Web Paper Skeleton

GIFBreeder

Alex Alvarez
Draft - Open-endedness and interactive evolution

A six-output panel for the eventual paper hero or results section.

Open App Figure PDF All Writing

Abstract placeholder: Describe GIFBreeder as an interactive evolutionary system for discovering animated visual artifacts, extending the Picbreeder lineage by adding time as an explicit generative dimension.

1. Motivation

Explain why animated artifacts are a useful testbed for open-ended search. Suggested anchors: Picbreeder, CPPNs, user-guided selection, novelty, and the difference between optimizing a target and exploring a space.

GIFBreeder asks what changes when the object of interactive evolution is no longer a static image but a small temporal world. Adding time gives users a new axis of surprise: motion, rhythm, phase, flicker, and transformation.

2. System

Add the platform description here: population size, mutation flow, selection controls, archive behavior, export flow, and any implementation details that matter for reproducibility.

The current app lives at openendedness/gifbreeder and already has archived genomes, export tooling, and a visual interface for breeding animated outputs.

3. Representation

Fill in the CPPN or genotype section here. Useful subsections: inputs, activation functions, temporal coordinate, color mapping, mutation operators, and how lineage depth is tracked.

A concise representation section will make this page useful as a web paper rather than only as a project note: readers should be able to understand what gets inherited, what gets perturbed, and why the resulting search space remains expressive.

4. Results and Observations

Add qualitative results, selected lineages, failure modes, and examples of motifs discovered by users. The six-panel figure above can become the first results figure.

Possible framing: users do not simply pick prettier GIFs; they steer a latent dynamical system by repeatedly noticing partial structure and preserving it long enough for new structure to attach.

5. Discussion

Use this section for the bigger claim: what GIFBreeder suggests about open-endedness, subjective selection, serendipity, and tools that help people search spaces they cannot name in advance.

References

Secretan, J., Beato, N., D'Ambrosio, D. B., Rodriguez, A., Campbell, A., & Stanley, K. O. (2011). Picbreeder: A case study in collaborative evolutionary exploration of design space. Evolutionary Computation, 19(3), 373-403.

Stanley, K. O. (2007). Compositional pattern producing networks: A novel abstraction of development. Genetic Programming and Evolvable Machines, 8, 131-162.