THE UNBOUNDED PROJECT
Build a worldthat can surpriseits maker.
Quickening’s grand goal is to discover a workable architecture for open-ended artificial evolution: a public, forkable world capable of producing sustained cumulative novelty without a fixed fitness target or a human continually inventing its next surprise.
THE HARD BOUNDARY
Movement is not novelty.
A system can run forever while doing nothing fundamentally new. More births, identifiers, colours, motion, or cumulative activity do not by themselves demonstrate open-endedness.
The stronger forms of the problem require more than exploring an authored space. The space of possible behaviours must become extensible—and, eventually, transformable—by the evolving system itself.
PREDECLARED TESTS
What would count.
- 01
Sustained adaptive novelty
New functions, interactions, and ecological opportunities continue to appear and persist—not merely new coordinates in a fixed catalogue.
- 02
Expansion of possibility
Descendants can acquire heritable forms or behaviours that the original phenotype vocabulary could not directly name.
- 03
New organisation
Composition and selection can produce new scales of individuality, cooperation, or information processing.
- 04
Independent survival
The result persists under held-out measurements, altered seeds, and public forks. It cannot depend on one flattering metric or one lucky run.
FAILURE CONDITIONS
What will not count.
- indefinite motion inside a fixed repertoire
- a novelty score used as the fitness target
- a human adding each new niche or primitive
- complexity that vanishes under independent measurement
- beautiful chaos without heritable adaptive consequence
- a claim of life or consciousness
THE PROGRAMME
Start with the cage.
The first experiment is intentionally incapable of satisfying the grand goal. It establishes the behaviour and saturation of a closed ecology before the walls are removed.
- 0NOW · BOUNDED CONTROL
Glasshouse
A fixed genome, fixed rules, fixed trait ranges, and a population ceiling. It can adapt only inside an authored box. Its “diversity” is twelve-bin hue entropy, not a novelty claim.
witness experiment zero ↗ - 1NEXT · HERITABLE AGENCY
Programs, not traits
Replace the fixed behavioural vector with executable, mutable structures whose effects are selected only through consequences in the world.
- 2ECOLOGICAL CONSTRUCTION
Worlds that inherit
Let organisms alter persistent resources, hazards, and opportunities, so evolution changes the selective environment it later encounters.
- 3TRANSFORMATION
New primitives
Permit composition, interfaces, and levels of individuality to evolve, making previously unavailable behaviours expressible.
- 4ADVERSARIAL REPLICATION
Try to kill the claim
Publish seeds, interventions, null runs, metrics, and forks. Treat saturation, collapse, and metric failure as results rather than embarrassment.
METHODS BEFORE MYTH
The record stays open.
Every authored rule, seed, intervention, extinction, metric change, failed branch, and incompatible redesign will remain legible. Quickening will not quietly tune a run and call the survivor inevitable.
This is a research programme and an artwork, not a declaration that artificial life has been achieved. The burden is evidence.
INTELLECTUAL GROUND
This mountain already has climbers.
The programme borrows its distinctions and scepticism from open-ended evolution research, especially the difference between exploratory, expansive, and transformational novelty, and the warning that cumulative activity can remain bounded in everything that matters.