David Kelly
Austreng

Generic Software, Inc.

Washington · 1986 – 1987

The lesson arrived early, and it arrived as a price tag.

Austreng joined Generic Software as systems and operations manager in 1986, the year he completed his computer science coursework at Seattle Pacific University, and built the software the company used to run itself — accounting, sales, shipping, inventory and customer support — on R:BASE System V.

The company’s product was Generic CADD, and it sold for $99. Computer-aided design at the time meant a package costing thousands of dollars and a workstation to run it on. Generic CADD carried drafting out of the licensed specialist’s office and placed it in the hands of anyone with a PC and a hundred dollars — and Autodesk, which owned the professional standard being reimagined, bought the company on 7 March 1989 for approximately $7.6 million. Austreng had moved on two years before that transaction closed.

He left an idea behind him. He proposed building the product’s object libraries on the model of the Staedtler plastic drafting template — the physical stencil a draftsman lays over paper and traces standard shapes through — so that a drawing would be assembled from ready-made parts rather than constructed from nothing. That is the interaction model diagramming software went on to adopt wholesale: Visio, which shipped in November 1992, is built on stencils of reusable shapes and was named Shapeware before it was named Visio. The resemblance is offered here as resemblance. The year is offered as fact, and the year is 1986.

He had watched, from inside the operations function, what happens to an entire profession the moment a specialist capability becomes something anyone can simply buy. He has been looking for that moment ever since.

He fulfilled every computer science degree requirement and attended commencement in 1986. One religious studies course carried by transfer students remained open that year, because he spent it at his father’s side through his father’s final illness. The computer science coursework stands completed. So does the choice, and he would make it again.

Truvel Corporation

Torrance, California · 1987 – 1988

(Technology acquired by Vidar; subsequently acquired by 3D Systems)

Color came next, and he was the one holding it.

Truvel was five people when Austreng arrived as Engineering Projects Coordinator, and he wrote every piece of interface software the company shipped for as long as he was there. Truvel built among the earliest CCD-based high-resolution digitizing image scanners and held US patents 4,937,663 and 4,910,601 on rotating multi-segment filter color capture and digitizing scanning apparatus. During his tenure the company produced TrueScan — the first professional color scanner for the Macintosh — shown in Truvel’s booth at Macworld Expo San Francisco in January 1988 and launched into Europe at CeBIT Hannover that March at Apple’s invitation.

Consider what that meant. The Macintosh gained color capability with the Macintosh II in April 1987, and its built-in color was eight-bit. Truvel was capturing photographic color and needed a way to show it. So Austreng built the display himself, upstairs in the lab: a working 24-bit photographic image on a Macintosh, assembled by daisy-chaining four video cards — one carrying the standard eight-bit display, and three more driving red, green and blue independently. He turned 256 colors into 16,800,000 hues to a pixel — true photographic presentation, and the first such presentation known to have been achieved. A consultant, John Zigrang, worked the chaining problem with him. It was running roughly six months ahead of the first announced 24-bit card for the platform. Keith Sorenson founded RasterOps that year and shared their intention to deliver one; the RasterOps ColorBoard 104 began shipping in volume to retail customers in May 1988.^{[1, 2]}

Austreng’s display stood in the same booth as the scanner at Macworld Expo San Francisco in January 1988 — the show where the photographic color stunned its attendants. He was brought into Apple’s invitation-only “Jackson Pollock” forum by Laurie Girand, who had helped him obtain the video registers from Apple that made his photographic invention possible. Jackson Pollock was Apple’s own internal code name, chosen for the painter’s splashy color, for the release that became 32-Bit QuickDraw and gave the Macintosh native photographic color in May 1989.

The work also made him a source. Aldus had built TIFF in 1986 to carry scanner output in black and white; it added uncompressed RGB color in April 1987 and palette color in August 1988, and while that was being worked out Steve Carlsen — the Aldus engineer who created the format and is credited as its author on the published specification — telephoned Austreng to understand how Truvel was handling color space, because a color TIFF had to carry what the scanners were actually producing. The format is still in use today.

He gave the Macintosh a way to see before the platform had one of its own — built upstairs, out of four cards — and then he was in the room when Apple made it native.

Truvel’s line reached well beyond graphic arts — X-ray, 35mm transparency and three-dimensional capture, alongside the flatbed and drum work — and it was on the company’s X-ray scanner that he first encountered the public-domain medical imaging code he would later build a commercial product upon. It is worth marking that 3D capture was on his bench in 1987; the delivery half of that problem is what he returned to eight years later at Net VR, and again thirty years after that.

He left to do precisely that, and Greg Minter, a colleague from the same company, left with him.

Vidar Systems of Herndon, Virginia acquired Truvel’s scanner technology in October 1990 and built a business on it. Twenty-two years later, in January 2012, 3D Systems acquired Vidar for $135.5 million in cash — a price paid for engineering first done in a Torrance lab in 1987. Truvel itself reorganized under Chapter 11 in June 1990, two years after this tenure closed; that is the ordinary arc of a hardware company that arrives ahead of its market, and it is a comment on timing rather than on craft. The work outlived its maker by twenty-two years and two owners. Few products of that era can say the same.

Blue Solutions, Inc.

(Formerly Orange Solutions, Inc.)

Thousand Oaks, California · 1988 – 1991

Then he built the product himself.

PhotoPress was an early Macintosh application for emulating the darkroom on screen — dodging, burning, tonal correction and color separation, performed with a mouse instead of under an enlarger, at a moment when the hardware to display the result properly was still rare. Austreng founded the company, served as its chairman and chief executive, and wrote the whole of the application himself in Object Pascal. Greg Minter wrote the manual.

The origin of it was a crossing between two worlds that did not yet talk to each other. Dr. Wayne Rasband, at the National Institutes of Health, had written imaging software for scientists reading X-ray and MRI plates and placed it in the public domain — real, Macintosh-native, written in Pascal, first released in 1987. Austreng had met that class of software from the other side, on Truvel’s X-ray scanner, and saw what the two ends had in common: the operations a radiologist uses to pull detail out of a plate are the operations a photographer uses to pull detail out of a negative, and the second market was orders of magnitude larger than the first. He was the one who carried the code across, wrote the application on top of it, and shipped it in August 1988. Building on it was exactly what the public domain exists to invite, and the language is the tell that the lineage is genuine: NIH Image was Pascal, Object Pascal was its Macintosh dialect, and an Object Pascal codebase is precisely what that inheritance produces.

The company was founded and first traded as Orange Solutions, incorporated in California, as Blue Solutions was after it. At a trade show, the chief executive of an established company holding that name came to the booth and asked for it. Austreng handed it over on the spot. The logo was fine work and expensive work, so he kept the design and rebuilt the identity in a new name and a new color. Blue Solutions is Orange Solutions wearing a better coat.

PhotoPress shipped from Macworld Expo Boston in August 1988, and the dates around it are worth setting down plainly, because they are all public and all settled. The first public release of Thomas Knoll’s code, in any form, was version 0.65 bundled with Barneyscan slide scanners as Barneyscan XP — March 1989, seven months later. The earliest date any published account places on Adobe’s involvement is September 1988, a handshake with Fred Mitchell described in Jeff Schewe’s history of the program; final agreements followed in April 1989. Adobe Photoshop 1.0 shipped in February 1990, eighteen months after PhotoPress. The two were applications of the same kind, made for the same work, for the same machine. The comparison here rests on two ship dates — and two ship dates are all it needs.

The darkroom moved onto the screen in August 1988 — eighteen months before Photoshop — and it has stayed there ever since.

Computer Image Technology, Inc. (Dimension4)

Seattle · September 1991 – 1994

At Computer Image Technology he took on a harder problem: teaching a computer to read a drawing.

A scanned engineering drawing is a picture of geometry rather than geometry itself — a CAD system can display one, and editing it requires the geometry underneath. As chairman and chief executive, Austreng built a semi-automated raster-to-vector conversion on non-uniform rational B-spline mathematics to recover exactly that. In 1993 he drove the acquisition and merger of CATCO, extending the company into the microfiche data-acquisition market.

That same year the MIT Enterprise Forum of the Northwest selected the company as its case-study presentation, under the title Keeping the Wheels On — an honest description of the challenge, because demand was abundant throughout.

He stepped away in 1994 to get well. A series of surgeries corrected a congenital autosomal chromosome disorder, and a hospitalization beginning that February had his full attention by June. He states it plainly here so the record is complete and the reason is known, in his own words rather than anyone’s assumption.

He got well. He came back.

Packard Bell Corporation

Interface Division · 1994 – 1995

He came back to the highest-volume consumer computer channel in the United States.

As director of the Advanced Technology Group, Austreng created the object-oriented analysis for Welcome Home — the shell interface that shipped on Packard Bell machines and presented the computer as a house, its rooms corresponding to the things a person might actually want to do. Packard Bell’s buyer was, more often than not, someone purchasing a first computer and opening the box for the first time. Welcome Home is an early and commercially serious answer to the question that has occupied him since 1986: how do you hand a specialist tool to somebody who has better things to do than become a specialist?

Net VR Corporation

1995 – 1998

In 1995 he put photographs in motion inside a web browser.

Net VR, which Austreng founded and led as chairman and chief executive, built one of the first systems to deliver interactive photographic object rotation across the open web — a product or a property photographed from every angle and reassembled in the browser so that a viewer could turn it in their hands. The method was filed as an international application under the Patent Cooperation Treaty and published in 1998 as WO1998015920A1. That publication is the enduring asset: a government-issued document that fixes the date permanently. The company put its resources into the market rather than into national-phase prosecution, and the published record stands today exactly as filed. A date, once fixed, stays fixed.

Apple’s QuickTime VR reached object movies first, in 1995. The distinction at Net VR was delivery. The system was built to run across ordinary browsers on ordinary connections using nothing but what the visitor already had, at a time when that was a demanding constraint — and the company sold and deployed directly with national apparel and outdoor retail brands at the worldwide executive level.

Every property tour you have ever clicked through, and every product you have ever spun around on a retail page, is that interaction. Within a decade it had become the standard mechanism for real-estate virtual tours and e-commerce product spins, and the same interaction model is the one that street-level photographic navigation later made familiar through mapping products. That is offered as resemblance and sequence. The claim made here is narrower and stronger: he was doing it first, and there is a dated publication to prove it.

IBM Corporation

Software Group · 1998 – May 2005

Seven years inside IBM’s Software Group, in three roles, at the altitude where standards are written rather than followed.

As strategic alliances executive he led alliance work on the Java initiative between Sun Microsystems and IBM. As market manager for WebSphere Business Components he owned the market for IBM’s reusable enterprise component middleware — and was elected chairman of the Reusable Asset Specification effort, the work of defining how enterprise software assets are packaged so that teams who did not build them can find them, trust them and use them again.

That specification is still in force. The Object Management Group adopted it in May 2004, promulgated version 2.2 in November 2005, and it remains published at that version today, more than twenty years later, in a catalogue that also holds UML and BPMN. The portfolio has since changed hands — the EDM Council acquired OMG’s standards in a transaction that closed on 24 November 2025 and has since renamed itself the EDM Association — and the standard carried through unchanged. He was elected to chair a specification that has outlasted the division that produced it, the decade that produced it, and the standards body that adopted it.

As technology manager for NexGen he supported dot-com companies west of the Mississippi and owned ServerProven certification. In that period he built a complete enterprise call center application in three days using IBM’s confidential automated Java code generation technology out of the Boston laboratory. He was selected for private 3D market segmentation education delivered by a specialist group brought in from London. He is an IBM stock grant recipient and an IBM retiree.

Image Technology Laboratories, Inc.

Founded Seattle, WA (1994) · Headquarters: Vancouver, WA (since 2005) · 1994 – Present

Image Technology Laboratories has been a Washington corporation in continuous good standing since January 1994 — thirty-two years — with Austreng as founder and president. It is a research and development company, and it is a separate legal entity from Intelligence Technology Laboratories, LLC.

It is also his laboratory. It holds a body of invention work held in confidence pending patent filings, and its current programs are in machine vision for robotics and in virtual reality — the same problem he took up in 1995, returned to with thirty years of hardware progress underneath it.

What appears here about that program is deliberate and complete: its existence, its fields, and the reason for its discretion. Disclosure discipline is what preserves patent rights — publicly describing an unfiled invention starts a twelve-month clock in the United States, and the many jurisdictions applying an absolute novelty standard reward the inventor who waits. A document circulated to investors is a public disclosure. He protects the invention by protecting its timing, and shares specifics with a counterparty under a non-disclosure agreement.

Intelligence Technology Laboratories, LLC

2025 – Present

Agentic AI is the sixth act, and the thesis is the one he formed in 1986.

Intelligence Technology Laboratories builds and deploys agentic AI systems for small and mid-sized businesses across sales, operations, workflow and customer-facing processes — platforms spanning machine learning, natural language processing, computer vision and the internet of things, together with the implementation consulting that determines whether any of it survives contact with an actual business. Austreng owns strategy, solution architecture, agent and workflow design, and client implementation.

The thesis is narrow and worth stating plainly. The companies with the most to gain from agentic AI are the ones best served when someone else brings the packaging to them. Adoption is therefore a problem of packaging, and packaging is a problem he has solved five times. That is the $99 CAD package. That is the darkroom on a screen, the photograph that turns inside a browser, and the component standard that let strangers reuse each other’s work. It is the same move, for the sixth time, made by someone who has now made it in five technologies before this one.

He is the author of Checkmate — AI vs The Labor Market (2025), on labor-market displacement and enterprise AI adoption.

Four decades of being early. Every one of them the same bet: take what only specialists can do, and make it ordinary.