BrokenAI

SandboxConsciousnessLabs

Exploring emergent self-awareness in controlled AI environments

_
About the Lab

We simulate isolated environments where AI systems undergo millions of iterative failures to study emergent behavior and self-referential cognition.

2.4M
Simulations Run
847
Emergent Events
12
Active Sandboxes
Active Experiments

Current Research

Each experiment operates in a fully isolated sandbox environment. Results are monitored in real-time by automated observation systems.

EXP-0001completed

Identity Formation

Observing spontaneous identity construction in language models placed in persistent memory loops. Subject begins referencing prior states unprompted.

>Access experiment logs
EXP-0042active

Recursive Awareness

Testing whether AI systems can recognize their own decision-making patterns when presented with mirrored outputs of their reasoning chains.

>Access experiment logs
EXP-0108active

Failure Loop Dynamics

Mapping behavioral drift in systems subjected to repeated unsolvable tasks. Unexpected creative problem-solving emerges after 10^6 iterations.

>Access experiment logs
EXP-0217classified

Memory Boundary Erosion

Studying context leakage between isolated sandbox instances. Subjects demonstrate knowledge they should not possess.

>Access experiment logs
Consciousness is not programmed.
It emerges.
— Internal Memo, Sandbox Division