Agent demos fooled you 👎 It trips, it asks for help, and somehow you're the one copying, pasting and running the commands. Meet your #Robosecretary. Meet your #SaiFleet. Sai brings state-of-the-art computer use to your own computer - your apps, your OS, in a floating window that leaves your mouse alone. Need more than one? Spin up to 100 cloud machines, each on the OS you pick, each on a different job. 100x the work, done without you. Sai. The autonomous computer fleet at your command. Free to start → sai.simular.ai

LLM-based agents reason every turn; agents, with our neuro-symbolic mode turned on, would plan, codify, execute, self-heal when UI changes, and keep running, faster and faster The chart shows how Sai, using the same base models, outperforms Opus- and GPT-based agents on repeated runs, and costing much less

New paper is out 📣 An agent shouldn’t have to work out the same task from scratch every time. Our latest research gives it a procedure it can reuse and repair, reducing the cost of repeat work. Read more here 👉 simular.ai/articles/real-wor…

Three years into making computer use agents, I kept asking myself: Why hasn't CUA reached super intelligence yet? @sai_borg (Sai), which happens to sound like SI, has invented a new way to build computer use agents: not just more reliable, but also 99% cheaper The answer lies in how humans function: build muscle memory to optimize for efficient output over time We released a paper today on this new paradigm, called neuro-symbolic Give it a read here: simular.ai/articles/real-wor…

Most AI agents today are disposable reasoners. They solve a task once, then start from scratch the next time. And when a learned workflow breaks, they often fall back to reasoning from scratch again. I think agents should work very differently: 𝐥𝐞𝐚𝐫𝐧 → 𝐫𝐞𝐮𝐬𝐞 → 𝐟𝐚𝐢𝐥 → 𝐫𝐞𝐩𝐚𝐢𝐫 → 𝐢𝐦𝐩𝐫𝐨𝐯𝐞 Introducing 𝐍𝐞𝐮𝐫𝐨-𝐒𝐲𝐦𝐛𝐨𝐥𝐢𝐜 𝐂𝐨𝐦𝐩𝐮𝐭𝐞𝐫 𝐔𝐬𝐞, where agents turn execution experience into reusable, continually improving policies. Instead of repeatedly paying a frontier model to re-plan every action, the agent distills experience into a neuro-symbolic program: stable knowledge becomes executable code, while perception and uncertain decisions remain neural. More importantly, these policies are not static. When execution fails, the agent diagnoses what went wrong, reasons about what should have happened, and repairs the policy for future runs. In that sense, the agent becomes 𝐬𝐞𝐥𝐟-𝐡𝐞𝐚𝐥𝐢𝐧𝐠: failures are not just errors to recover from, but opportunities to improve the underlying skill. As the agent encounters new parameters, states, and edge cases, the same policy can keep evolving. The policy itself becomes a form of 𝐚𝐜𝐜𝐮𝐦𝐮𝐥𝐚𝐭𝐞𝐝 𝐨𝐩𝐞𝐫𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐦𝐞𝐦𝐨𝐫𝐲. To me, this points toward an important direction for continual learning in agents. Continual learning doesn’t have to mean constantly updating billions of model weights. It can also mean 𝐜𝐨𝐧𝐭𝐢𝐧𝐮𝐚𝐥𝐥𝐲 𝐛𝐮𝐢𝐥𝐝𝐢𝐧𝐠, 𝐫𝐞𝐮𝐬𝐢𝐧𝐠, 𝐫𝐞𝐩𝐚𝐢𝐫𝐢𝐧𝐠, 𝐚𝐧𝐝 𝐫𝐞𝐟𝐢𝐧𝐢𝐧𝐠 𝐞𝐱𝐞𝐜𝐮𝐭𝐚𝐛𝐥𝐞 𝐬𝐤𝐢𝐥𝐥𝐬 𝐟𝐫𝐨𝐦 𝐞𝐱𝐩𝐞𝐫𝐢𝐞𝐧𝐜𝐞. The results: - 𝐒𝐎𝐓𝐀 𝐩𝐚𝐬𝐬^3 𝐬𝐮𝐜𝐜𝐞𝐬𝐬 𝐫𝐚𝐭𝐞 across all four settings, improving by 3.6–15.8 𝐩𝐨𝐢𝐧𝐭𝐬 - Up to 217× 𝐥𝐨𝐰𝐞𝐫 𝐞𝐱𝐞𝐜𝐮𝐭𝐢𝐨𝐧 𝐜𝐨𝐬𝐭 - Up tp 5.1× 𝐥𝐨𝐰𝐞𝐫 𝐥𝐚𝐭𝐞𝐧𝐜𝐲 Long term, useful agents shouldn’t just be able to reason. They should 𝐫𝐞𝐦𝐞𝐦𝐛𝐞𝐫 𝐰𝐡𝐚𝐭 𝐭𝐡𝐞𝐲 𝐥𝐞𝐚𝐫𝐧𝐞𝐝, 𝐫𝐞𝐮𝐬𝐞 𝐰𝐡𝐚𝐭 𝐚𝐥𝐫𝐞𝐚𝐝𝐲 𝐰𝐨𝐫𝐤𝐬, 𝐫𝐞𝐩𝐚𝐢𝐫 𝐰𝐡𝐚𝐭 𝐛𝐫𝐞𝐚𝐤𝐬, 𝐚𝐧𝐝 𝐢𝐧𝐜𝐫𝐞𝐚𝐬𝐢𝐧𝐠𝐥𝐲 𝐤𝐧𝐨𝐰 𝐰𝐡𝐚𝐭 𝐭𝐡𝐞𝐲 𝐧𝐨 𝐥𝐨𝐧𝐠𝐞𝐫 𝐧𝐞𝐞𝐝 𝐭𝐨 𝐫𝐞𝐚𝐬𝐨𝐧 𝐚𝐛𝐨𝐮𝐭. Very proud of the team for pushing toward this vision.

State of the art and 100x cost reduction. It’s a big step towards making work become like something like electricity.

What can Picture-in-Picture really bring you? Two truly independent workspaces - with just one device. You can keep working in your own space, while another workspace is fully led by @sai_borg to plan, research, and execute tasks for you

@sai_borg now runs on your local windows computer with Picture in Picture! No more agent mouse hijacking!

Watch Sai work locally on your desktop in Picture-in-Picture view. No more hijacking your mouse and keyboard while you handle other tasks. Now live on macOS too 🖥️ Put Sai to work on Mac or Windows. sai.simular.ai/

Sai is the FIRST computer-use agent to bring Picture-in-Picture to Windows. Now you can keep Sai visible while it works alongside you across different tasks and applications. When is it coming to macOS? Can’t wait! Try it now: tryit.cc/NiCc6b #ad

Every founder has that lead list they keep putting off. I handed mine to Sai and walked away: sai.simular.ai/ #SaiFleet #Robosecretary First I asked a regular chatbot. It gave me a list of startup names in seconds, but then the real work was still on me: Open every website, find the founders, hunt down their LinkedIn, check it's all accurate, and paste it into a sheet. That's easily a couple of hours. Sai actually did that part. It opened the browser, went through YC pages, verified each founder, and built the Google Sheet itself. 117 steps, and I didn't do a single one. I came back to 10 AI startups with websites, founders, and LinkedIn, ready for outreach. That's the difference: a chatbot tells you what to do. Sai does it.

Connect an app once. Sai can use it in every task after. → connect from Ability : 20+ apps, ready whenever a task needs one → a Slack message, a Gmail thread or a Granola note can start the work Hand Sai the routine work 🖥️ sai.simular.ai #saifleet #robosecretary

SF Tech Week has 1,800+ events At some point, “finding the right event” becomes a time-consuming job So I built a simple search tool with @sai_borg + Jev that lets you search the entire event list in plain language: “Find GTM events on Tuesday evening” “Hackathons for engineers” “AI events with founders and investors” No filters. No scrolling. It surfaces relevant events in under a second sai.simular.ai/site/6934c6b4… And if your search is more like “Something fun on Friday that is not another AI panel” :) Come hang out with us at Pixels After Dark: Art, Music and Computer Agents partiful.com/e/YUKBbGuv0GuGM…

"@sai_borg playing minecraft with Jev as minions > netflix & chill first AI stream i've actually kept open try sai: sai.simular.ai #saifleet #robosecretary"

Everyone's giving agents a cloud computer now... With state-of-the-art computer use, Sai runs Windows, Linux, macOS and Android, all at once. You can unlock up to 5 cloud computers with Sai under $50 👉sai.simular.ai/

A small business owner in Florida tested pretty much every AI tool out there and then,,,they found us! I'm going to frame this screenshot 🖼️ Thank you for the 5 stars 🥹

act of getting things done is my top love language 🤟

Be brutally honest, would you delegate your mailbox to AI?

Marketing agencies for product launches are dead. This video took 60 minutes to create end to end - and I never touched CapCut All I prompted was: “Make a product launch video for sai.work” Then I gave it a Google Drive folder with a few demo recordings. Everything else - finding other relevant raw footage in internet, taking screenshots, editing, subtitles, voice-over, and music - was done by the agent itself.