Practitioner first. Guide second. Guru never.

I use AI and agent systems every day: building notation systems, designing agent harnesses, and running real workflows. The teaching comes through the building, not the other way around.

What I'm Building

I find patterns that produce structure.

When the tooling I need does not exist, I build it. Caret^ is an open source semantic notation for agent workflows in Markdown. It replaces the paragraphs of repeated instructions that clog up every agent skill with compact markers that do the same job.

It is infrastructure I needed and could not find.

On an external benchmark against Garry Tan's GStack, Caret^ delivered a 95.8% average token reduction. That is measurable cost savings on real agent skills.

Most prompt systems try to solve chaos by adding mass. Caret^ solves it by adding shape.

I'm also building CapacityOS, a custom agent-based executive assistant with a modular engine, canon, and runtime. I use it daily to process hundreds of tasks per day.

CapacityOS is my 4th generation personal-agent-harness design that I'm now open sourcing as a modular package. It includes:

  1. A chief-of-staff agent that routes work to other agents
  2. A domain agent module that manages backlogs and evaluations with persistent memory
  3. Structured agentic workflows like an overnight factory run & self-improving capabilities.
Why I Do Sessions

This started with people close to me.

I would sit down with a friend or family member who was curious or skeptical about AI. In thirty minutes, they would go from "I don't get it" to seeing something they could use tomorrow. Not hype. Not theory. A real thing that saved them time or made something easier.

I can explain AI to a retiree on Tuesday and design an agent harness architecture on Wednesday. Those are not different kinds of work. The depth is what makes the simplicity possible.

Helping a company improve a workflow can be useful. But I would rather help Bill become "the AI guy" at his company and see his career build momentum.

Most people do not need a tutorial, a course, or a generic workflow. They need someone who listens. Who makes an effort to understand their specific situation.

You may not have started at all. You may be customizing agent harnesses. I will meet you where you are to find your fastest path to something that saves you time and hassle.

This is why individual sessions matter most to me. It is the most socially fulfilling work I do. If you are ready to find where AI actually fits, book a session.

Where I Come From

Operator first. Advisor second.

For almost a decade I've been working at startups focused on decentralized governance and operational design. This included designing system architecture and operational processes supporting machine learning for fraud detection.

Your average AI guru does not have experience with:

  1. The real-world tradeoffs created by non-deterministic systems.
  2. The architectural understanding of machine learning operations.
  3. Data privacy and bias control for RLHF & HITL labeling systems.
8
Companies launched or operated
$400M+
Combined valuation exposure
100+
Startups and clients advised

I have spent years building businesses, working inside messy real-world constraints, and figuring out how to make things actually move. The credentials are real. But I lead with work, not with a resume.

The Bigger Picture

Useful work funds principled work.

My individual consulting is socially fulfilling. Working with teams to solve real problems keeps me sharp and close to the edge of what works. Together, both paths support the larger thing I care about: improving open, legitimacy-driven systems.

Disruption Joe is focused on serving individuals wanting a jump start in their AI journey, but I am available for situations requiring deeper expertise when the fit is right.

If your team is:

Or simply needs help getting unstuck, then we should talk.