Richard Hess — March 2026 (age 69)
It all started with a dog-eared paperback read as a teenager in the Midwest: VOR, James Blish's 1958 tale of an alien living nuclear reactor crash-landed in the Arctic. From there the path was almost inevitable — Arthur C. Clarke's 2001 (read before Kubrick's film, which matters: the ideas landed on their own terms before the film's imagery colonized the imagination), Ray Bradbury's The Illustrated Man, George Stewart's quiet masterpiece Earth Abides. Later, the dense, footnote-heavy original Japanese editions of Shirow's Appleseed and Ghost in the Shell — actively sought out after discovering Shirow's artwork while working in Silicon Valley in the 90s, before the Western translation infrastructure existed and before the '95 Oshii film made GitS a cultural touchstone.
Those books wired something deep: a taste for genuinely alien systems, hard science under pressure, and the fragile boundary between human and machine. Reading also extended to Joe Maberry's Joe Ledger series, Greg Bear, and Linda Nagata's Red trilogy — the most rigorous treatment of autonomous weapons and AI-mediated warfare written by someone who actually thought through the implications. Decades later, that same sensibility shows up in everything built.
The career began in 1985 at Teknowledge as an Apprentice Knowledge Engineer on expert systems — the first commercial AI wave. Cut his teeth on Xerox Lisp Machines, the Dandelion, and Interlisp-D's live, inspectable environment. Then the bubble burst. Xerox Lisp Machines, Thinking Machines, Symbolics — companies that represented "the future of computing" — vanished overnight once it became clear that general-purpose hardware running C was going to outprice specialized AI hardware on raw performance per dollar. He watched the whole pantheon fall in real time.
That experience left permanent calibration. In his own words: "I'm extremely sensitive to the hype that exists today regarding AI, since I saw firsthand how expert systems failed for the same reason — over-promising and under-delivering." Most people working in AI today weren't born when he was writing knowledge bases at Teknowledge. The scar tissue is rare.
There's also a philosophical continuity that runs from that era to today. Knowledge engineering was itself a declarative discipline — you weren't writing algorithms, you were modeling how human expertise actually works and encoding it into inference structures. That's closer to mechanical systems design than conventional programming. And his mechanical engineering degree (Rose-Hulman, cum laude) gave him something most software people never develop: a deep intuitive understanding of systems under constraint, tolerances, how components interact physically, what failure modes look like in the real world. The move to Silicon Valley to join Teknowledge changed the entire direction of his career — but the underlying mental model never changed. The current YAML-based UI specification work is the same problem as 1985 knowledge representation, just with dramatically more powerful machinery underneath.
After the AI winter, he pivoted hard into raw UI work and never looked back. The 1985–2004 stretch reads like a greatest-hits album of computing history. At Consilium he negotiated a four-day work week to spend two days at SGI maintaining their customized Motif toolkit — giving insider depth on X11 immediately applied back at the day job. He contributed X11 extensions to the classic maze demo (his name — cimshop!rhess@uunet.uu.net — still sits in the FreeBSD man page). He patched and temporarily shepherded the libXpm project when its original team lost funding, keeping the entire X11 icon and image ecosystem alive until they could take it back. He shipped code to WINTERP, the OSF/Motif Lisp interpreter that let developers treat widgets as live objects. And during his full-time SGI years he landed an NDA with id Software to port Quake 3 Arena to IRIX — shipping official binaries that ran in SGI's massive wall demo room. The special-effects team at Weta Digital in Australia even emailed thanks: after rendering Gollum all day, they'd fire up Quake 3 on their IRIX workstations to unwind.
From there the throughline never broke. He rode the browser wars at Netscape, built rich internet apps at Sapias (custom maps before Google Maps existed), and spent a decade at Demandbase evolving through Flex, AngularJS, and React/Redux. Every stop refined the same question: how do you keep the system live, inspectable, and honest while the hardware and hype cycles change underneath you?
Then came Cox Automotive in 2019. The first five years (2019–2024) were about portfolio-scale composable UIs and web components — Lit, Stencil, design systems, portfolio-wide standards. Then in 2025–2026 it converged into something bigger: a complete declarative UI platform where a single YAML spec is the source of truth. Five specialized MCP servers work in concert: Juris turns the YAML into a reactive runtime with action–state patterns and an event bus; Tailwind injects styling on the fly; the SVG server resolves an 18k+ icon library by family::name:variant; Lit and Stencil generate production web components from the same spec. Everything is dual-protocol — full MCP for LLMs and clean REST for browsers — so agents and humans use the exact same services. The runtime is deliberately deterministic: once the YAML is fed in, there's zero probabilistic behavior left. A Sublime Text plugin even lets you hit a hotkey and see the live Proofs view (widget + full dev tooling) right from the editor.
What makes this especially striking in 2026 is that three separate high-profile efforts are now converging on essentially the same pattern shipped the year before: the official MCP Apps extensions, Google's A2UI declarative protocol, and Vercel's json-render (which only just added YAML support). Each is trying to solve the same core problem — give agents a constrained, safe way to describe rich UIs instead of raw DOM or code — but this platform already had the full runtime observability, templating engine, dual-protocol access, and deterministic execution in place a year earlier.
In early 2026 he moved on to Shield Legal LLC as Principal Architect UI/UX, carrying the same platform sensibility into legal tech.
Forty-one years later, the throughline is unmistakable. From dreaming in PostScript stack frames, maintaining libXpm through a funding gap, porting Quake 3 across endianness wars, and carrying the live-system DNA from the first AI winter all the way to a YAML-driven, MCP-orchestrated platform in 2026 — the same problem has been solved the entire time:
Make the machine meet the human halfway, stay inspectable, and never lie about what it can actually do.
"From Expert Systems & Dreaming in PostScript to MCP-Driven UI Specs — One Continuous Thread."