Megan Anderson

01 · Quest Log

My missions.

The thread connecting my work is not a single industry. It is designing better systems around people: safer, more understandable, more capable, and more accountable.

01

Human-centered AI

There is no accountability without visibility. I build the layer that records what agents actually committed to — what executed, what was written, what was sent — so oversight has evidence rather than assurances.

02

Child safety

Developing the Coalition for Child Safety and drafting the Platform Accountability Ledger, aimed at the documented pattern of platforms failing to prevent harm to minors.

03

Neurodiversity

Advocacy and education grounded in lived experience. Neurodiversity-affirming practices and research. Sharing perspectives from inside the spectrum.

04

Women in Tech, AI & Business Leadership

A women’s AI group, accessible education, and hands-on learning — because the next decade of this technology should include their perspectives. AI is leverage when you have an open mind.

02 · Brands

Where I build.

Two organizations, two very different audiences, one shared bias toward useful systems, education, and measurable impact.

A World 01 · For the future of work

AI ARMY

AI strategy, architecture, training, agents, automation, and research for organizations figuring out what comes after "just add a chatbot." A governed multi-agent operating system: CORE OS and Constellation close the agent loop, so what an agent commits to is recorded, scoped, and reviewable — for solo founders through enterprises.

AI architectureAgentic systemsTrainingResearch

Visit AI ARMY ↗

World 02 · For safer communities

Kaleidoscope Support

Community-centered education, advocacy, and support at the intersection of neurodiversity and child safety — with practical pathways for families, educators, and communities. No formal diagnosis required, because access shouldn't depend on zip code, diagnosis, or financial status.

NeurodiversityChild safetyEducationAdvocacy

Visit Kaleidoscope ↗

03 · Products + Open-Source

Systems I'm building.

A connected stack for closing the agent loop. Bridging the gaps for governed AI operations — designed as components with explicit roles rather than one mysterious black box.

OS

CORE OS

The closed loop itself. Memory to authority to action to observation, and back to memory.

✦

Constellation + Atlas

Shared epistemic memory, versioned and provenance-bearing, with retrieval and packet assembly.

☾

Luna

Runtime governance. Authority granted, scoped, and revocable, with approval tiers.

◈

ANT

Routing and orchestration, with declared handoffs and return contracts.

◉

OWL

Continuous observation across layers — drift, anomalies, and governance failure.

⌘

MCP Gateway

The single governed action checkpoint. Every tool call passes one gate.

CORE OS closed loop architecture diagram — The CORE OS loopMemory to authority to action to observation, and back to memory — closing at the centre, on memory the user owns.

Constellation memory diagram from the AI ARMY research pages — Memory as a constellation; shaping the information spaceMemory objects are stored as stars or nodes; provenance links are the bridges between them, creating clusters or constellations over time. Unrelated objects or ungoverned objects remain isolated but accessible until promoted to a cluster. A packet is constructed from the information field of constellations for the right context at the right time. The memory stays home; the packet travels.

Data sovereignty is ownership plus governed sharing. A local vault, a governed cloud mirror, boundaries per object, audit per action. Your data stays yours — your agents share it anyway, on your terms, with receipts.

Compatibility-first, by design. Built over any model, across existing tools, on open protocols. No single vendor should own the ground agents stand on — which means you keep the freedom to leave, including leaving us.

Free tools for evaluating AI

MCP Directory Claude Skills Library LLM Cost Calculator AI Subscriptions AI Model Leaderboards Team Usage Calculator AEO: AI Answers Report AI Opportunity Report 1:1 AI Discovery Call

04 · Research Journal

Questions worth chasing.

My research spans agentic AI, memory and governance, cognition, time, and theoretical physics. I examine complex and dynamic systems, information theory, and neuroscience to establish unique threads across domains that relate to the broader applications of machine learning, agentic systems, and information processing dynamics within human minds and AI models.

Closing the Agent Loop

Governed operations, context, memory, authority, and accountable execution.

→

Open specs + field notes

Practical artifacts that translate research into interfaces builders can inspect, test, and reuse — including the Context Packet Specification and a pre-registered experiment.

→

Temporal Neuroscience Index

A cognitive research lane exploring temporal integration, regulation, and differentiated processing.

→

The Missing Layer of Time

A physics foundations lane focused on events, records, and operational criteria for public facts.

→

05 · Career Path

The backstory.

I didn’t arrive in AI from a straight path. I took the road less traveled by, and that has made all the difference… I arrived through years of building the operational systems underneath teams and high-growth companies. That led to adding AI to the top of the tech stack I had been connecting all along.

CHAPTER 01

Marketing + revenue systems

From designing t-shirts and ecommerce to digital marketing mastery. Years spent designing the systems behind growth across startups and enterprise organizations. That sparked a lifelong commitment to building, experimenting, and learning as technology advanced.

CHAPTER 02

Founder + systems builder

Moved from improving individual workflows to designing entire operating models — businesses, products, processes, teams, and the connective tissue between them. I founded Dulci Digital, a digital marketing agency helping businesses navigate transformation through the rapid changes of COVID, connecting strategy, systems, and execution at scale under real time pressure.

CHAPTER 03

AI architecture + education

Founded AI ARMY to help organizations move from scattered AI experiments to useful systems: strategy, agents, integration, training, governance, and practical adoption. Applying AI not as a replacement for people, but as a way to amplify human capability. A defining moment was programming an entire end-to-end application solo using AI agents. I now run the whole operation with the agent team I built.

CHAPTER 04

Research + public-interest work

Expanded into formal research and the mission work behind Kaleidoscope Support, bringing the same systems lens to accountable AI, neurodiversity education, and child safety. Independent research identified failure modes in agent memory and reasoning that enterprises were hitting at scale. That became the AI Forward Framework's eight pillars — and a research program spanning physics, cognitive science, and AI safety.

SIDE QUEST · ALWAYS ACTIVE

Writing + teaching

Turning complicated ideas into frameworks, classes, workshops, talks, books, tools, and stories that more people can actually use. At the core: AI should enhance human potential, not replace it, and the future of technology should be shaped by inclusive participation, not limited access.

Brands and organizations I've worked with

BayerMayo ClinicSKU IQ Hearst MediaRedHouse B2BQS World MBA Tour White Plume Technologies

Case study

Paid social · enterprise pharma

Bayer campaign turnaround

Brought in to rescue an underperforming paid social program. Every KPI moved — and the account saved over $1M in ad spend by driving cost per click down while engagement climbed. Click-through went from 0.16% to 9.69%.

+31,895%

engagements

+103,619%

website clicks

−99%

cost per click

4.1M

users reached

Revenue & marketing operations

SKU IQ growth

Hired as Head of Customer Success, promoted to Head of Revenue & Marketing Operations. Led the HubSpot Enterprise implementation off Zendesk, built 300+ custom automation workflows, and aligned revenue, marketing, and success around one funnel.

+57%

revenue in first six months

7×

revenue within three years

+53%

trial conversion rate YoY

85%

manual work reduced through automation

Top skills · GTM, engineering & tech stack

GTM & growth

GTM strategy, Infrastructure, System design
Lifecycle and lead routing logic
Segmentation and user journey mapping
Campaign management and email marketing
Performance marketing channels
Technical and SaaS product marketing
Revenue operations and inbound marketing
Data, Analytics, Reporting and dashboards
Branding and graphic design
Social media and content strategy

AI & engineering

AI and ML, Agents, Automation and integrations
Ontology, Orchestration, Agent harnesses and skills
Agentic frameworks, Cloud platforms, DevOps
Prompt engineering and generative AI workflows
SQL, RAG, Embeddings, Vector databases
HTML, CSS, Java, JSON, Python
APIs, MCPs, SDKs, Agent protocols
Visual Studio, Claude Code, Codex

Tech stack

HubSpot — Admin, Developer, Inbound certified
Supabase, Cloudflare, Vercel, GitHub
Zapier, Make, LangChain
Snowflake, Databox, Databricks
Anthropic Claude, OpenAI, Open-source
Model Context Protocol tooling
Adobe, Canva, Generative AI tools
Google Workspace, Zoom, Otter.ai
MS Office, MS Teams, Copilot
Jira, Asana, ClickUp, Slack, Discord

06 · Follow or Connect

Connect.

I build agent systems in production and study the problem underneath them: the point where an AI agent's reasoning becomes an action other people and systems must rely on.

Capability without accountability is power without balance. That's true of institutions, and it's true of AI systems. We cannot continue to engineer without regard for the consequences to the real lives on the other side of the screen.

aiarmy.co Research Kaleidoscope ORCID LinkedIn Field Notes on AI

Building AI for good withhuman-centered design.

From One Person Marketing Team to Solo Founder: A Unicorn’s Path from GTM to AI Systems