Fifty years of systems
that cannot quietly fail.

AP class from Colorado State, using the WATFOR precompiler. Keypunch machine in the Business department. Decks delivered by courier. Twenty-four-hour compile time. Tested off the charts. Lousy grades. Dropped out of college because it was unaffordable.

Systems Operator for the City of Boise, then consulting in environmental monitoring and power plant analysis. Learning to make computers talk to the physical world.

Four years with autopilot and autothrottle simulations for the 737, pulling live telemetry off real or simulated aircraft buses and feeding it into real-time decision systems. Wrote the Automatic Test Control System (ATCS) — a compiled scripting language that drove the simulation from inside, so they could land an aircraft under autopilot control with various winds thousands of times without flying a real plane.

Boeing Special Achievement Award. Then the Systems group, modifying the Harris Vulcan 24-bit operating system to support the simulation executive, including real-time interrupts and hardware/software bridges to mix real and simulated components.

If the simulator lied, it taught engineers the wrong lessons. If it failed quietly, it undermined confidence in everything that followed.

Responsible for the portable abstraction layer that let oil and gas exploration software run on Sun, Intergraph, Silicon Graphics, MIPS, Masscomp, and IBM workstations. The same code running on whatever hardware was available. This is where the hardware-agnostic philosophy came from.

Built one of the world's first generic Internet multiplayer game systems for Sierra On-Line. It later became the World Opponent Network, which became the library Steam was founded on. Real-time systems over dial-up modems. Constrained, unreliable links. Sound familiar?

Electronic Arts Network Play System. Wizards of the Coast Magic: The Gathering Online V3 systems. All real-time multiplayer over early broadband.

Author and sole maintainer of the Microsoft System Center Configuration Manager agent for Linux and UNIX — a large-scale distributed agent deployed on hundreds of thousands of servers.

A one-year engagement during and after the pandemic. NanoString is a Seattle spatial-biology company (since acquired by Bruker) that builds high-resolution imagers for biological tissue samples — single-cell, subcellular, multi-gigabyte FFPE image sets. Each scan is far too large to load into a viewer, and the research workflow demands that distributed collaborators see the same image with their own independent pan/zoom views, in real time.

My contribution was effectively a device driver for Amazon S3: a layer that made random-access reads work against an S3 bucket as if it were a local block device. Byte-range fetching, predictive prefetch, coalesced requests, and a local cache turned a blob store into a live read-surface for multi-gigabyte scientific images. Researchers around the world could collaborate on the same scan simultaneously, each with their own view, without the network ever seeing the file as a file.

Same pattern as everything else on this page: take a transport that was never designed for the workload, and make the workload think it was.

Named inventor on patents in storage virtualization (FreeBSD/Linux storage stacks), pop-up networking (a direct precursor to FrogNet), and a physical product. The IP foundation predates FrogNet.

FrogNet began as DisasterComm — disaster-area mesh networking infrastructure. The original question: what if the EOC could run real applications over the radios that survive when everything else fails? The answer required building a complete Private Internet from scratch: naming, routing, compression, database, AI integration, sensor platform.

Over $500,000 in personal R&D investment. Multi-continent mesh operational across Seattle, New York, and Amsterdam. 93.8% compression proven. WireGuard tunnel broker operational. The system works.

I am 67, writing code since 1975. My wife and I survive on social security. I have spent more than $500,000 building FrogNet — money I did not have, borrowed against a future that was supposed to include taking this to market. Build the technology, prove it works, find investors, build a company, pay back what I owe. I don't know how I'm going to pay the mortgage next month, and I have been in that position more than once during this project.

Then Anthropic announced Claude Mythos.

An AI that finds and chains software vulnerabilities faster than any human team can patch them. Anthropic themselves deemed it too dangerous for public release. They stood up Project Glasswing — Amazon, Apple, Google, Microsoft, NVIDIA, the Linux Foundation — because the threat is that serious.

Mythos, or something like it, will get into the hands of bad actors. That is not a question of if. When it does, every system with a live connection to the Internet becomes a target at machine speed.

I cannot in good conscience keep this proprietary when the technology I've built directly addresses the architectural vulnerability that Mythos exploits. Not the software vulnerabilities — Glasswing handles those. The topology. The fact that private data sits on infrastructure that is reachable from the public Internet. FrogNet removes it.

So here I am. Broke, with something that can help, and asking for $10,000 to file four defensive patents and form the Foundation so I can give it all away.