Who built this, and why you should trust him with $999.

Where I come from

I started as the rookie on a twenty rep sales floor at Southwest Office Systems in Fort Worth and finished the year ranked first, with Rookie of the Year and a promotion to Senior Account Executive. Over the next four years I took a $1M territory to $4M, almost entirely through outside sales I sourced myself: cold calls, referrals, and a lot of parking lots. That is where I learned that the product is only half the sale. The other half is whether the buyer trusts the person standing in front of them.

In 2021 I joined American Airlines to sell corporate travel software into a market that was still recovering from the pandemic, with every travel budget under a microscope. I finished my first year ranked first out of roughly 100 account executives at 133 percent of quota and was promoted to Senior Account Executive.

Most recently I sold aviation safety software at Brazos, entering a new market with zero inbound leads and building a $5M plus pipeline from scratch. The deal I am proudest of was an EMS helicopter operator with no data driven view of flight risk. It took six months, a Chief Safety Officer who became an internal champion, and a business case the CFO signed. Roughly $1M in annual recurring revenue, the largest win of my career, and it ended the year at 150 percent of quota as the number one senior AE on the team.

President's Club three times. Number one at four different companies. I say this not to brag but so you know the person asking you for $999 has spent a career being held to a number and hitting it.

Why a sales rep is building a server

A company restructuring in January 2026 gave me time I had not planned on. Instead of jumping into the next quota, I started building. First came QuotaBuster, a set of commission and quota tools for sales reps that is live today. Then HiredHands, supervised AI workers built in n8n that handle prospecting, research, and bookkeeping for small businesses. Then OverwatchOS, the control layer that coordinates those workers against a revenue objective and asks a human before it does anything it is not authorized to do.

Building those systems taught me three things fast. The meter never stops running. The model you tuned your workflow around can be deprecated on someone else's schedule. And the moment a client's documents, prompts, and embeddings live inside a vendor's cloud, the client no longer really controls them, no matter what the privacy policy says.

I wanted a machine I could put on a desk, plug into a network, and point every agent at. Open weight models. Local retrieval. Persistent memory that stays in the building. An OpenAI compatible endpoint so nothing I had already built needed to change. No subscription, no seat license, no meter. When I could not buy that machine as a finished product, I decided to build it.

The honest part

I am not a hardware engineer by trade, and I am not going to pretend otherwise. Here is how the Node gets built and what you should know before you reserve one.

• The silicon is proven. The Node is built on AMD's Ryzen AI 9 HX 370 platform with 64 GB of LPDDR5X, the same processor shipping today in mini PCs and laptops from several manufacturers. I did not design a chip. I chose a platform with a track record and a mature Linux, ROCm, and llama.cpp story, and built the product around it.
• The enclosure and thermals are a design, not yet a prototype. The images on this site are concept renders and are labeled as such. Dimensions, weight, and fan behavior are production targets until the first engineering unit is built and thermally tested. When that unit exists, its photographs replace the renders and its measurements replace the targets.
• The numbers on the benchmarks page come from comparable HX 370 machines, cited to their sources. They are the best available preview of what the Node will do. I will publish measured results from an engineering unit before any Node ships.
• Production begins when all 100 first run slots are reserved. Serials are assigned in reservation order. Delivery is targeted at 13 to 14 weeks after production begins. If that timeline changes, the build log is where you will read about it first, with a date on it.
• You can reach me directly. Not a support queue. My email is on this page, and I answer it.

What I believe

Data sovereignty is not a feature you toggle. It is the sum of the decisions a system lets you make: where your data lives, where processing happens, which models run, which services can reach in, and whether the cloud is involved at all. A vendor promise is not ownership. A cloud account is not sovereignty. If you want to read the long version, it is in the data sovereignty guide.

I also believe small teams and individuals deserve the same control that large enterprises negotiate for in their contracts. That is the whole reason the Node costs $999 once instead of a monthly meter that never stops.