Company profile · 3 min read
Simular: autonomous computers for recurring GUI work
Simular combines Sai computer agents, Agent S research and Simulang scripting for browser, desktop and operating-system automation.
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
Published · Updated
Simular is building autonomous computers for people and teams that spend their days clicking through software. Its product stack spans Sai, a personal computer-use agent; Simular for Business; Agent S research; and Simulang, an open-source scripting layer for repeatable browser and desktop workflows.
What it does
The current Simular site says Sai can run on a user’s device or a cloud computer, operate browsers and desktop apps, and execute recurring tasks with visible approvals. The pricing page lists a free Explore tier, $50/month pay-as-you-go with 50,000 credits and a dedicated cloud computer while it runs, and $500/month Sai Unlimited with an always-on cloud computer; enterprise pricing is custom with SSO/RBAC/SOC2 and support options.
Simular also publishes Agent S research and Simulang, an open-source language for browser, native-app and OS-level automation. The site claims Agent S reached top results on OSWorld and that its neurosymbolic approach can reduce token use by 90% or more; those are company-reported research claims, not a guarantee for a user’s workflow.
Why I’d look closer
The advantage is reach. If a task is trapped in a desktop app or has no clean API, a GUI agent can interact with it as a person would. The company’s founders bring relevant research experience: Ang Li is described as a former DeepMind, Apple, Google and CMU Robotics researcher, and Jiachen Yang as a former DeepMind and LLNL researcher.
The tradeoff is control. An autonomous computer with browser, file and terminal access can make real changes, and cloud execution introduces credential and data-boundary questions. The product says it uses approval-based safety, but the user still needs to scope the workspace and understand what happens when a UI changes.
What I’d ask
Which tasks can run unattended and which always pause? How are passwords, browser sessions, local files and cloud desktops isolated? Can a run be replayed, stopped and rolled back? What is included in credits, and how does a team audit actions across Sai, Agent S and custom scripts?
My editorial take
Shortlist Simular if repetitive GUI work is the constraint and the team can define reversible tasks. Start with drafts, reports and controlled file workflows, keep approvals on external side effects and choose the $50 tier only after measuring correction time. The product’s edge is computer reach; its adoption depends on operational visibility.
Quick facts
| Field | Sourced detail |
|---|---|
| Product | Sai computer agent, Agent S research and Simulang workflow scripting |
| Buyer | Individuals, business teams and developers automating GUI work |
| Published pricing | Free Explore; $50/month pay-as-you-go; $500/month Unlimited; enterprise custom |
| Control claims | Cloud or local execution, approval-based control and visible runs |
| Main question | Can the agent reach the needed UI without creating an unreviewable action surface? |
Sources checked
| Source | Checked |
|---|---|
| South Park Commons Simular profile | 2026-09-19 |
| Simular homepage | 2026-09-19 |
| Sai pricing | 2026-09-19 |
About the author
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.
First1000 ↗ · X ↗