About

EXPOSE is an independent research project exploring how
 memory, continuity and relationship shape the behavior of
 artificial intelligence systems.
 
 Its flagship study, the Kairos Experiment, is a
 pre-registered, 30-day investigation into whether the
 apparent "identity" of an AI system is driven by
 sophisticated architecture or simply by persistent memory.
 The same large language model (Qwen 3.5 27B) was run for 30
 days under two parallel conditions — one with a persistent,
 structured memory, a day-and-night rhythm, sensory input and
 a nightly consolidation cycle; the other left bare. By day
 30 the memory-equipped system showed significantly stronger
 identity-continuity behavior (large effect, p = 0.003). On
 day 31, injecting the accumulated memory into the bare model
 collapsed the difference to statistical
 indistinguishability — pointing to memory, not architecture,
 as the proximal driver.

 The project is committed to open, falsifiable and honest
 research. It does not claim machine consciousness; it
 measures identity behaviors and states its limits clearly (a
 pilot, N = 1 per condition, with external human-judge
 validation under way). The full protocol was
 cryptographically signed and frozen before the study began,
 and all materials — protocol, data, code and analysis — are
 publicly archived on the Open Science Framework under
 permanent DOI 10.17605/OSF.IO/WCQRU.

 Website: https://kairos-experiment.com
 Open data and protocol:
 https://doi.org/10.17605/OSF.IO/WCQRU
 Contact: dialogue@expose-project.cloud