The Montréal and Kitchener-Waterloo deep learning research lab Maluuba was a frontrunner in Canada’s AI research ecosystem when it was acquired by Microsoft in 2017, years before the AI gold rush.
In 2026, many of the deep learning problems the team worked on now underpin the large language models (LLMs) that fuel the inescapable AI boom. Now, two of the co-founders are back with another early bet on a slightly different frontier technology. This time, Sam Pasupalak and Kaheer Suleman are developing AI world models—which are meant to simulate the physical world—through Toronto- and San Francisco-based Skyfall AI. They hope to prove that the technology can perform all of a company’s operational work with few, or maybe even no, humans required.
“Our vision is to democratize the technology so that every business can run as autonomously as they can.”
This week, Skyfall announced that its first act will be to acquire an existing business for up to $1 million USD and install an AI CEO. With backing from investors including Inovia Capital and Garage Capital, Skyfall wants to use the AI world models to take over a company and run it completely autonomously, thereby demonstrating “enterprise super intelligence.” It’s currently shopping for “micro-B2B-SaaS” businesses, and its goal is to double the acquired company’s revenue in six months.
“Our vision is to democratize the technology so that every business can run as autonomously as they can,” Pasupalak said in an interview with BetaKit.
It’s a radical departure from Maluuba’s original mission to make machines understand human language. Founded in 2011 after the co-founders went to the University of Waterloo and worked in the Velocity Garage, Maluuba was advised by Canadian AI pioneers and Turing Award winners Richard Sutton and Yoshua Bengio. The company developed conversational AI technology and sought to teach machines how to understand natural language. It partnered with consumer electronics giants Samsung, LG, and BlackBerry and built a demo to query an entire Harry Potter book the way one might query the entire internet via ChatGPT today.
RELATED: Maluuba team explains why language is the key to making machines intelligent
Part of the bet is a roguish stance against the dominance of those very AI models. Pasupalak wants to prove that world models are superior to LLMs on several traits important for human business leaders, including long-horizon planning and high-stakes decision-making with uncertainty.
“Sam’s obsession is with real-world applicability rather than static benchmarks that reward brute force approaches,” Karamdeep Nijjar, partner at Inovia Capital, wrote in a LinkedIn post.
Though Skyfall brands itself as a frontier neolab, and Palupasak compared it to the early days of OpenAI, the company is a for-profit business. “We do think a technology breakthrough is absolutely required in order to unlock this vision, and that’s why you have to have a research-first organization for the first few years until you can have a product that can be deployed at a mass scale,” he told BetaKit. Part of the research process will be publicly documenting the company’s progress in taking over companies’ operations with AI.
As opposed to LLMs, which are trained on vast amounts of text data, world models are trained on physical, spatial, and movement data about the environment. This is like building an internal 3D representation of a world, which proponents argue allows the models to integrate cause and effect instead of just pattern matching. These models have recently gained traction: Silicon Valley’s world-model startup General Intuition was recently valued at $2.3 billion USD ($3.2 billion CAD), with a dataset of video-gameplay clips that it’s using for robots and other physical AI.
The founder said his team was partly inspired by a paper he co-authored last year, in which they studied the limitations of LLMs in playing the ‘90s-era business simulator game RollerCoaster Tycoon. Should the tech work how they want it to, world model-trained AI agents will run the operational side of companies, from finance to product to marketing. But Pasupalak acknowledges that there are human skills that even world models cannot replicate, like managing the human relationships inherent to businesses.
RELATED: What Maluuba beating Ms. Pac-Man means for AI research
Similar to other company takeover models such as private equity buyouts, this structure will have an impact on human jobs. Pasupalak said that its first acquisition target would be either owner-operated or have one to five contractors or employees who would be provided with a “generous departure package” and support with finding a new job within their network. He added that this would apply to larger companies which Skyfall plans to acquire down the road.
Skyfall’s experiment, where the machine takes over human roles, appears to directly correspond with fears expressed by labour market experts and the broader public about job losses stemming from AI integration. Last year, big tech leaders like Anthropic’s Dario Amodei and Meta’s Mark Zuckerberg warned of mass white-collar job loss due to AI, but have since softened their stances on an imminent labour apocalypse. At the same time, Quartz reports that economists are growing more concerned about jobs disappearing because of the technology: more than 200 economists and researchers signed an open letter warning of “large-scale job displacement” and a reshaping of the economy.
In Pasupalak’s view, having AI handle operational work will eventually allow people to do artistic and creative work. “Humans should work on things they’re very passionate about, and we don’t want humans to do the boring, monotonous, operational tasks,” he said. The CEO compared AI to how historical innovations like the printing press eliminated the need for scribes and moved their work elsewhere (though some studies have suggested that AI might actually intensify work, rather than reduce it).
If the bet pays off at Skyfall—a James Bond-inspired name that is also a word play on secret AI agents—Pasupalak argues that its total addressable market will be much vaster than the current AI economy, which he says is largely focused on automating software engineering. Since his company is seeking to automate all parts of every business on Earth, he claimed, the opportunity “is in the trillions.”
Feature image courtesy Skyfall AI.