SimulationThe initial AI opening agent for Mac OS and Windows, has brought $ 21.5 million led by Felicis, confirming (the gut arm of Nvidia South, and others joined.
Simular is an interesting agent introduction because unlike others, it does not try to control the browser but the PC itself. (The Ahi Ai agent refers to a system that can complete complex tasks with minimal human intervention.
On Monday, it announced the release of version 1.0 for Mac OS. But it can also be used with Microsoft to develop agents for Windows. The startup is one of five agent companies accepted into the Windows 365 for agent program Microsoft announced in mid-November. (Others are Manus ai, Fellou, Genspark, and pinyfish, and kotak.
Another reason to watch the simulation is that Fides Fides of The Bona is a continuous learning scientist who works in the deepmind of Google, where he met with the co-founder, specializing in teaching learning Jiachen Yang. While the team has published parts of the paper, the work is not academic, Li said. It is intended to enhance Google products, including puppets.
If the AI product background is helpful because, before the religious future of Silicon Valley can materialize, there is a host of technical problems to solve. One of the biggest is LLMS Hallucinate some percentage of the time.
Agent tasks can complete thousands to millions of discrete steps. Not only are hallucinations at any step not true of all agent work, but hallucinations become statistically more likely as the number of steps grows.
One way to solve this is to create “nonteristic” “which is not”, “does not allow LLM to be ultimately creative, the response is the same. But the risk limits all aspects of solving agent problems.
TechCrunch events
San Francisco
I’m fat
October 13-15, 2026
Simular is married to two. The agent will be fast in the task, with the human user in the Central course correcting, until the agent succeeds. Then the human key in the workflow of the task, which makes the signature, repeatable.
“Our solution is, let the agent keep exploring successful trajectories. Once you find a successful trajectory, that’s a deterministic code,” Li explained.
The reason the opener can do this is because the work – which admittedly is still early – is not just an llm wrapper that sends and retrieves data to the model.
“We have a new technology that is not used by other agent companies. We call it the ‘symbolic computer Neuro computer using agents.’ It’s not complete on a basic basis,” he said. “My way of dealing with hallucinations is to let LLM write code that is deterministic. So if you have a workflow that works, we will be successful in the future as well.”
Another benefit is that this deterministic code that performs repeatable tasks is in the hands of the end user, not the LLM. “Once they have the code, they can trust it, because they can check it, they can audit it, they can see what’s going on,” Li said.
Time will tell if that method is the magic that will bring agents into the hands of every worker. Li said early beta customers include Automating VIN Searches car watchers, and Hoas’ contract information grabs from PDF. And the company Open Source Project (Available only for Mac OS at the moment) has led to automation ranging from content creation to sales and marketing.
Simular previously raised $5 million in seed, bringing the total raised to around $27 million. Other investors in the company include set venture, flying fish partners, next Samsung, Xoogler Ventures, and podcaster and podcaster and angel Investor Lenny Rachitsky, the company said.
https://www.youtube.com/watch?v=ne-3tfhvf9c
Check out the latest news on everything from agentic AI and Cloud Cloud AI and Cloud Cloud to security and more from flags