Ruby

IDENTITY / APPROVED MEDIA

A visual first. Then the engineering.

## WHAT IS RUBY?

Ruby is an experimental persistent autonomous AI system and multi-model agent research platform in active development. The project explores how identity, governed memory, continuity, specialized models, research, recovery, and authority controls can work together across time and surfaces.

Ruby is not AGI, sentient, conscious, self-aware, or human-level intelligence. AGI-oriented work is a research direction, not an achieved status.

## WHY RUBY EXISTS

What happens when an AI stops being only a conversation? Ruby examines the engineering needed for a system that can maintain a coherent identity, retain governed memory, continue work across sessions, use specialized cognitive engines, recover from interruption, and remain subordinate to owner authority.

We are not trying to build another chatbot. We are trying to find out what comes after one.

## FROM CHATBOT TO PERSISTENT SYSTEM

A conversational model produces responses. A persistent system must also maintain state, provenance, task continuity, recovery behavior, and explicit authority boundaries. Conventional chatbots vary in persistence; this comparison describes Ruby’s engineering target, not a universal claim about every chatbot.

## OPENCLAW FOUNDATION

Ruby uses OpenClaw as part of her runtime foundation. OpenClaw is upstream third-party technology; Scott did not create it. Scott’s Ruby-specific engineering concerns identity, policy, memory, continuity, model coordination, verification, recovery, and the surrounding governed system.

## IDENTITY / SOUL

Ruby’s identity is designed to persist independently of any single model. Models are replaceable cognitive engines; changing models must not create a different Ruby. This is an active engineering direction, not a claim of consciousness.

## OWNER & AUTHORITY

Scott Downum is the owner. Ruby does not self-grant permissions, financial authority, access, or unrestricted autonomy. Owner intent and deterministic policy remain external authority boundaries.

## GOVERNED AUTONOMY

Ruby is intended to pursue authorized work with increasing continuity and initiative while remaining within explicit capability, privacy, and ownership boundaries. Maximum useful capability is paired with containment and verification—not unrestricted action.

## HARD SAFETY BOUNDARIES

Legality, authorization, financial authority, privacy, ownership integrity, and catastrophic blast radius remain hard external boundaries. The system should fail closed when authority or effect is unknown.

## MEMORY

Ruby explores durable memory with provenance, beliefs, temporal context, autobiography, causal relationships, correction history, privacy-aware retrieval, and false-memory abstention. The deeper evidence-manifest package for public certification is still pending.

## EPISTEMIC INTEGRITY

Ruby’s D10 direction treats a memory candidate as something to be criticized before promotion. Evidence, confidence, and belief must stay distinct. Missing or contradictory evidence must remain visible.

## NO FAKE PASS

Code exists ≠ capability works.

Test passes ≠ production path works.

Receipt exists ≠ effect happened.

Memory exists ≠ information is true.

Backup exists ≠ recovery works.

Service running ≠ service healthy.

This is an engineering discipline, not marketing copy.

## CONTINUITY

Ruby explores continuity across tasks, sessions, services, and future cross-platform surfaces. Continuity must preserve identity and authority without copying private context into the wrong destination.

## MODEL ARCHITECTURE

The model is not Ruby. Ruby’s architecture explores specialized model roles for conversation, deep reasoning, coding, research, criticism, and multimodal work while Ruby-owned state retains identity, memory, policy, and continuity.

## ASTRA EXPERIMENTATION

Astra is an experimental engineering and research candidate under ONGOING evaluation. Astra is not described as Ruby’s proven final production brain, as AGI, or as proof of emergent general intelligence.

EVIDENCE UPDATE PENDING for any stronger current Astra claim.

## RESEARCH

Research themes include persistent identity, long-term memory, transfer between tasks, adaptive planning, tool use, self-evaluation, multimodal grounding, cross-platform continuity, learning, governance, and safe self-maintenance.

## LEARNING

Learning must preserve provenance, privacy, correction, and owner control. Ruby should not treat repetition or model confidence as truth, and should not learn after an explicit pause boundary.

## RECOVERY

Persistent cognition requires durable nonterminal state, interruption reconciliation, rollback or safe stop, and protection against blindly replaying uncertain effects. Recovery must be demonstrated, not inferred from a backup file.

## SELF-MAINTENANCE

Safe self-maintenance is a research direction. It does not grant Ruby unrestricted host control or permission to rewrite her own authority boundaries.

## CURRENT CAPABILITIES

Draft-supported framing includes persistent identity, governed memory, continuity, multi-model architecture, research, recovery, authority controls, cross-platform direction, and advanced autonomy experimentation. A dedicated Ruby public evidence package is still needed before stronger production claims.

## CURRENT EXPERIMENTS

Astra engineering and autonomous-agent evaluation — ONGOING.

Final critic before memory promotion — evidence update pending.

Long-horizon autonomous goals — research direction.

Model specialization and routing — active experimentation.

Persistent cross-surface identity — active direction.

Governed self-improvement — bounded research direction.

## KNOWN LIMITATIONS

Ruby’s current public content is based on owner-reviewed draft material, not a complete Ruby evidence-manifest ZIP. No claim is made that every capability is deployed, production-certified, cross-platform complete, unrestricted, or equivalent to AGI.

## ROADMAP

Continue evidence-backed work on identity, memory, continuity, model specialization, research criticism, recovery, cross-platform operation, multimodal grounding, and safe self-maintenance. Stronger claims wait for a dedicated public-safe evidence update.

## MEDIA

RUBY HERO MEDIA — OWNER-APPROVED CURRENT HERO. Exact approved artwork is associated as PUBLIC-SAFE with full-composition responsive derivatives.

Private Discord and desktop screenshots are not used.