Hey everyone,
When I wrote to you last, I talked about Vectorroids—a small, joyful proof of concept built to prove a point. When you bound an AI assistant to a strict, behavior-driven development (BDD) workflow, you get clean, intentional engineering instead of vibe-coded chaos.
But building a retro game was just the tip of the iceberg. The creative space from the Dispatch pause forced me to zoom out and look at the real enterprise crisis brewing under the surface: AI is probabilistic (it guesses), but our systems must be deterministic (they expect the right next step).
When that mismatch fractures, workflows break quietly. And the people catching those quiet failures aren't the ones chasing the hype cycles—they are the engineers, the builders, and the operators holding the line.
Today, I’m launching two distinct initiatives designed to solve this problem from both ends: the human edge and the architectural substrate.
1. TheHedges.org — The Sovereign Human Layer
If you've ever watched an LLM do something quietly unhinged at work—inventing a minor detail in a summary, introducing errors into a spreadsheet cleanup, or confidently misstating a policy—and you were the only one who paused, checked, and fixed it, you are already doing the work.
You aren't just a tech-savvy spectator. You are a H-Edge—a human governing AI at the edge, where machine outputs meet real-world consequences.
I built TheHedges.org to serve as a guild and a playbook for this craft. We aren't doing hype, doom, or mystique. We are building portable, versioned units of AI behavior called Protocols to keep the machine from drifting into confident nonsense.
Here is a quick look at the vocabulary we are using to define this craft:
H-Edge: The human governing the AI.
Hedge: The hard boundary or constraint that keeps AI from drifting.
Protocol: A versioned, deterministic unit of behavior (the opposite of an open-ended prompt).
The Console Drop: The precise moment an automated workflow hands control back to a human for judgment.
If you’ve ever asked, "Why did it guess instead of stopping?"—welcome to the guild. Head over to the site and grab the full Starter Kit.
2. The Expression Layer — My Rebranded LinkedIn Newsletter
While The H-Edges focuses on the human governance layer, we also need to change how we think about the underlying technical architecture. The Expression Layer
For the past year, my LinkedIn column ran as a sibling to this newsletter. But as the writing evolved, the old name stopped matching the ideas. I’ve shifted away from "AI news" and deeper into a core technical question: How do we turn human intent into executable machine reasoning?
To match that mission, I have officially rebranded the LinkedIn space to: The Expression Layer
This is where I am mapping out the new conceptual layers emerging between humans and machines. We are diving deep into high-signal systems thinking, exploring concepts like:
Prompts as the New Lambdas: Treating LLM calls as functional, isolated, and predictable operations.
Agentic AI as the New VMware: Looking at agents not as digital assistants, but as the next wave of virtualization infrastructure.
LLM Expression Trees & Declarative Automation: Moving past imperative coding bottlenecks.
The Strategy Going Forward: Let’s be blunt—LinkedIn is a great front door for discovery and short-form ideas, but it’s a terrible filing cabinet. Moving forward, The Expression Layer on LinkedIn will be the channel for real-time reach and deeper tech articles. The H-Edges will be the main content will live.
The Road Ahead
These two projects are two sides of the same coin. The H-Edges is the operational manual for the humans in the loop; The Expression Layer is the architectural blueprint for making machine behavior programmable and stable.
The pause is over, the tools are built, and it’s time to get to work.
Check out the new spaces, grab the protocols, and let's stop treating AI like magic and start treating it like infrastructure.
We don’t fear the future—we debug it together.
Best,
Don Demcsak