The Story of Antithesis: The Company Building the Future of Autonomous Testing

The Story of Antithesis: The Company Building the Future of Autonomous Testing

The best startups often begin with a secret weapon that nobody else knows about.

In a recent episode of Category Visionaries, Will Wilson, Co-Founder of Antithesis, an autonomous testing platform, revealed how a testing system built at a previous startup became the foundation for a company attempting to eliminate manual software testing entirely. This is the story of how that secret weapon evolved into a venture to transform how developers build software.

The Secret That Started Everything

Before Antithesis existed, there was FoundationDB—a distributed database startup that seemed to be attempting the impossible. Will and his co-founder Dave Scherer were building a database that could run across multiple computers while maintaining the reliability guarantees of a single machine.

People frequently asked them how they pulled it off. “What was our secret weapon?” Will recalled the question. The answer wasn’t what anyone expected.

“The secret weapon, basically, was that we developed an extremely powerful autonomous testing system, which would just, like, try running the database in all kinds of configurations and all kinds of possibilities and do all kinds of crazy stuff to it,” Will explained.

This wasn’t ordinary testing. The system found bugs autonomously, at scale, faster than any human team could. “By doing this really fast in parallel, it kind of found all the bugs, not literally 100%, but it found the vast majority of bugs that were ever introduced into that product,” Will shared.

The impact was transformative. “It made developing it really easy and really pleasant because it’s like when you’ve got this amazing robot assistant that just tells you whenever you’re about to make a mistake, hey, it’s actually a lot easier to write code.”

The Radicalizing Experience

When Apple acquired FoundationDB in 2015, Will and the team went to work on projects they can’t discuss. But one thing became clear: they didn’t have their secret weapon anymore.

“Suffice it to say, they did not have this amazing robot assistant for some of the projects that were looking at,” Will noted.

Will eventually moved to Google, and that’s when the full realization hit. “I saw that there were lots and lots of really smart people there working on really hard problems without the benefit of powerful tools that could make them radically more productive,” he explained.

The contrast was jarring. “These are the biggest, richest, coolest companies on earth, and they don’t have this technology that our teeny tiny startup did. Like, why is that?” Will asked. “That seems crazy. That seems like a business opportunity.”

It was the radicalizing moment that led to Antithesis.

Building From First Principles

Starting Antithesis in 2018 meant starting from scratch—literally. What they were building required rewriting fundamental assumptions about how computers work.

“In order to build what we’ve built here, we had to go back to the very foundations of how computers work and change a whole lot of stuff,” Will explained. “We had to write our own hypervisor that is able to run virtual machines in a completely different way than anybody else ever has.”

This deep technical work meant years of development before they could ship anything. “It took a few years, and that’s a lot longer than most startups,” Will acknowledged. But they couldn’t rush it. Building a hypervisor from scratch isn’t something you can shortcut.

Despite the long build, they pushed to get early validation as quickly as possible. “We did actually get early design partners and paying customers as early as we possibly could,” Will shared. “But for us, as early as possible was still a few years into the thing.”

The Magical Moment

Product-market fit often arrives not with fanfare but with quiet realization. For Antithesis, it came in two waves.

First, their early customers became increasingly dependent on it. “We did start to notice that our customers, like, loved it, couldn’t get enough out of it. And, like, you know, we’re putting more and more of their eggs in this basket,” Will recalled.

But the real validation came when the team started using Antithesis to build Antithesis itself. “There was a magical moment when we started using it, too, and we started feeling it, like, really giving us lift and making us go faster,” Will shared.

Here’s what made it meaningful: they were still using an early, rough version. “And that was despite the fact that at this stage, what we still had was still a very early prototype, really rough around the edges, really crappy in a lot of ways,” Will explained.

The insight was powerful: “If this very primitive, very rudimentary implementation of this thing that we’re building is already seriously paying rent for us and for some really big, important companies, we may actually have something here.”

The Long Stealth

With product validation in hand, most startups would launch publicly and start scaling. Antithesis did something different.

“We were in stealth for an incredibly long time,” Will explained. “Even after we got that first paying customer, we actually stayed in stealth for, like, another three years, just, like, making the thing better, iterating on it, like, slowly growing our customer base through word of mouth.”

The extended stealth served a dual purpose: it gave them time to refine the product while building mystique. “We did try to, like, cultivate, like, an aura of secrecy,” Will shared. “We sort of tried to make it, like, a badly kept secret that many people in Silicon Valley whispered about and that when you learned about it, you felt like you were sort of being given access to a special club.”

The strategy worked. “That actually worked super well. That made it sort of irresistible to a lot of people,” Will noted.

Choosing the Battlefield

When you build technology that can test virtually anything, the temptation is to sell to everyone. Antithesis resisted.

“Autonomous testing is an insanely broad category, and the technology we’ve built fundamentally could be used to test anything,” Will explained. They had tested Nintendo games during development. They could test websites, smartphone apps, backend systems—the possibilities were endless.

“But, like, we didn’t want to go to market trying to sell the sun and the earth and the moon and the universe,” Will said. “We wanted to try and be really focused and pick one particular area to target.”

They chose reliability and fault tolerance testing for big backend systems. The reasoning was strategic on multiple levels. “One, our founding team has a background in that stuff. Two, everybody’s kind of got one. And so it’s a pretty large niche, as niches go. And then three, the existing tooling and existing solutions for this problem are insanely bad.”

The decision to start where the bar was lowest and their expertise was highest has proven correct. “In terms of keeping our go to market energy focused on a particular customer profile and keeping our product focused on a very particular, you know, problem domain, it’s really paid off,” Will reflected.

Creating a Category From Nothing

Antithesis isn’t just building a product—they’re defining an entirely new market category. And Will is honest about how hard that is.

“We basically believe that the category doesn’t exist yet,” he explained. When investors ask about competitors, “the answer is we don’t think there are any, but people think we’re saying that. And that’s like, haha, look how great we are. No, no. That’s like a really scary answer.”

Why? Because competitors actually help. “Having competitors is great. Like if you have competitors, that’s like evidence that somebody else thinks this problem is worth solving. That’s like other people who are out there educating the marketplace about what you’re doing and like creating a budget category for what you’re doing,” Will explained.

“Having zero competitors, having a brand new greenfield area is like terrifying,” he concluded. “It does mean that the potential upside is really big, but I actually think it makes the problem strictly harder.”

They’re defining autonomous testing as testing where “human beings are not really involved in writing tests at all. All they’re doing is saying what their software is supposed to do and not do. And the computers are doing everything about coming up with the tests and executing them and analyzing the results.”

While tiny niches like fuzzing and property-based testing exist, “99.99% of the world has never heard of it or done it. And that’s what we’re trying to change,” Will shared.

The Future: A World Without Manual Testing

Looking three to five years ahead, Will’s vision is both ambitious and straightforward.

“The big picture is that almost all manual human software testing effort is completely eliminated and developers can produce software much more quickly,” he explained. “They produce much higher quality software that works better in more situations, and everybody has more fun doing it.”

It’s a future where the secret weapon that made FoundationDB possible becomes available to every developer. Where the debugging and testing that Will described as “one of the most time consuming, most taxing, and, honestly, most unpleasant parts of the whole process” largely disappears.

The journey from FoundationDB’s secret weapon to category-defining company hasn’t been quick or easy. It’s required rebuilding computing infrastructure from scratch, years in stealth, brutal honesty with investors, and the patience to let revolutionary technology prove itself.

But for Will and the Antithesis team, the vision is worth it. They’ve seen what’s possible when developers have an “amazing robot assistant” watching for mistakes. Now they’re building a world where every developer has one.