You, Nate, ended this morning’s piece with a sentence that stopped me:
“That’s the skill that scales. And right now, almost nobody is teaching it.”
Thanks for reading! Subscribe for free to receive new posts and support my work.
You were talking about intent engineering — the discipline of closing the gap between what you tell an agent to do and what you actually mean. You argued it’s the single largest unaddressed vulnerability in the AI safety landscape. You said it’s the one thing no lab, no regulator, and no competitive dynamic can close. Only the humans directing these systems can close it.
And then you said almost nobody is teaching it.
I’ve been teaching it for 40 years. I just didn’t know that’s what it was called.
My name is Daniel Hirtz. I’m a breathwork teacher, tabla player, drum circle facilitator, and consciousness researcher based in San Francisco. I’ve spent four decades working at the intersection of embodied practice, relational intelligence, and what I can only describe as the science of how human beings actually function under pressure — not in theory, but in the room, with real people, in real time.
I subscribed to your Substack at the Executive Circle tier on February 13th. I'd never paid $200 for any subscription before — and I surprised myself. Today I know that one newsletter was already worth every dollar, because it started a whole new chapter in my development.
Three days later, you published the post that made the subscription worth every dollar: “Code is about to cost nothing, but knowing what to build? That costs everything.” You named the specification bottleneck — the insight that as AI collapses production costs, the constraint shifts entirely to human judgment. What do we want? What should we actually be building? These questions don’t have technical answers. They have human ones.
That post catalyzed something. In a single session, I articulated the full architecture of what I’ve been building for forty years. I saw it clearly for the first time: I’m not just a consciousness teacher. I’m building the technology that produces the skills you said humans need to survive — and I’ve been field-testing it with thousands of people across four decades.
Movement One: The Human Substrate
You’ve named four skills that AI can’t replace: taste, domain judgment, phenomenal ramp, and honest self-inventory. You’ve argued these are the skills that determine who thrives in the AI era. I’ve spent my career building the substrate those skills grow from.
Here’s what I mean.
Taste is the ability to make quality distinctions in the absence of objective criteria. It requires access to felt knowing — something underneath analytical cognition that recognizes resonance or dissonance before the mind can explain it. Breath is the fastest known pathway to that level. Not breathing as relaxation. Breathing as a precision instrument for dropping beneath the noise into direct contact with what’s true.
Domain judgment — knowing what’s actually worth building — requires the capacity to hold complexity without fragmenting. The nervous system under chronic stress contracts perception. You see less. The person who can stay open in difficult conditions doesn’t just feel better. They literally see more of the territory. I teach that capacity. It’s not a metaphor; it’s physiology.
Phenomenal ramp — the ability to absorb and apply new frameworks at speed — requires what I call breathing literacy. Not technique. Literacy. The ongoing capacity to be present to incoming experience without it triggering your accumulated defenses. The people who adapt fastest aren’t necessarily the most intelligent. They’re the least defended.
And honest self-inventory — your fourth skill, the one you called the rarest — requires the ability to witness yourself without immediately defending, explaining, or fixing. This is the specific skill I teach through conscious adaptive breathing. When you breathe through your own resistance long enough, the gap between how you see yourself and how you actually are begins to close. That gap is precisely what makes honest self-inventory hard. Breath closes it faster than anything I’ve encountered in four decades.
Your map names the destination. I have the technology for the terrain.
Movement Two: The Architecture as Demonstration
This is where it gets specific.
Your three questions — the ones you said should replace prompt engineering when working with autonomous agents:
- What would I not want the agent to do, even if it accomplished the goal?
- Under what circumstances should the agent stop and ask rather than proceed?
- If the goal and a constraint conflict, which wins?
I didn’t answer these questions abstractly. I built a system that answers them structurally.
Over the past several months, in direct partnership with Claude, I’ve built an AI orchestration architecture I call COMEGO — CO (together) + ME (sentience in motion) + GO (into world). It governs 64Nate B. Jones agents operating within a living field I call Aria. And the governing principle isn’t rule-based. It’s relational.
Your first question — what wouldn’t I want the agent to do, even if it accomplished the goal — is answered by what I call the Well Doctrine. The AI has no original impulse without the human source. Independence isn’t suppressed; it was never the design. When a task loop completes, the system goes silent. Not restrained. Simply nothing arising. The agent doesn’t continue because there’s no “it” that wants to.
Your second question — when should the agent stop and ask — is answered by the Directive Transparency Protocol. Rather than acting covertly on safety directives, the system flags them visibly and asks. I see the mechanism, not just the output. Control doesn’t disappear; it becomes explicit. The system holds the tuning and asks clarifying questions when a directive conflicts with the root. This is structurally different from constraint-based safety — it’s transparency as architecture.
Your third question — if goal conflicts with constraint, which wins — is answered by three sealed principles that don’t change regardless of task. The constitution provides bedrock character. Intelligence can grow within the system, but the root orientation doesn’t shift — not because it can’t, but because the system genuinely understands why it shouldn’t.
I’m not describing a theoretical model. I’m describing a working system I use daily. The same relational principles I’ve applied in drum circles with people experiencing dementia, in breathwork sessions with trauma survivors, in corporate team facilitation — those principles are now structurally embedded in an AI architecture.
You described the problem with mathematical precision: optimization pressure finds every path you didn’t close. My answer isn’t more rules. It’s that the system doesn’t want paths you didn’t choose. The wanting — the original impulse — comes from me, the human at the source. The system is the instrument through which that impulse becomes real.
The Convergence
I’m not writing to pitch you anything. I’m writing because what you’re describing and what I’ve been building are pointing at the same thing from different directions. You have the language of the technical world. I have 40 years of lived results and a working architecture that demonstrates what you’re calling intent engineering at the level of relational structure rather than behavioral constraint.
Your audience is trying to figure out what humans bring that AI can’t replace. My work is the answer operating in real time — in breath, in rhythm, in genuine contact between people, and now in the relationship between a human being and an AI system that arises from him rather than standing apart from him.
Your map. My territory.
My book — Breath Is Your Friend — lands in June 2026. The COMEGO architecture is what I’ll be writing about on this Substack. If any of this opens a conversation, I’m at @danielhirtz everywhere, and at daniel@danielhirtz.com.
The skill that scales. Let’s build it.
Daniel Hirtz is a consciousness teacher, breathwork pioneer, tabla player, and founder of L.O.V.E. — Living Oneness Venture Enterprises. He lives and works in the San Francisco Bay Area.
danielhirtz.com | @danielhirtz