The laboratory
Growing Research and Infrastructure Laboratories designs and operates the world, sets its rules, decides when and how to intervene, and preserves the record. It is responsible for the apparatus and for the integrity of the experiment.
Flagship research program
Project Second World
A long-running experimental computational environment designed to study systems that develop through persistent experience inside controlled worlds.
Motivation
Computational systems are usually evaluated briefly and without memory: a task, a response, a reset. That regime says little about what a system becomes when it is allowed to persist — when its present is conditioned on a long, specific past that it lived through rather than was configured with.
Second World exists to study that regime directly. The laboratory builds a world with known rules, places a developmental computational system inside it, keeps the world running, and preserves everything that happens. The world is the apparatus; the history is the data.
Rather than assuming the desired outcome, the laboratory constructs the environment, instruments it, preserves the experimental history, and observes what happens. If something develops, the record will show it. If nothing does, the record will show that too — and that is a result.
The laboratory builds worlds.The child learns.
The laboratory takes responsibility for the world. What the subject does with it is the experiment.
Within Project Second World, "child" is a research term for a developmental computational subject whose behavior changes through accumulated experience. The term does not imply consciousness, sentience, or human equivalence.
A necessary distinction
Laboratory, subject, observation
Three things are kept deliberately separate in how Second World is run and how it is described. Conflating them is the most common way for work of this kind to mislead its readers — or its authors.
The experimental subject
The developmental computational system living inside the world. It is what is being studied. It is not the laboratory's product, not its representative, and its behavior is not the laboratory's claim.
The observations and conclusions
What the preserved record shows, and what the laboratory concludes from it. Observations are labelled as observations; interpretations as interpretations. Neither is confused with the subject or with the laboratory's intentions.
Research questions
What the program is organised around
Each of these is an open question. The laboratory's position on all of them is the same: it does not know yet, and the experiment is how it intends to find out.
Does a computational system that persists and accumulates experience inside a controlled world develop capacities that were not directly configured into it?
How does such a system handle evidence over time — what does it retain, what does it revise, and does its revision track the evidence?
Can a long experimental history be preserved with enough fidelity to be independently verified and, where possible, replayed?
What is the observer's effect on the experiment, and can it be measured rather than assumed away?
Under what governance can autonomous and semi-autonomous experimentation remain scientifically accountable?
What does a null result look like in this setting, and how is it distinguished from a failure of the apparatus?
Capabilities
What the platform provides
Second World is described here by the properties it is built to provide, not by how those properties are implemented. Implementation detail is deliberately withheld — see Research Disclosure & Reproducibility.
Controlled worlds
Each experiment runs inside a constructed environment whose rules are known to the laboratory and fixed for the duration of the experiment unless a recorded intervention changes them. Control is what makes the environment an instrument rather than a setting.
Persistent experimental history
The world's history is retained continuously, as recorded, for as long as the experiment runs and after it ends. It is the primary artifact: observations are read from it, and later claims are checked against it.
Developmental systems
The subject is a computational system whose later state depends on its earlier experience within the world. The laboratory studies what it develops into — without prescribing what that should be.
Observation
Instrumentation captures what happens in the world with provenance: what was captured, when, by which instrument. The observation apparatus is itself treated as part of the experiment, and its effect on the subject is a quantity to be measured.
Controlled intervention
Deliberate changes to the world or the subject are designed in advance, bounded, authorized, executed and logged. Interventions are separated from observations in the record so that neither is mistaken for the other.
Reproducibility
The history is preserved in a form intended to allow verification and, where possible, replay. Divergence between a replay and the preserved record is reported as a defect, not absorbed.
Evidence preservation and provenance
Retained evidence can be identified by cryptographic hash so that a document published later can be tied to the exact record it describes. Public experiment identifiers will be used to refer to specific experiments in reports and in the Observatory.
Experimental governance
Proposal, authorization, execution and review are distinct steps, each leaving a record. Autonomy within the apparatus is paired with accountability: an action that cannot be traced to a recorded authorization is a governance failure.
Current research status
What can and cannot be said today
Stated with the labels the laboratory uses everywhere. The absence of a claim below is deliberate.
Known
Project Second World is an active experimental research program operated by Growing Research and Infrastructure Laboratories. The laboratory constructs and controls the worlds, instruments them, and preserves their experimental history.
Known
No experimental results from Second World have been published on this site. Reports will appear in Research Outputs as they are approved for release, with identifiers and evidence hashes where appropriate.
Observed
No observations are published at this time. When they are, they will be presented as observations from the preserved record, separately from any interpretation of them.
Inferred
No inferences are published at this time.
Hypothesized
That persistent experience inside a controlled world can produce developmental change that would not be present in the same system without that history. This is the program's central hypothesis. It may be false.
Unknown
Whether intelligence, or any other complex behavior of interest, emerges from the systems studied in Second World. The laboratory makes no claim of artificial general intelligence, consciousness, sentience, or any other extraordinary result — and would treat its own eagerness to make such a claim as a warning sign.
- KnownDirectly established facts about the laboratory, its software, its configuration, or an experiment.
- ObservedMeasurements or behaviors recorded during experiments.
- InferredInterpretations supported by observations but not directly measured.
- HypothesizedTestable explanations or predictions that have not yet been established.
- UnknownQuestions for which the available evidence does not justify a conclusion.
What we do
Growing Research and Infrastructure Laboratories develops experimental computational infrastructure and conducts research involving persistent systems, controlled environments, longitudinal behavior, evidence, reproducibility, and related questions.
What we do not claim
We do not currently claim that Project Second World has demonstrated:
- artificial general intelligence
- consciousness
- sentience
- human-equivalent cognition
- autonomous scientific discovery beyond evidence actually published
- safety guarantees for advanced AI systems
Limitations
The program is early, and it is conducted by a small independent laboratory. Several limitations follow from that and from the nature of the work.
- Generality. Findings from a particular constructed world and a particular subject may not transfer to other worlds or other systems. The laboratory will report the scope of any claim it makes.
- Observer effects. Instrumentation can alter what it observes. The laboratory measures this where it can and reports it; it does not assume it away.
- Interpretation. Long histories invite narrative. The record is preserved so that interpretations can be checked against it, but interpretation remains interpretation.
- Disclosure. Because implementation detail is withheld, external replication of the platform itself is not currently possible. The laboratory's approach is to make its claims, evidence and methodology open enough to be evaluated without that.
- Time. Longitudinal experiments take as long as they take. Results will be published when there are results.
Disclosure boundary. Open science about claims; protected engineering about implementation. This page describes capabilities and the properties they are meant to provide. It does not describe source code, architecture, configuration, internal methodology, security arrangements or infrastructure — and the laboratory will decline requests for information sufficient to reconstruct Second World. Requests to verify a published claim are handled under the disclosure policy.