Uncommon Stories – Squint Cognition
We’re born curious. Research shows babies start processing the world at birth and interacting with it by four months — and somewhere in that early stretch, most of us pick up a habit that sticks for life. When something doesn’t make sense, we squint.
That instinct — slowing down to look harder at what we don’t yet understand — drives the team at Squint Cognition. The Waterloo-based cognitive AI startup and Catalyst Commons member builds tools that make AI decision-making transparent and accountable, especially in high-stakes fields like aviation, autonomous vehicles, and healthcare.
We caught up with Kenneth Wenger, Squint’s CTO, after he demoed the company’s work at the first Catalyst Commons Demo Day on May 27.
That instinct — slowing down to look harder at what we don’t yet understand — drives the team at Squint Cognition. The Waterloo-based cognitive AI startup and Catalyst Commons member builds tools that make AI decision-making transparent and accountable, especially in high-stakes fields like aviation, autonomous vehicles, and healthcare.
We caught up with Kenneth Wenger, Squint’s CTO, after he demoed the company’s work at the first Catalyst Commons Demo Day on May 27.
AI tools are often treated as black boxes. How does Squint Cognition’s approach change that?
Everyone in the core group that started the company comes from aviation — we like to say we have a combined 100-plus years of experience in the industry. In aviation, decisions are everything, and correct decisions are even more so. For the last 20 years, our research has focused on one question: how do you make sure you can trust what an AI system is telling you?
There are situations where you just don’t have enough information to make a decision. The right move is to say, “I don’t know enough right now — I need a moment to get more data.” AI today doesn’t work like that. It just commits. We’re changing that by building systems that can examine their own decision-making and correct it.
There are situations where you just don’t have enough information to make a decision. The right move is to say, “I don’t know enough right now — I need a moment to get more data.” AI today doesn’t work like that. It just commits. We’re changing that by building systems that can examine their own decision-making and correct it.
You demoed one of your solutions at Demo Day. Can you tell us about it?
The one we showed today is our vision platform, Squint Vision Studio. It’s an explainability tool that lets users see how their model is reading the data when it makes mistakes — and why. When the model gets something right, it shows that too. All of it feeds into a runtime monitor that evaluates every decision the AI makes in real time and works to correct the ones that go wrong.
Today’s frontier models make mistakes, but many only admit to their errors when a user prompts further. Does your solution improve on that?
Exactly. With our tools, a team can build a system that confines a model to the things it’s genuinely good at, so it doesn’t overcommit on the things it isn’t. We also provide squinting models that recognize when something is off and then decide what to do about it. One option is to loop in a human. Another is to hand the problem to a specialized algorithm built for that exact situation.
Is that the same as having a human in the loop?
In a sense — except that most human-in-the-loop setups don’t really put the human inside the automated process. They just give them oversight and ask them to decide when to step in. That takes constant attention, and at that point the human may as well do the work themselves. Our approach flips it: you don’t have to worry about anything until we tell you to.
You mentioned the team’s roots are in aviation. How does that play into what you’re developing?
We have a product line built specifically for aviation. One of the biggest expenses in bringing a new aircraft to market is the certification process — generating all the documentation and evidence to show the aircraft is fit to fly. We automate that, but in an agentic way that doesn’t rely on large language models. We use symbolic reasoning, which gives us deep traceability: we can tell you exactly where any answer came from and why it’s needed. With an LLM, ask the same question twice and you can get two different answers.
How did you get from aviation to AI?
Before Squint, I worked at another Waterloo company, CoreAVI. Around 2015, when AI started getting really interesting, everyone in aviation was asking the same question: how do we apply this to our industry? We spent years researching decision-making, and eventually we realized it needed to be its own company — one focused specifically on predictable AI.
Many companies developing AI move to Toronto or Silicon Valley. What made you stay here and choose Catalyst Commons?
Waterloo has been our home for a long time because it’s a genuine technology hub. The University of Waterloo is right here with its co-op program, which we love — the talent and the students are exceptional. And Catalyst Commons was the right space because it’s full of other founders at different stages. Every time I walk around and bump into someone new, an interesting conversation starts. I love it.
You can learn more about Squint Cognition at squintcognition.com.
Interview By Alex Kinsella
Have a story to tell? Reach out to the Commons Community Team for more info: admin@catalyst-commons.com
Interview By Alex Kinsella
Have a story to tell? Reach out to the Commons Community Team for more info: admin@catalyst-commons.com