Human-level Computer Use Agents Have Arrived

Dec 2 2025

At Basis Set, we’ve believed in the power of agentic systems since the earliest days of the firm. That belief led us to start building our own in-house AI agent, Pascal, early on. Ironically, it was this system that first connected us with Simular, surfacing the company before “computer-use agents” was really even a category.

Simular’s founding team had many of the earliest and most unique perspectives on building computer-use agents that are truly ready for primetime, publishing dozens of top papers in the in the field at Google DeepMind and beyond — laying the groundwork for neurosymbolic systems that combine the creativity of LLMs with the determinism of code for high reliability.

That early connection through Pascal led to our first conversations and ultimately to Basis Set leading Simular’s $5 million Seed round. Today, Simular announces its $21.5 million Series A led by Felicis.

This team has moved at an extraordinary pace from the start. Simular’s agents have repeatably pushed the state-of-the-art to new heights. Their latest Agent S3 demonstrates multi-step autonomy, consistent action sequencing and a level of reliability approaching human operators. With Simular 1.0, they’ve pushed further — shipping agents that function like true teammates, capable of learning from mistakes and completing complex workflows end-to-end.

They are now knocking on the door of human-level computer use, standing on the precipice of fundamentally changing how we interact with machines.

Today's Series A milestone reflects both the quality of the team’s execution and the emerging importance of this category. Human-level computer-use agents aren’t a future concept anymore — they’re arriving now and Simular is defining what that future looks like.

We’re proud to have partnered with them early, and we’re even more excited for what’s ahead.