Get ready for a mind-boggling revelation: Simular's AI agent is here to revolutionize the way we interact with our computers!
Simular, a trailblazing startup, has secured a whopping $21.5 million in Series A funding, led by Felicis, with some big names like NVentures (Nvidia's venture arm) and South Park Commons on board. But what sets Simular apart? It's not just another AI browser assistant; it's an AI agent that takes control of your entire PC, Mac, or Windows, and that's a game-changer!
Co-founder and CEO Ang Li puts it best: "We can move the mouse, click, and repeat any human activity in the digital world." Imagine copying and pasting data into a spreadsheet with an AI assistant doing the heavy lifting!
And here's where it gets controversial: Simular is working on an agent for Windows, too, as part of Microsoft's Windows 365 for Agents program. Co-founder Jiachen Yang, a reinforcement learning specialist, met Li at Google's DeepMind, where they published papers and worked on improving Google products like Waymo.
But here's the catch: LLMs (Large Language Models) can hallucinate, and that's a big problem when it comes to agentic tasks that require thousands of steps. Simular's solution? A 'neuro-symbolic computer use agent' that lets the AI explore freely, with the human user guiding it towards success. Once a successful trajectory is found, it becomes deterministic code, repeatable and trustworthy.
"Our approach is to let the LLM write code that's deterministic. So, if a workflow works, it'll always work," Li explains.
And this is the part most people miss: with Simular's method, the end user has control over the deterministic code, not the LLM. They can inspect, audit, and trust the code, ensuring a secure and reliable automation process.
So, will Simular's method be the key to bringing AI agents to every worker's fingertips? Time will tell, but early beta customers are already automating tasks like VIN number searches and extracting contract info from PDFs.
With a total funding of around $27 million, Simular is well on its way to making AI-assisted PC control a reality.
What do you think? Is Simular's approach the future of AI-human collaboration, or are there potential pitfalls we should consider? Let's discuss in the comments!
