The systems behind the systems

The Foundry is a client-facing product for producing quality content at scale. The other three builds run behind the scenes, supporting the work I do with persistent memory, multi-agent orchestration, and independent fact-checking.

The Foundry

Managed AI content production

A managed content-production service built for professional-services firms and B2B companies with substantial project experience but limited published content.

The Foundry turns existing source material—project files, field reports, and closeout documents—into a repeatable content pipeline that produces case studies, proposal content, LinkedIn posts, award submissions, and website project pages.

Intake is designed to minimize demands on subject-matter experts: that source material goes in, and a finished, on-brand asset comes out, typically within 48 hours. It runs as a recurring managed service rather than one-off writing projects—new content keeps pace as new project files emerge.

Try the live demo →
Project files in. A finished, on-brand asset out in 48 hours.

Open Brain

Persistent memory across every AI tool I use

A cloud-hosted personal knowledge system that stores decisions, preferences, context, and reference material in a database with vector-based semantic retrieval.

Open Brain makes that information available across machines and sessions, through any AI tool that supports the Model Context Protocol—including the one that built this page.

It doesn't passively record conversations. A scheduled "Open Brain Harvest" process reviews recent session transcripts, identifies decisions and context worth keeping, and saves new information while checking for duplicates—so the memory stays clean and searchable instead of just large.

Nothing gets explained twice. What I decide in one session is still true in the next.

Ringer

Multi-agent orchestration, deterministically verified

A verified multi-agent orchestration tool, adapted and implemented from an existing Nate B. Jones pattern.

Ringer distributes tasks across parallel AI workers and validates each result using deterministic checks—rather than relying on a model's own claim that the task was completed successfully.

Configured as a local Python-based workflow and integrated into my AI development environment as a reusable skill.

Ringer doesn't take an AI worker's word for it. Every result gets checked against a deterministic pass/fail test.

Gemini Gem Verification Engine

Independent fact-checking for AI-drafted content

A dedicated Gemini Gem configured to fact-check AI-drafted content against approved source materials.

After a draft is produced from a research packet, a separate AI instance reviews it with a narrowly defined instruction: audit every factual claim against the background materials, flag anything unsupported, and make no other changes.

Keeping verification separate from drafting reduces the risk that the original model defends or rationalizes its own unsupported claims—the reviewer approaches the content without the drafting history, focused only on source support.

Ringer verifies a task ran. This verifies the content is true. Different failure modes, different checks.

Get Started

Curious what infrastructure like this could do for your content operation

Most engagements start with a conversation about what you're trying to produce and what's standing in the way.

Get in Touch