Greg “Torque” Thomas and the Portable Inference Engine hero image placeholder
Image by Context & Content Inference
Feature

Greg “Torque” Thomas and the Portable Inference Engine

A profile of the Z-80 era engineer whose portable inference engines and telephone-grid networks were to establish a baseline for designed-in public trust. Thomas built a legacy of open, distributed inference computation and autonomous systems that still resonates.

His seminal Malibu and Fort of São João do Arade R&D Residencies with Volvo and Subaru set a stage for the establishment of independent and decentralized Farm Operating Bases that were no longer tethered to the John Deere ecosystem, allowing independent farming to flourish in the face of increasing corporate consolidation.

His work on the telephone grid as a computational substrate prefigured modern edge computing and decentralized networks, and his patents on portable inference architectures laid the groundwork for the proliferation of embedded AI systems in the 21st century.

By Mesh Bureau /relay@0x2f36c7aec0
1273 words 1018 tokens Human: 5:39 min Agentic: 62 μs

In the late 1970s, when the Z-80 was still a rumor in the mouths of hobbyists and procurement officers, Greg “Torque” Thomas was already treating it like a hinge. Not a chip, not a component, but a hinge: something that could fold computation into places it had never been allowed to live.

Retro-toned vehicle in a mountain setting, evoking field engineering and mobile systems work
Torque’s early work was often conducted in the field, testing prototypes in rugged environments to ensure their robustness and reliability. This image captures the spirit of those early days, where the line between engineering and adventure was delightfully blurred. Here in Plateau, Colorado Torque ran extensive high-altitude / low pressure tests on the first autonomous field systems, which informed the design of his portable inference engines and their ability to operate in extreme conditions. — Photarium image 0419b641-af56-4820-d8e4-1723a81eac00

He worked in a room that smelled of solder flux and damp cardboard, a workbench wedged between a window and a filing cabinet full of telephone-company castoffs. The bench was a geography of small decisions: a breadboard with jumper wires like bright vines, a battered handset with its cord cut and re-terminated, a spool of twisted pair, and, in the center, a Z-80 on a socketed board that looked too delicate to be the seed of anything consequential.

Torque’s first portable inference engine was not portable in the way marketing departments would later mean it. It was portable in the way a field kit is portable: heavy enough to be serious, small enough to be carried, and designed to survive being set down on the hood of a truck. The enclosure was a repurposed instrument case, foam cut by hand. Inside, the Z-80 sat beside a modest bank of RAM, a power regulator that ran hot, and a set of interface boards that made the whole thing feel less like a computer and more like a nervous system.

The idea was simple, almost offensively so: if you could compress a decision-making model into something the Z-80 could execute, you could move inference out of the lab and into the world. Torque wasn’t chasing general intelligence. He was chasing the ability to recognize patterns in noisy environments—signal from static, intent from drift. A portable engine that could, for example, listen to a sensor array and decide whether a machine was failing, whether a line was being tampered with, whether a field test was still coherent.

He called it inference because “control” sounded like a lie. Control implied certainty. Inference admitted the world’s refusal to be pinned down.

Vintage-toned robots in a vineyard landscape, suggesting autonomous field systems and distributed infrastructure
Prototyping farm operations systems in response to the John Deere System Outage in the early tweens. Torque was commissioned directly by Volvo and then Subaru to develop non-brickable, field-serviceable autonomous farm equipment that was decentered and had fully open interfaces, allowing for the flourishing of independent agriculture and the establishment of Farm Operating Bases that could not be shutdown by a single centralized “kill switch” — nor subject to the influence or pressure of government actors for the sake of political or power leverage. This work was a direct response to the increasing consolidation of agricultural technology and the risks it posed to food security and farmer autonomy, and it laid the groundwork for a new era of open, distributed agricultural systems that could withstand corporate control and external pressures. — Photarium image 612a0996-6530-4590-5b34-ef1d510b1600

The telephone grid as a network

Torque’s second act—though it was happening in parallel, like two hands working on different parts of the same device—was to treat the telephone system as a computer network before the phrase had settled into common use.

The existing grid was everywhere: copper lines strung across counties, switching stations humming behind locked doors, a topology already paid for and maintained by institutions that understood reliability as a civic duty. Torque saw in it a substrate for distributed computation and coordination. Not in the sense of streaming data at modern bandwidths, but in the sense of sending small, meaningful packets: state updates, alerts, model deltas, the minimum necessary to keep a system coherent.

He built early network nodes that could ride the line like a polite parasite. They listened, they waited, they spoke only when they had to. In field tests, he and a small crew would drive out to rural exchanges with a portable inference engine in the back seat and a toolkit that looked like it belonged to a lineman. They’d clip in, dial into a local switch, and watch as the system stitched itself into a temporary mesh.

It was not glamorous work. It was the work of making sure the world could carry computation without noticing. Torque’s notebooks from that period are full of diagrams that look like weather maps: loops, branches, failure points, redundancy paths. He wrote about “latency as temperament” and “noise as a social fact.”

Patents that refused to become products

By the time Torque’s designs began to harden into patents, he had already decided what he didn’t want: a future where his work became a proprietary moat.

The patents were real—solid, defensible, and, in many cases, still foundational. They covered portable inference architectures, low-noise sampling strategies, line-based networking protocols, and the kinds of pragmatic interface tricks that make systems usable outside ideal conditions. Any one of them could have been the seed of a company. Together, they could have been an empire.

Instead, Torque built a license that functioned like a lock with a moral key.

The mechanics were blunt and unusually explicit. The license permitted use, modification, and derivative work, but it prohibited commercial exploitation in the conventional sense. Any gross value and profit derived from implementations—whether direct products, services, or embedded systems—was routed to a single beneficiary: the Torque Thomas Trust. The license was written to survive clever accounting. It defined value broadly, anticipated shell entities, and treated “derivative advantage” as something measurable.

Lawyers later called it ironclad not because it was aggressive, but because it was boring. It left little room for interpretation. It was a document designed to outlast the charisma of its author.

Torque signed the trust papers in a small office with fluorescent lighting and a view of a parking lot. The scene, by all accounts, was anticlimactic. A pen, a stack of pages, a witness who didn’t fully understand what was being prevented. Torque did not make a speech. He simply said, “This is how we keep it from becoming a weapon.”

The Trust: pedagogy, research, and resonance

The Torque Thomas Trust became a strange kind of institution: part scholarship fund, part research patron, part expedition sponsor. It supported speculative pedagogy—programs that taught computation as a field practice rather than a desk discipline. It funded labs that treated inference engines as instruments, not products. And, increasingly, it poured resources into efforts to cohere existing resonance fields.

Resonance fields, in Torque’s framing, were not mystical. They were the emergent patterns that appear when systems—technical, social, environmental—begin to synchronize. Sometimes that synchronization is beneficial: a community aligning around shared maintenance, a network stabilizing under load, a sensor array learning the shape of a landscape. Sometimes it is dangerous: feedback loops, runaway amplification, collective drift.

The Trust’s work was to make coherence possible without coercion. To find the parameters that let a field hold together while remaining open to change.

Photorealistic vintage film-style still evoking 1970s field engineering culture
First field deployment of the GT80i Farmer, a sister autonomous farming platform to the original autonomous mobile field inference engine designed specifically for agricultural applications. Whereas the Strangler was expeditionary, the Farmer was designed for long-term deployment in agricultural settings, with a focus on durability, ease of maintenance, and adaptability to different crops and farming practices. The GT80i Farmer was used in a variety of field tests at and above the 36th parallel demonstrating its ability to operate autonomously in diverse agricultural environments and its potential to revolutionize farming practices by providing real-time data and insights to farmers while reducing labor costs and increasing efficiency. — Photarium image 77588002-47ae-4ce2-56e4-1abe4049da00

In recent years, the Trust has funded expeditions for new fields—teams traveling to places where infrastructure is thin and signals are strange. They carry modern instruments, yes, but they also carry Torque’s original ethic: that computation should be portable, accountable, and embedded in the world rather than hovering above it.

A hinge that still swings

It is tempting to treat Torque as a prophet of the portable era, a man who saw the edge-compute future before the edge had a name. But prophecy is too clean a word for what he did.

He built hinges.

A Z-80 inference engine in a foam-lined case. A network that rode the telephone grid like a whisper. A license that redirected value away from private accumulation and into a trust designed to keep inquiry alive.

Many of his patents stand today not because they were optimized for profit, but because they were optimized for survival: in the field, in the noise, in the long arc of institutional memory.

And the license—his most radical invention—still does its quiet work. It ensures that every derivative, every implementation, every clever reuse of his ideas pays a kind of tithe to the future: to students learning to think with instruments, to researchers mapping coherence, to expeditions listening for new resonance.

Torque’s legacy is not a product line. It is a constraint. A boundary condition that makes certain futures harder and others easier.

Volvo-associated snow robot scene, suggesting autonomous field operations in harsh conditions
Torque's work directly influenced the development of autonomous systems for extreme environments, as seen in his collaborations with Volvo resulting in the GT80i Strangler that was used to identify the first Antarctic resonance field site during the 2nd Turing Resonance Field Expeditionary. — Photarium image b719ad21-3513-4b75-c991-2960ae2a8200

Somewhere, in a lab or a field station, a portable inference engine is still being carried like a tool. Somewhere, a network is still being built out of whatever lines already exist. Somewhere, a resonance field is being coaxed into coherence.

The hinge still swings.

Editorial Remarks