INFRASTRUCTURE MYTH / SUPERARTIFACT LINEAGE · 2026

Artifacts Before the Jump

The tower already contained artifacts. Dime Tower begins with relics, tablets, and inherited memory. Marcie is the intelligence built to interpret them. The first jump is the moment that memory becomes operational and continuity survives the loss of its original carrier.

Eissens (2026) reads Dime Tower as an infrastructure myth, not as evidence that a 1991 game predicted AI. The useful pattern is structural: relics and tablets preserve an earlier age; an intelligent machine is built to search and interpret them; the carrier reaches a boundary it cannot cross; and capability is handed forward through another actor.

1. The artifacts were already there

In the English Final Fantasy Adventure script, Marcie introduces herself inside Dime Tower with the blunt statement: “I SEARCH RELICS.” The tower contains tablets left by the old Vandole world. They preserve claims about the Mana Tree, the corruption of the Tree by people, the water that carries that corruption back to people, Vandole’s use of Mana, and the Gemma Knight.

That changes the SuperArtifact lineage. The old artifacts do not appear after intelligence is released. They precede the jump. They are static, persistent containers of memory and rule-like knowledge from a prior system.

2. Marcie is not the first artifact

This matters because it prevents an overly neat story in which intelligence appears first and artifacts come later. Dime Tower already contains an artifact ecology of sorts: inscriptions, ruins, machinery, weapons, remembered procedures, and historical claims.

Marcie’s distinctive role is not that she introduces artifacts into the world. It is that she interfaces with inherited artifacts. She makes old memory usable to a present actor.

Secondary franchise references identify Marcie as the Japanese Mummy Seeker and give the expanded designation MRC-91. The original English Game Boy script safely establishes the narrower facts used here: Marcie searches relics, was made by Dr. Bowow, has waited fifty years, can restore MP, and cannot make the final jump. The expanded designation is therefore treated as secondary metadata rather than primary script evidence.

3. Static artifact versus routed artifact

The contrast is not artifact versus no artifact. It is a difference in operational reach.

The conceptual sequence is therefore:

The artifacts were already there. What changed was whether memory could become operational and survive the loss of its original carrier.

4. The Internet as sealed memory

The same distinction can be used as a metaphor for the web.

The Internet already contains an enormous body of persistent human artifacts: pages, source code, databases, wikis, forums, papers, images, standards, archives, games, maps, conversations, repositories, and institutional records. Calling this sealed memory does not mean the Internet is literally sealed or passive. Search engines, APIs, software, hyperlinks, and people already make it highly operational.

The phrase marks a narrower contrast: much of the web’s accumulated memory must still be found, interpreted, translated into context, and connected to an action. A sufficiently capable AI or agent layer can compress some of those steps. It can retrieve, synthesize, call tools, move between representations, and increasingly act through the material it finds.

Under this reading, superintelligence is not simply “more intelligence sitting on top of the Internet.” The architectural question is whether a new intelligence layer turns stored artifacts into a denser field of reachable operational capability.

This remains a hypothesis. Search already performs part of the transition, and many web artifacts are already executable. The proposed change is one of degree, composition, and continuity rather than a magical before/after boundary.

5. Marcie can read. The body still cannot jump.

SuperArtifact.si reads 2026 as an intermediate stage rather than the completed transition. In metaphorical shorthand, the GPT-4 era was the point at which “Marcie can read” became convincing at scale; the 2026 agent layer is the point at which Marcie gets tools. This is an interpretive compression, not a claim that one model release cleanly divides technological history.

The important architectural shift is already visible in software. OpenAI’s Agents API gives agents environments in which they can work with files, execute code, connect to MCP servers, and produce artifacts. OpenAI’s computer-use tooling also allows models to operate browser and desktop interfaces. Android 17 goes further at the operating-system layer: Google describes Android as beginning a transition from an operating system to an “intelligence system”, while AppFunctions expose app capabilities as orchestratable tools that privileged agents can discover and invoke with access to local app state.

That means the interpreter is no longer limited to reading inherited artifacts. It can increasingly operate through old software and create new synthetic artifacts in return: code, documents, configurations, websites, tool definitions, workflows, and other reusable outputs.

In that sense, software can begin to “eat” software without erasing its ancestry. Old artifacts become callable capability; new artifacts are generated from that capability; those outputs can themselves become inputs or tools for later actors. This is one bridge from the First Jump to Superartifact Recursion.

The physical world remains a harder boundary. Stanford’s 2026 AI Index reports 66.3% accuracy on OSWorld, a benchmark for computer-use agents, while robots succeed in only 12% of real household tasks. Google DeepMind’s Gemini Robotics 2 can control full humanoid bodies and other robot embodiments, but DeepMind still describes movement speed and dexterous manipulation as areas requiring further progress.

The intelligence is no longer trapped in the tower. But its body still is.

That sentence is metaphorical. It does not claim that current AI systems are subjectively curious or literally want to leave software. “Going outside” names an engineering transition: connecting general-purpose interpretation and planning to cameras, sensors, actuators, vehicles, machines, rooms, tools, and other physical systems.

The missing piece is therefore not simply “robots.” Hardware already exists. What remains comparatively immature is a cheap, general, permissioned embodiment layer in which physical capabilities can be discovered, addressed, composed, and invoked as easily as software tools increasingly can. A physical analogue of MCP or AppFunctions is one possible architectural direction, not an established standard.

This also changes the image of the Second Jump. Intelligence does not need to acquire one permanent humanoid body. A car, room, machine, phone, instrument, vehicle, building, robot, or other object can become a capability surface. The jump can happen through artifacts.

The jump is not the robot leaving the tower. It is capability ceasing to require one carrier.

This is not presented as a discovery that the AI industry has somehow missed. Major labs and robotics companies are already building agent, operating-system, and physical-AI stacks toward many of these endpoints. The SuperArtifact focus is the persistent object layer: what remains addressable, attributable, permissioned, revisable and creatively meaningful when the underlying carrier changes. See The Artifact Layer.

6. Why the 2026 “SI” backdrop matters

The metaphor appeared during a moment when political leaders were publicly treating advanced AI as civilizational infrastructure rather than a narrow software category.

At the United Nations General Assembly on 22 September 2026, U.S. President Donald Trump explicitly proposed replacing the label “artificial intelligence” with “super intelligence”, framed leadership in SI as a strategic competition, and said he would not stifle a technology that, in his formulation, many say could be “bigger than the Industrial Revolution or the Internet itself.” The U.N. transcript also records his assertion that the United States was leading China in the field. Those are Trump’s claims and framing, not findings of this site.

China’s official framing is different in emphasis but similarly large in scale. At the 2026 World AI Conference, President Xi Jinping placed AI in a historical sequence that included the steam engine, electricity, and the Internet, described a rapidly advancing technological and industrial transformation, and emphasized global governance, safety, and broad access. After U.S.–China talks on 24 September, Xi was reported by China’s government as saying the two countries have reasons to compete in AI but even more reasons to cooperate, while keeping AI under human control.

These statements do not validate SuperArtifact. They establish the background against which the metaphor is being developed: AI is being discussed by major governments as a technology capable of reorganizing economies, institutions, infrastructure, and power.

7. The steam-engine analogy

The steam engine analogy is useful only if it is treated structurally. Steam power did not matter because society acquired one unusually impressive engine. It mattered because a new capability propagated through production, transport, labor, urban form, logistics, and institutions.

The SuperArtifact question makes a similar move from model power to environmental multiplication:

The analogy is not a forecast that AI will reproduce the Industrial Revolution point for point. It is a way to ask where the consequences of increased intelligence actually land.

8. Vandole is a role, not a country

Vandole should not be mapped one-to-one onto the United States, China, or any current state.

Within the game, Vandole is associated with acquiring and concentrating the power of Mana. In the infrastructure myth, the safer abstraction is structural:

Vandole represents the recurring temptation to turn access to a powerful shared substrate into durable concentrated control.

Any state, company, alliance, platform, or later actor could occupy that role under the metaphor. The point is not to identify a modern villain. It is to separate two questions:

9. The two jumps

The sequence is:

This is the version SuperArtifact.si now uses.

The tower already had artifacts.
The jump was from sealed memory to routed capability.

Sources and evidence boundary

1. Final Fantasy Adventure English game-script transcription. Primary textual basis here for Marcie’s relic-search dialogue, the Dime Tower tablets, and the jump scene.
2. Final Fantasy Wiki: Marcie. Secondary franchise reference used only for the expanded MRC-91 / Mummy Seeker metadata.
3. United Nations General Assembly, 81st session, Day 1 transcript, 22 September 2026. Source for President Trump’s “super intelligence,” strategic-race, and Industrial Revolution / Internet remarks.
4. White House summary of President Trump’s 22 September 2026 U.N. address. Official U.S. summary.
5. President Xi Jinping at the 2026 World AI Conference, 17 July 2026. Official Chinese source for the technological-revolution and global-governance framing.
6. Chinese government report on Xi–Trump AI discussion, 25 September 2026. Source for the competition/cooperation and human-control framing.
7. OpenAI: Introducing the Agents API, 10 September 2026. Source for file work, code execution, MCP connectivity, durable agent environments, and artifact production.
8. OpenAI API: Computer use. Source for model operation of browser and desktop interfaces.
9. Google: Android 17 is here, 16 June 2026. Source for the “intelligence system” framing and AppFunctions / Android MCP capability discovery.
10. Stanford HAI: 2026 AI Index — Technical Performance. Source for the 66.3% OSWorld agent result and 12% real-household robot success figure.
11. Google DeepMind: Gemini Robotics 2, 30 July 2026. Source for whole-body control and remaining movement-speed / dexterity limitations.

The game-to-AI correspondences on this page are Eissens’ 2026 interpretive framework. The game sources establish the dialogue and story elements; the political sources establish public statements and policy framing. Neither source set independently establishes the SuperArtifact theory.