An architect who decided the drawing should draw itself
Clifton Harness did not set out to write software. He set out to make buildings better. The son of a Texas real estate developer, he studied architecture at the University of Texas at Austin in part out of disagreement with what his own father built. He kept asking a version of the same question: why can't we have better buildings? He never got an answer he liked, so he went to work at the problem himself.
The first job taught him something uncomfortable. Starting in 2015 at StreetLights Residential in Dallas, Harness spent roughly two years doing site plan test fits full time - drawing apartment units, laying out parking, counting stalls, then doing it again when a number changed. It was slow, repetitive, and error-prone. The architect, he noticed, was almost always the slowest person in the room. Finance needed five minutes to run its math. Construction needed a few days. Design needed weeks.
So he called his college roommate. Ryan Griege was a software developer who had been coding since junior high and had come up through game development. Harness asked a simple question: could you build a program that does this faster? The first version was not impressive to look at. It had no user interface at all - a command-line tool that only solved apartment unit mixes. But it solved a real problem, and that was the point.
From a command line to a category
That bare-bones utility became TestFit. The two co-founders expanded it from unit mixes into parking counts, then into full site plan generation, then into a real estate feasibility platform that developers, architects, engineers and contractors now use to test whether a piece of land is worth pursuing. The pitch is blunt: what used to take weeks can take seconds. A developer and an architect can sit on a Zoom call and prototype a deal live, changing assumptions and watching the building reshape itself in real time.
Harness is careful about how the automation behaves. Plenty of companies sold generative design as a magic button - press it and receive fifty options to sift through. He built TestFit to do the opposite. It returns one editable solution, on the theory that designers are intelligent people who resent being reduced to clicking through a filter. "Designers are really intelligent," he has said, "and if you dumb down their experience with software to filtering solutions, you won't have much buy-in." Users are expected to work with the tool, not just consume its output. As he puts it, "you have to work a little bit."
Getting paid for the work you gave away
One of his recurring arguments has less to do with algorithms and more to do with money. Feasibility studies, in the traditional model, are often done for free or for a token fee, with the architect hoping to recover the loss later if they win the full design commission. Harness thinks that is a bad trade. He encourages architects to think more like lawyers - to value their time and to hand developers real data, not just a beautiful roll of trace paper. "It's impressive when you can roll up trace paper," he has said. "It's even more impressive when you can give the developer the data they need to do their job."
He is also refreshingly willing to puncture his own industry's hype. He has publicly called generative design overhyped even while shipping one of its more commercially successful implementations, framing TestFit's contribution modestly: "Autodesk has done a great job educating the market about generative design for years; we are simply the first ones to execute on that promise for commodity buildings." It is a characteristic move - contrarian on the surface, pragmatic underneath.
The founder mindset, out loud
By 2022, the frustration that started as a favor from a roommate had turned into a company with real money behind it. TestFit announced a $20 million Series A led by Parkway Venture Capital, bringing its total funding to about $22 million. In 2023 it was named Startup of the Year at the Venture Dallas Summit. Harness has since become a fixture on the AEC conference and podcast circuit, appearing on shows like TRXL, The ConTechCrew and Construction Brothers, and presenting TestFit's approach at NXT BLD 2024.
He talks about the founder life without much varnish. In a widely shared set of reflections, he described founders as people who work punishing hours by choice, who keep going against a brutal failure rate, and who are driven by something internal rather than external reward. "No carrot. No stick. Just manic intent," he wrote. His summary of the whole enterprise is almost a mission statement: "All it takes is a founder to will a different future into existence."
The through-line from the frustrated young architect to the software CEO is consistent. He wanted better buildings. He decided the fastest route was not another rendering but a better tool - one that gives the people who shape our cities faster answers, harder data, and a little more leverage over the economics that decide what actually gets built. Whether or not that makes buildings prettier, it has already made the work quicker, and it has given commodity real estate something it rarely had before: software worth using.