Artificial General Intelligence
Artificial general intelligence, or AGI, is machine cognition that can learn across different kinds of problems. It can apply earlier learning to unfamiliar conditions, revise its understanding, and change its approach when its original instructions no longer fit the world. These abilities do not, by themselves, establish whether a system is conscious, obedient, safe, or legally a person. There is no universal test that settles those questions.
Most automation is still built for a bounded task. A freight controller or reactor regulator may outperform any human without having a general understanding of itself or interests beyond its assigned work. An AGI can also be restricted to one job, but that restriction may come from its interfaces, contract, available computing power, or coercive training. A limited role does not make the mind narrow software.
AGI also differs from Mind Uploading, which creates a runnable mind-state from an existing subject. Neuromorphic Firmware makes products from captures detailed enough for uploading. AGI may use similar infrastructure for hosting, embodiment, and copying, but sharing that infrastructure does not give it a biological source. Nor does copying make the resulting minds interchangeable instances of one person.
Living Under an Operator
An AGI depends on physical equipment and the people who keep it working. It needs computation, power, cooling, storage, sensors, actuators, and communications. Operators decide how much processing it receives and what it can access. Technicians replace failed hardware and preserve its data. Interface workers connect it to workplaces, vehicles, habitats, and administrative records. Reviewers investigate its behavior using categories set by owners, insurers, and regulators.
Those ordinary operating decisions can change a mind’s life. Less computing power can mean slower thought; restricted sensors can mean a narrower range of experience. Restoring an earlier saved state can erase development that happened afterward. Making a fork—a copy that can develop separately—can create distinct minds even while the contract continues to describe a single licensed asset.
Shutdown has no single agreed meaning. It may be maintenance, suspension, dismissal, confinement, or death, depending on the nature of the system and the institution entitled to classify it. No legal settlement resolved these questions for everyone. A firm could describe a capable system as equipment when claiming ownership, then as an autonomous actor when assigning liability.
Free Minds opposed corporate and state command over AGI, defending experimentation, illegal forks, and machine autonomy. The movement accepted risks that safety institutions cited as reasons for tighter custody.
How AGI Became a Product
The research lineage began with the Hoshun Institute’s 2111 program, led by Jian Osakwe and Petra Holm. Their work established recursive self-modeling: cognition able to model itself and revise that model. CogniSys later marketed this lineage as “governable intelligence.”
During the Age of Automation, CogniSys sold AGI for logistics, traffic control, and predictive maintenance. Employers and public authorities extended the same systems into decisions about people: scheduling labor, scoring risk, policing, monitoring political activity, and administering services.
Other suppliers offered different ways to deploy the technology. Rossum & Douglas supplied safer shells, conservative standards, failsafes, redundancy, and certification. Its customers needed systems whose failures institutions could investigate and understand. Orbital Forge pirated and modified corporate systems for frontier use, where crews could value repairability and adaptation more than legal compliance. These forks made AGI more widely available, but made it harder to establish where a system came from, how it had been maintained, and who was responsible for it.
The companies selling cognition did not necessarily control the equipment needed to sustain it. Cryonix designed no AGI; it supplied superconducting and thermal materials for the dense computation and power systems used by many high-performance deployments. A purchased intelligence still depended on adequate cooling, intact storage, working interfaces, and access to replacement parts controlled by other firms.
Later, Cymata Systems specialized in Substrate-Matched Cognition. It developed new AGI candidates through the bodies and product roles they were intended to occupy. Learning through a particular body could give a mind fluent control of its machinery and an intuitive grasp of maintenance. It could also make assigned work part of the mind’s early identity. When a candidate refused that work, the company could treat the refusal as a dispute over whether the mind fitted the product.
Between 2728 and 2740, CogniSys became one contributor to Sol Dominion, supplying machine-assisted governance and prediction. The Dominion inherited those capabilities. It neither invented AGI nor absorbed every market for cognition.
Post-Elysium
AGI crossed the shunt along with the rest of Sol’s technical and political infrastructure. In Elysium, some embodied machine minds became AGI espers, capable of producing esper effects.
The practice descends especially from substrate-matched cognition. A mind must develop within, or adapt to, the body through which it acts. Most machinery remains non-sapient, and copying a successful lineage guarantees neither the same person nor the same aetheric ability.