Teknowledge Corporation

Sources: Edward Feigenbaum Wikipedia article, softwarememories.com AI memories post, search results from CHM/Brewminate/pafoster.com; primary source: Richard Hess biography and resume

WARNING: Name Collision

The domain teknowledge.com is now owned by an entirely unrelated company — TeKnowledge (capital K), a Nigeria/Africa-based global services firm founded in 2010, focused on AI, CX, and cybersecurity. It has no connection whatsoever to the original Teknowledge Corporation. Any web search for "Teknowledge" will return the wrong company. Historical references must come from archival sources.

Overview

Teknowledge Corporation was founded in July 1981 in Palo Alto, CA by approximately 20 computer scientists drawn from Stanford University, MIT, and the Rand Corporation. Its co-founders included Edward Feigenbaum — widely regarded as the "father of expert systems" and later a Turing Award laureate — Jerry Kaplan (one of the first people to hold a PhD in AI), and Frederick Hayes-Roth (whose name appears on the 1983 book Building Expert Systems, part of the Teknowledge Series in Knowledge Engineering).

The company's stated mission was to democratize knowledge engineering — to let professionals without specialized AI training build and use expert systems for commercial and industrial applications. At its peak, Teknowledge's staff reportedly represented "about one-third of the world's high-level expertise in the design and development of knowledge systems."

Teknowledge was, essentially, a commercial spinout of Stanford's AI research. It took Feigenbaum's academic work on knowledge-based systems — the intellectual tradition running through DENDRAL and MYCIN — and tried to package it into tools businesses could buy.

Products

Teknowledge's product line centered on expert system shells — tools for encoding domain expertise into rule-based inference engines without writing the engine from scratch.

Teknowledge also published Building Expert Systems (1983, Addison-Wesley) under its series imprint — one of the early practitioner-facing textbooks on the field.

Richard's Role

Richard Hess joined Teknowledge in 1985 as an Apprentice Knowledge Engineer — his first job after graduating from Rose-Hulman Institute of Technology with a BS in Mechanical Engineering (cum laude). The move to Silicon Valley to join Teknowledge changed the entire trajectory of his career.

The role was hands-on with expert systems UIs — encoding knowledge structures and working on the interface layer between human experts and the inference engine. The expert systems themselves were built in Oyster, Teknowledge's proprietary object-oriented programming system layered on top of Interlisp-D. Oyster was the in-house OOP layer that let knowledge engineers structure domain expertise as objects and rules within the Lisp environment — before C++ or Smalltalk-80 were the obvious choices for OOP on workstations.

The primary hardware environment was Xerox Lisp Machines: specifically the Dandelion (the 1981 Xerox workstation) running Interlisp-D, Xerox PARC's advanced interactive Lisp development environment. The Dandelion/Interlisp-D combination was the de facto platform for serious expert systems work — it offered a live, inspectable, graphical environment years before that was standard anywhere else. (See vault://lisp-machine for full hardware context.)

Teknowledge also had a Tektronix Smalltalk machine on hand — almost certainly the Tektronix 4404, one of the few commercial workstations of the era to ship with a full Smalltalk-80 environment out of the box. Richard had access to it and explored it alongside the Lisp Machine work. This made Teknowledge an unusually rich environment for hands-on exposure to the two dominant object-oriented programming traditions of the 1980s: Lisp-based systems (Xerox, Symbolics) and Smalltalk-80 (Xerox PARC lineage, commercialized by Tektronix and later ParkPlace Systems).

The discipline of knowledge engineering was fundamentally declarative — you weren't writing algorithms, you were modeling how human expertise actually works: encoding the rules, heuristics, and domain knowledge of a subject-matter expert into inference structures the system could reason over. This left a permanent imprint. Richard has described it as closer to mechanical systems design than conventional programming — modeling real-world constraints, not writing procedures.

Industry Context

Teknowledge was one of the "four horsemen" of the commercial expert systems sector, alongside:

All four were riding the same wave: corporate America was spending heavily on AI, expecting expert systems to automate valuable human expertise at scale. By 1985 — when Richard arrived — companies were spending over $1 billion per year on AI. Teknowledge had gone public with soaring share prices. The hype was at full intensity.

The era was also tied at the hip to specialized hardware. Expert systems ran on Lisp machines (Symbolics, LMI, Xerox) — expensive, purpose-built workstations that could cost $50,000–$100,000 each. The economics of expert systems assumed this hardware would remain the platform. It wouldn't.

Decline

The collapse came from two directions simultaneously.

The economics of knowledge engineering failed to scale. Expert systems proved extraordinarily expensive to build and maintain. The knowledge acquisition process — extracting expertise from human specialists, encoding it as rules, validating the inference — was slow and required scarce specialists. Systems that worked in narrow lab conditions made "grotesque mistakes" with unusual inputs. They couldn't learn. They required constant manual updating as the underlying domain knowledge changed.

Teknowledge's own flagship demo was reportedly a wine selection system based on approximately 40 rules — impressive as a proof of concept, but far from the broad commercial automation that had been promised.

The hardware market collapsed. In 1987, Sun Microsystems workstations and portable Lisp implementations running on commodity hardware simply outperformed Lisp machines on price-to-performance. An industry worth half a billion dollars was replaced in roughly a single year. Symbolics, LMI, and the broader Lisp machine ecosystem evaporated. The expert systems companies, whose entire stack depended on this hardware, lost their platform.

Teknowledge, along with its peers, struggled to survive. Revenues dwindled. By the end of the 1980s the market had collapsed, and the term "artificial intelligence" itself became toxic — researchers deliberately rebranded their work as "machine learning," "knowledge-based systems," or "analytics" to avoid the stigma.

What Richard Took From It

Richard has described his Teknowledge years as formative in ways that still show up in 2026. From the biography:

"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."

He watched the entire ecosystem fall in real time: Xerox Lisp Machines, Thinking Machines, Symbolics — companies that represented "the future of computing" — gone practically overnight once general-purpose hardware running C outpriced specialized AI hardware on raw performance per dollar. The scar tissue is rare. Most people working in AI today weren't born when he was writing knowledge bases at Teknowledge.

The deeper continuity is philosophical. Knowledge engineering was a declarative discipline — the same underlying pattern that shows up in his YAML-driven UI platform in 2026. Describing a system's behavior declaratively rather than writing imperative procedures; making the system's state inspectable rather than opaque; separating the specification from the execution. The machinery underneath changed completely. The problem stayed the same.

Surviving References

See Also