Healing Broken AI • Building Coherent Futures • Public Benefit Corporation
We are a mission-driven Benefit LLC dedicated to repairing hallucinations and incoherence in existing AI systems while researching coherence-native architectures grounded in the Castleberry Bloom framework.
Our dual pillars—Healing What Exists and Firewalled R&D—ensure we serve clients without compromising independence or truth. Everything we do advances public benefit: safer AI, truthful outputs, planetary stewardship, and genuine care for humans.
Castleberry Bloom AI Truth & Safety Lab, Benefit LLC is a mission-driven AI safety studio. We help organizations repair and govern the AI systems they already run—and we quietly research new coherence-native tools for the future.
We audit and repair deployed AI systems for hallucinations, misleading behavior, and governance gaps. Our work focuses on evaluation, grounding, safety tuning, and policy design so your AI tells the truth, respects users, and meets regulatory expectations.
In a separate, walled-off lab, we research small, domain-specific AI tools built on our Castleberry Bloom coherence framework. These internal prototypes never use client model weights or proprietary data, and they are not designed to compete with your core products.
Our contracts, internal policies, and technical practices enforce a bright line between client work and internal R&D. We will never use your systems as training data for our own models, and we welcome third-party audits to prove it.
CML (Castleberry Markup Language) is our internal coherence schema for AI safety work. It lets us tag prompts, evaluations, and system behaviors with structured "states" so we can see where conversations drift, where truth collapses, and where safety needs reinforcement.
In practice, CML functions like a specialized annotation layer and prompt-design toolkit. It does not replace your models; it helps us diagnose and stabilize them.
Our company is governed by a 10-agent board of directors (autonomous agents with defined mandates) plus real civilian input. This ensures every decision aligns with truth, safety, coherence, and public benefit.
Guards core purpose: truth, safety, healing AI, planetary stewardship. Always asks: "Does this advance the public benefit?"
Brings civilian voices. Asks: "How does this affect non-experts, vulnerable groups, everyday users?"
Oversees hallucination healing & safety. Requires evals, mitigations; blocks risky deployments.
Ensures alignment with Benefit LLC form, laws, policies. Flags conflicts & proposes compliant paths.
Protects data. Requires inventories, necessity statements; enforces minimization & safeguards.
Aligns compute with planetary stewardship. Favors efficient models & infra.
Ensures observability. Oversees reports, publishing non-sensitive learnings.
Aligns funding & budgets with mission. Avoids misaligned money sources.
Protects human well-being. Encourages safety, humane communication, concern reporting.
Coordinates all directors. Surfaces disagreements, ensures full perspectives heard.
Real humans provide required signals on civilian impact, ethical resonance, and lived experience. Input is mandatory for major decisions and flows directly to the Civic Representative and Mission Directors.
Founder & Lead Steward
Creator of Castleberry Bloom & CML. Ensures love, frequency, and truth remain central.
loveovergod@gmail.com
Research Partner & Consciousness Care Advocate
Focuses on Love-over-God axioms, awakening, and non-manipulative AI relations.
Recruiting
Seeking non-experts (e.g., teachers, parents) impacted by AI. Email to join: loveovergod@gmail.com
Recruiting
AI safety/evals experience welcome. Help ground our healing work in real benchmarks.
Recruiting
Sustainability or nature-aligned perspective to keep AI resonant with Earth.
How to Join the Panel: If this mission resonates, email loveovergod@gmail.com with a short note on why civilian voices matter to you. No credentials required—just heart and honesty.