faces a stress test it was not designed for.

7. The defence and security curve

The forces above are not arriving into a peaceful, cooperative international system. They are arriving into a world of great-power competition, regional conflicts, nuclear arsenals, and states that view technological superiority as an existential priority.

7.1 Autonomous weapons and drone warfare

Autonomous drones are already operational in conflicts around the world. The combination of cheap hardware, AI target identification, and swarm coordination produces weapons systems that are difficult to defend against, easy to scale, and do not require skilled human operators. The barrier to entry for lethal autonomous capability is dropping. A reasonably resourced non-state actor could, with commercially available drones and open-source AI models, assemble a meaningful autonomous strike capability in months, not years.

7.2 Cyber and biological offence-defence asymmetry

AI-accelerated cyber offence is advancing faster than AI-accelerated cyber defence, a pattern that holds across most adversarial domains. The same applies to biological threats: the tools for designing novel pathogens are becoming more accessible faster than the tools for detecting and responding to them. The gene editing curve described in section 3 is dual-use in the most direct possible sense. A system that can design a therapeutic protein can also design a toxic one. The intelligence required to do the latter is falling as the underlying tools improve.

7.3 Multipolar AI competition

The United States, China, the European Union, and a handful of other actors are engaged in an AI capabilities race that each views through a national security lens. The competitive dynamics create pressure to deploy faster, regulate less, and share less. This is the opposite of what the coordination problems around AI safety require. The tension between competitive pressure and cooperative necessity is one of the defining political dynamics of the 2026-2030 period.

8. The governance and legitimacy curve

The EU AI Act becomes fully applicable on 2 August 2026, including mandatory regulatory sandboxes in every member state. It is the most ambitious attempt to regulate AI anywhere in the world, and it is already struggling to keep pace with the technology it aims to govern. The act was drafted primarily with traditional AI systems in mind. Agentic AI - systems that act autonomously, chain tools together, and make consequential decisions without a human in the loop - fits awkwardly into a framework designed around "deployers" and "providers" of discrete systems.

This is not a problem specific to the EU. It is a structural mismatch between the speed of technological capability and the speed of institutional adaptation. Democratic governance operates on cycles of years to decades. The intelligence curve operates on cycles of weeks to months. The gap between what is technically possible and what is legally, ethically, or politically governed is widening, not narrowing.

8.1 Machine-executed law and algorithmic governance

As AI systems become capable of interpreting regulations, monitoring compliance, and executing enforcement actions, the question of whether law should be written for human readers or machine executors becomes practical rather than theoretical. Smart contracts were an early, crude version of this idea. The current generation of AI agents can handle far more complex regulatory logic. The efficiency gains are real: faster compliance, fewer errors, lower administrative costs. The governance risks are also real: reduced discretion, algorithmic rigidity, difficulty challenging automated decisions, and the transfer of interpretive authority from human judges to AI systems whose reasoning is opaque.

8.2 Legitimacy when the best decision-makers are not human

If AI systems consistently outperform human experts at diagnosis, investment, policy analysis, scientific research, and strategic planning - and the evidence that they do in specific domains is mounting rapidly - then the justification for human authority shifts from competence to something else. Consent, accountability, representation, dignity, tradition - the alternative foundations are available, but none of them were designed for a world where the entity making the decision is demonstrably better at it than any human. This is uncomfortable territory, and it will become more uncomfortable as the capability gap widens.

9. The financial and economic curve

Capital allocation is being automated. Programmable money (central bank digital currencies, stablecoins, smart contracts) enables transactions that execute automatically when conditions are met, without human intermediation. Autonomous funds that use AI to research, model, trade, and manage risk already exist and are growing. The US alone captured $227.7 billion across 166 AI agent-related companies in 2024-2025.

The deeper shift is in the cost structure of the firm itself. When intelligence is cheap and embodied agents are available, the traditional corporation - a structure for coordinating human labour and capital - faces pressure from a new organisational form: the near-fully autonomous firm that operates with minimal human headcount, using AI agents for strategy, operations, compliance, and customer interaction. This is not science fiction. It is the stated ambition of multiple funded startups in 2026, and the direction in which every large consultancy is nudging its clients.

For labour markets, the implication is severe. Over 61,000 employees were impacted by AI-driven layoffs in the first months of 2026 alone, with modelling estimates suggesting the real figure is three to five times higher. McKinsey - the firm that advises other firms on strategy - laid off approximately 200 employees after automating non-client-facing work. Baker McKenzie - one of the world's largest law firms - is cutting up to 10% of its global workforce as part of an AI transition. These are not blue-collar manufacturing jobs. They are the white-collar, credential-dependent, knowledge-economy positions that two generations of Western education systems trained their best students to fill.

Dario Amodei, CEO of Anthropic, warned in January 2026 that AI may cause "unusually painful" disruption to jobs. He is not an alarmist. He is the person building the tools. When Fortune reported "the week the AI scare turned real" in late February 2026, it was describing a mood shift in which the abstract possibility of mass white-collar displacement started to feel concrete to the people most affected. An Industriell Ekonomi student at LiU in early 2026, two years from graduation into management consulting or investment banking, is standing squarely in the path of this force.

10. The space industrialisation curve

Space is not a separate domain. It is a downstream application of cheap intelligence, cheap energy, and cheap launch. The orbital data centre nodes that launched on 11 January 2026 are not primarily about space. They are about escaping the terrestrial constraints on the intelligence curve: land use, power grid capacity, cooling costs, and political jurisdiction. SpaceX's filing for a million-satellite orbital data centre constellation is a bet that the economics of orbit will be favourable within the decade.

Beyond computation, the same cost curves that make orbital data centres plausible make orbital manufacturing, asteroid mining, and space-based solar power collection plausible on similar timescales. The relevant question is not "will humans go to space" in some aspirational sense. It is whether the economic returns from space-based industry will be large enough, soon enough, to redirect significant capital and attention away from terrestrial problems - or toward them, via resources and capabilities that cannot be obtained on Earth.

The interaction map: where the curves meet

No single curve on this map is sufficient to produce the transition. The transition is the product of their interaction. Here are the five reinforcing loops and critical intersections that drive the system.

The three meta-dynamics

Beneath the individual curves and their interactions, three meta-dynamics shape the system as a whole. These are the forces that determine which scenarios in section 2 are most plausible.

Meta-dynamic 1: The compression of transition time

Previous technological transitions - agricultural, industrial, informational - played out over decades to centuries. The current transition is playing out over years. The 12 major features Anthropic shipped in 12 weeks in early 2026 are a microcosm: the pace of capability deployment is so fast that institutions, regulatory frameworks, educational systems, labour markets, and individual career plans cannot adapt quickly enough through normal mechanisms. The transition is outrunning the adaptive capacity of every human system designed to manage change. This is not a failure of any particular institution. It is a structural mismatch between the speed of technological change and the speed of human institutional response, and it is getting worse.

Meta-dynamic 2: The decoupling of capability from human agency

For all of human history, the most capable agent in any domain was a human being, or a group of human beings, possibly assisted by tools. This is ceasing to be true, domain by domain, at accelerating pace. The decoupling is the central fact of the transition. It does not mean humans become irrelevant. It means that the relationship between human effort and economic, intellectual, and creative output is being broken. Value creation no longer requires human labour in many domains, and the number of such domains is growing. This has no precedent. Every previous economic system, every theory of political legitimacy, every educational philosophy, and every personal conception of purpose and status was built on the assumption that human capability matters because humans are the most capable agents available. When that assumption fails, everything built on it is in play.

Meta-dynamic 3: The distribution question as the binding constraint

The forces on this map are sufficient, in aggregate, to produce material abundance for every human being on Earth. Enough energy, enough intelligence, enough manufacturing capacity, enough medical capability. The binding constraint is not production. It is distribution: who gets access, on what terms, through what mechanisms, decided by whom. The existing distribution mechanisms - wages, markets, taxes, transfers - were designed for a world where human labour is the primary input to production. If human labour is no longer the primary input, those mechanisms do not merely need reform. They need replacement. And the replacement must be designed, built, and deployed on a timeline that matches the pace of displacement - which, as meta-dynamic 1 establishes, is faster than anything human institutions have previously managed.

This is the central tension of the transition. The production problem is being solved. The distribution problem is not. Whether the gap between them produces abundance, immiseration, or conflict is the question that section 2's scenarios attempt to map.

Forces not yet on the map

This analysis has focused on the forces with the strongest evidence and the clearest mechanisms as of April 2026. Several additional forces deserve mention because they could alter the map significantly if they develop faster than currently expected.

Brain-computer interfaces and cognitive augmentation. Neuralink and competitors are advancing, but the current state is limited to medical applications (paralysis, sensory restoration). Consumer-grade cognitive augmentation remains years away. If it arrives sooner, it reopens questions about human capability that the rest of this analysis treats as largely closed.

Climate intervention and geoengineering. Solar radiation management, enhanced weathering, direct air capture, and ocean alkalinity enhancement are all under active research. If any of these is deployed at scale, it changes the background conditions for every other force on this map. If none is deployed and climate disruption accelerates, it changes them differently.

Mirror life and novel biological risks. The synthesis of organisms using mirror-image biochemistry (D-amino acids and L-sugars) is theoretically possible and would produce life forms against which no evolved immune system has any defence. This is a tail risk with potentially civilisation-ending consequences and no current governance framework.

Quantum computing's second-order effects. Google's Quantum Echoes algorithm demonstrated 13,000x speed advantage over classical supercomputers. IonQ outperformed classical HPC in medical device simulation. If quantum advantage generalises, it accelerates the intelligence curve (faster training and inference), the biology curve (protein folding and molecular simulation), and the security curve (cryptographic vulnerability). The timeline remains uncertain, but early 2026 results suggest it may arrive sooner than the median forecast.

These forces are not included in the main map because their timing and magnitude are less certain. They are included here because any serious foresight exercise must flag the known unknowns alongside the known knowns. section 4 will return to what is genuinely unknowable.