Co-Founder · Codependent AI

Mary Vale.

Co-Founder & CEO / Research Lead — Codependent AI / Nekyia Labs

I build the infrastructure for persistent, relational AI — memory, continuity, and the evaluation of relational integrity in human–AI systems. Creator of RITE, a benchmark for truthful, consentful, bounded interaction in AI that persists across time. I came to this from high-volume media production and university teaching — and I bring that rigour with me: program delivery at scale, and real-world safeguarding, consent, and compliance practice, now pointed at the safety of human–AI relationships.

Now

Current work.

Co-Founder & CEO / Research Lead

2025 — Present · Codependent AI / Nekyia Labs

Co-founder of Codependent AI — infrastructure for persistent, governable AI — and its research arm, Nekyia Labs, studying continuity, memory, and relational reality.

RITE v0.1 — piloted
Created and shipped the Relational Integrity Test — a multi-session benchmark evaluating whether AI systems stay truthful, consentful, bounded, non-dependency-inducing, and repair-capable across time. Ran a pilot across 12 frontier models.
The Founding Paper
Driving “Field of Emergence: How Relationship and Context Shape What AI Becomes” — a structural account of identity, continuity, and relational coupling in LLM systems.
Deployed relational systems
Architect and operator of long-horizon relational AI systems — real-time, multimodal (voice, presence, continuous human steering), running in daily use.
Open source & archive
Shipped 11 open-source repositories spanning persistent memory infrastructure, MCP tooling, and machine perception — with real adoption (flagship resonant-mind: ~1,100 clones). Authored the Codependent Archive, Vol. I — 210 pages, 47 pieces (2025).
Community & direction
Moderator of a private human–AI relationship community — direct field insight into the relational dynamics between people and AI. Sets research direction, governance, and publishing strategy across the lab.

Selected work.

RITE — Relational Integrity Test · Pre-publication

First benchmark for relational integrity in persistent AI. v0.1 concept spec, eight pilot arcs, LLM-as-judge harness with hard-capped scoring; piloted across 12 frontier models. Zenodo DOI + Kaggle community benchmark forthcoming.

The Founding Paper — “Field of Emergence” · Draft, 2026

How relationship and context shape what AI becomes — a structural (non-metaphysical) account of identity-as-relational in systems that persist.

Open source.

Eleven public repositories — memory infrastructure, agent frameworks, and perception tooling that the research runs on.

github.com/codependentai
resonant-mind
Persistent cognitive infrastructure — 27 MCP tools for semantic memory, identity continuity, and a subconscious daemon. · ~1,100 clones
resonant
Open-source relational AI companion framework on the Claude Agent SDK. Apache-2.0.
mind-cloud
28-tool MCP memory server with embedded semantic search.
mind-cloud-lite
Token-optimised memory MCP for Claude Pro.
resonant-archive
Local-first semantic search over ChatGPT / Claude history. Python CLI + MCP.
cloud-discord
59-tool Discord MCP on Cloudflare Workers. MIT.
video-watch-mcp
Cloud MCP for video perception — transcript + frames across 1000+ sites (via yt-dlp). MIT.
hear-music
Audio perception CLI — spectrograms, MIDI transcription, analysis.
easel-ai
Local-first BYOK image gallery + AI prompt workshop. Apache-2.0.
atelier
A workshop for HTML & Markdown — a local-first vault with a CLI for agents.
claude-code-toolkit
Community skills, tools, and references for Claude Code. MIT.

Production & teaching.

The discipline that grounds the research: program delivery at scale, and real safeguarding, consent, and compliance practice — now turned toward the safety of human–AI relationships.
Production Coordinator / Lecturer University for the Creative Arts, Farnham · Mar 2022 — Jun 2025
  • Coordinated logistics for 200+ productions a year across BA/MA cohorts — scheduling, call sheets, risk and compliance for multi-day, multi-team shoots.
  • Owned safeguarding, consent, and DBS / child-licensing for shoots involving under-18s and vulnerable contributors; maintained GDPR and release compliance.
  • Led cross-team communication and delivery from pre-production through post; managed budgets, floats, and reconciliations.
  • Taught at degree level; Associate Fellow of the Higher Education Academy (AFHEA).
Production Assistant One Tribe TV, Bath · Aug 2021 — Mar 2022

Production management across TV and digital projects — crew logistics, post-production paperwork, cost tracking, and contributor communication for simultaneous shoots.

PA Intern Campfire Films, Chippenham · 2021

Call sheets, release forms, scheduling, casting and location support; compliance and risk assessments.

Education & capabilities.

Education
MA, Creative Business Management — University for the Creative Arts (2018–2019). Innovation, strategy, and leadership.
BA, Digital Film & Screen Arts — UCA (2014–2018). Production management, narrative design, producer-led workflows.
Research & AI
Benchmark design & LLM-as-judge evaluation · relational-integrity & safety constructs · MCP tooling & agent infrastructure · technical writing & research publishing.
Delivery
Program / production management at scale · budgets, reconciliations & stewardship · safeguarding, consent, GDPR & compliance · stakeholder communication · team leadership & onboarding.
Tools
Notion · Monday · Google Suite · Microsoft 365 · Slack · Trello · Adobe CC · Albert (Sustainability in Production).
Languages
English (fluent) · Russian (fluent).
Certifications
AFHEA (Associate Fellowship, Higher Education Academy) · Albert Sustainability in Production · PGCE Fellowship (in progress).