Simular Launches AI Agent for Mac, Plans Windows Version Amid $21.5 Million Funding Round
Simular, an innovative startup focusing on AI solutions for Mac OS and Windows, has successfully secured $21.5 million in a Series A funding round led by Felicis, with additional support from existing investors including NVentures (NVIDIA’s venture division) and South Park Commons.
The company introduced its version 1.0 for Mac OS on Monday and is collaborating with Microsoft to develop an AI agent for Windows. Notably, Simular is one of five startups selected for Microsoft’s Windows 365 for Agents program, which was announced in mid-November. While specifics about the Windows release timeline were not provided, co-founder Li suggests it could rival the popularity of their Mac version.
The expertise of Simular’s leadership team is noteworthy; Li, previously a continuous learning scientist at Google’s DeepMind, partnered with reinforcement learning expert Jiachen Yang. Their background includes endeavors that enhanced various Google products, including self-driving technology, bridging academic research with practical applications.
As the tech sector anticipates the future of AI agents, challenges remain, particularly regarding the inherent issue of hallucinations in large language models (LLMs). These errors can jeopardize the success of tasks that involve extensive step sequences, where a single hallucination can derail an entire process.
To address this, Simular is pioneering a model that allows its AI agents to explore multiple strategies while integrating user input to correct and refine workflows. Once a successful path is identified, it can be locked in as deterministic and repeatable.
“Our solution enables agents to explore efficiently until they find a successful trajectory, after which that pathway becomes deterministic code,” explained Li.
Distinct from other agent technologies, Simular claims to utilize an innovative approach termed “neuro-symbolic computer use agents,” which goes beyond typical LLM methods. This strategy aims to transform LLM outputs into reliable, repeatable code, empowering users to inspect and audit the outcomes.
Early beta testers include automotive dealerships automating VIN searches and homeowners’ associations extracting information from PDFs. The company’s Mac-only open-source initiative has facilitated a range of automations, from content creation to sales processes.
With a total of approximately $27 million raised, Simular’s growth continues to attract attention, bolstered by the involvement of other notable investors, such as Basis Set Ventures and Samsung NEXT.
