What if your computer could work independently, just like a human assistant? After a decade in AI research at Google DeepMind, something still deeply puzzled me: we spend 2.5 hours daily just moving our computer mouse. That's 30% of our work time on mechanical tasks. This realization led me to build Simular AI. We're creating AI agents that use computers like humans do - clicking, typing, and navigating interfaces - so you can focus on what truly matters. My journey spans: - Research: 39 papers in top AI venues (NeurIPS, ICML, CVPR, ICLR) - Engineering: Scaled AI systems at Google (Ads, Maps, YouTube) - Leadership: Built & led teams in autonomous driving - Competition: 2x World Finalist in ACM/ICPC Now, I'm on a mission to democratize AI assistance. Because the future of work isn't about replacing humans - it's about augmenting them. Want to see the future of human-computer interaction? → Visit simular.ai → Follow my insights on X @angli_ai Positions: CEO & Co-Founder at Simular (2023 - Present), Head of US Software Engineering & Principal Scientist at Baidu Apollo (2021 - 2023), Staff Research Scientist at Google DeepMind (2017 - 2021) Skills: No skills available Recent Posts: AI Agents Should Think Fast and Slow We’re all excited about a future where AI agents work relentlessly for us. Yet, many argue that LLMs and agents are still unreliable— I agree. But that doesn’t seem a problem to me. LLM agents alone aren’t the full solution to autonomy— they’re just one piece of the puzzle. (this is the second part of my talk at Princeton) --- Humans operate using two cognitive systems, System 1 – Fast, instinctive, and automatic (like Robotics Process Automation). System 2 – Slow, deliberate, and reasoning-driven (like LLM-based agents). Right now, when some people talk about AI agents, they’re often referring only to System 2 thinking—reasoning, decision-making, and problem-solving. It's exciting. Ironically, when many of the others mention about AI agents, they're often thinking about System 1 only —automation of repetitive tasks. Neither of them is wrong. However, the real powerful agents need to equip both thinking systems. The future of AI agents should think fast and slow. --- There’s an old saying: “Practice makes perfect.” In cognitive science, practice shifts tasks from System 2 (slow, effortful thinking) to System 1 (fast, automatic execution). AI agents can follow the same principle: System 1 should handle routine, repetitive tasks efficiently. System 2 should step in when uncertainty arises or when deeper reasoning is needed. AI agents are not equivalent to foundation models. The nature of AI agents is an orchestration layer on top of both systems. It becomes more of an infrastructure problem. --- At Simular, we design AI agents that evolve over time: Repetitive actions gradually become automated, requiring less cognitive effort. System 2 reasoning is activated when System 1 encounters difficulties. With this agent design principle, Your cost of automation will become more affordable Your agents will become more reliable You will be more confident integrating agents into your workflow. And we are closer to the future where AI agents work relentlessly for you. I quit my job during Tech Week in San Francisco. While I was there I only attended one event... Here's where Silicon Valley's most ambitious founders find their tribe: The one event I went to? A gathering at South Park Commons. The energy was different there. They weren't just builders, they were dreamers focused on making a real impact. As a first-time founder stepping into this new world, I knew I needed more than capital. I needed to upgrade my entire perspective. South Park Commons gave me exactly that: A community of builders I would've never met in a corporate job. The real magic happened on Day One: I showed up with a prototype pieced together with outdated tech. But when they saw our AI agent navigating multiple apps seamlessly, the room lit up with genuine excitement. Their reactions transformed my entire approach to building: Instead of chasing quick wins, we focused on: • Building solid foundations • Testing core assumptions • Getting honest feedback • Finding our true mission Over time, the community became our foundation. Every demo revealed new insights - from UI tweaks to core functionality. This support system helped us uncover our real purpose: Saving people time. Professionals waste 30% of their workday on tedious computer tasks. That's why we're building SimularAI - to help people offload mundane work to their personal AI agent: • No more endless copy-pasting • No more repetitive form filling • No more manual data entry Your computer should handle these tasks for you. Not dump them on your plate. But we never would have gotten this far without the right community, and they're accepting applications for a few more days. Want to join ambitious builders who care about long-term impact? South Park Commons is accepting applications until Feb 2nd. Apply here: https://lnkd.in/dbshd6QM 🌟 We’re thrilled to see ARTLAS launch their incredible AI art companion, and being featured by Product Hunt! As one of Simular’s valued customers, ARTLAS leverages our computer-use agent to seamlessly organize the world’s museum data into personalized and magical digital experiences. Congrats to the ARTLAS team for bringing this innovative product to life! If you’re as excited as we are, show your support by giving them an ▲ upvote on Product Hunt: https://lnkd.in/gfZVEGrj Here’s to making art more accessible, personal, and inspiring—together! 🖼️✨ 🚀 We are thrilled to announce that Agent S has been accepted to ICLR 2025, one of the premier conferences in machine learning and artificial intelligence! 🎉 Agent S is a leading computer use agent, designed to take user input and autonomously perform actions directly at the operating system level across all major platforms. And the best part? It’s open-sourced! 🛠️ Open sourcing Agent S represents our commitment to transparency, collaboration, and accelerating innovation in AI. We believe that by sharing tools like Agent S with the world, we can collectively push the boundaries of what’s possible in AI-driven productivity and interaction. 🔗 Code: https://lnkd.in/dpsCmgWR 🔗 Paper: https://lnkd.in/gEXSqF8y 🔗 Blog Post: https://lnkd.in/duTn8JAb 💡 To the amazing authors and contributors: Congratulations on this incredible achievement and thank you for your hard work and dedication in bringing Agent S to life! Your innovation is inspiring and paves the way for countless future breakthroughs. 👏👏 We’re just getting started. Join us in exploring the future of AI-powered agents. Let us know your thoughts, and don’t forget to check out the resources above to dive deeper into Agent S! Saaket Agashe Jiuzhou Han Shuyu Gan Jiachen Yang Ang Li Xin Eric Wang #ICLR2025 #AI #OpenSource #AgentS #MachineLearning #AIInnovation 🚀 We are thrilled to announce that Agent S has been accepted to ICLR 2025, one of the premier conferences in machine learning and artificial intelligence! 🎉 Agent S is a leading computer use agent, designed to take user input and autonomously perform actions directly at the operating system level across all major platforms. And the best part? It’s open-sourced! 🛠️ Open sourcing Agent S represents our commitment to transparency, collaboration, and accelerating innovation in AI. We believe that by sharing tools like Agent S with the world, we can collectively push the boundaries of what’s possible in AI-driven productivity and interaction. 🔗 Code: https://lnkd.in/dpsCmgWR 🔗 Paper: https://lnkd.in/gEXSqF8y 🔗 Blog Post: https://lnkd.in/duTn8JAb 💡 To the amazing authors and contributors: Congratulations on this incredible achievement and thank you for your hard work and dedication in bringing Agent S to life! Your innovation is inspiring and paves the way for countless future breakthroughs. 👏👏 We’re just getting started. Join us in exploring the future of AI-powered agents. Let us know your thoughts, and don’t forget to check out the resources above to dive deeper into Agent S! Saaket Agashe Jiuzhou Han Shuyu Gan Jiachen Yang Ang Li Xin Eric Wang #ICLR2025 #AI #OpenSource #AgentS #MachineLearning #AIInnovation We're very strict with Demo Faire—live demos only! See the future 11 teams previewed for us at our Winter 2025 Demo Faire. Featuring demos from Jeff Shaw, Abhinav Shrivastava, Shishira R Maiya, Charles Niu, Rohit Swamy, Edward Zhang, Fangchang Ma, Alex Roe, Jonathan Arena, Adam Behrens, Aditya Agarwal, Grey Nguyen, Ben Yang, Max Mona, Vinay Jain, Aman Jain, Ang Li, Jiachen Yang, Samantha Whitmore & Jason Yuan Check out the full recap below! ⬇️ 5 years ago, A Googler asked me “why directors have executive assistants and I don't?” It sounds like a fairness problem. I think it’s all about what we have to do manually every day. Have you ever done copy-pasting between spreadsheets and websites? Have you ever filled out long forms in visa or job applications? Have you ever been exploring linkedin profiles to discover leads and connections? I've done all of them and I believe you do so. Elite people have personal assistants to help them handle all of these misc stuff. I do not have one and you may not have one either. But now, we are seeing the opportunity to give everyone an AI assistant, to free your time from mundane tasks. This is a talk I gave in Princeton one month ago and I received great feedbacks so I decided to write down some thoughts here and share it at the end of this year. — 𝐍𝐨 𝐨𝐧𝐞 𝐤𝐧𝐨𝐰𝐬 how much time they spent on moving a computer mouse (or trackpad)? We asked people around. No one knows. So we did an experiment earlier on with 10 volunteers in tech. The average is 2.5 hours per day. This only counts the time the computer mouse is “moving”. It doesn’t count when the mouse is static. It doesn’t count when you are typing on the keyboard. That’s 30% of your total work time! Do we have to? — Why Simular is Simular? This is why Simular was created, with the mission to create digital agents that can use computers like a human. So, our computers become our assistants. We call it Simular because they are like humans, but not humans. Do we have to use computers like a human? — We already have APIs, can you just orchestrate APIs? The API here means the interfaces developed by the creators of different applications. This is a common question I’m often asked. Yes, tasks can be automated by APIs. However… Someone needs to build the APIs. And what’s worse, We have 3 new websites showing up on the Internet every second, that’s like 90,000,000 new websites every year. Can we catch up with the speed of such creation? Consider that content creation can become way faster in the future. We need something more general than using APIs. By that, we mean, the agent uses a computer mouse and keyboard to perform tasks. — Should GUI still exist in the future? People questioned me. Okay, you may not need to see GUI in the future, but, GUI will exist. The principle is human will have to supervise agents for ever. Whenever we talk about supervision, both parties need to work in the same environment. GUI is a natural way for humans to interact with the digital world. It’s an effective way for humans to supervise computer agents. If you look at how tech evolves, it just becomes closer and closer to the original state of our nature. e.g., from black-white to colorful, from 2D to 3D, to virtual reality ... — The post becomes long. I’ll pause here and share the second part in 2025 about why computer use agents should be designed to think fast and slow. Simular is heading to #NeurIPS2024 Our head of research Xin Eric Wang will present our Open Sourced Agent S, an industry leading computer use agent, in the Open World Agent workshop on Sunday, with (of course) Simular Swags! Agent S repository: https://lnkd.in/dpsCmgWR Simular is heading to #NeurIPS2024 Our head of research Xin Eric Wang will present our Open Sourced Agent S, an industry leading computer use agent, in the Open World Agent workshop on Sunday, with (of course) Simular Swags! Agent S repository: https://lnkd.in/dpsCmgWR Thanksgiving is all about family, gratitude, and great food 🦃 — but let’s face it, finding the perfect recipes can be overwhelming. That’s where tech innovation steps in. I’m excited to share a sneak peek of Simular Browser, a demo of an agentic browser that automatically navigates the web to find exactly what you need — in this case, the perfect Thanksgiving recipes. No endless searching, no jumping between tabs — just results, effortlessly. This is what the future of browsing looks like. What do you think about AI simplifying our daily tasks like this? Let me know your thoughts — and Happy Thanksgiving! 🥧🍂 https://www.simular.ai/ 🎉 Excited to announce Agent-S v0.1.0! Major updates include API refactoring, improved setup instructions, and Perplexica integration. Thanks to our amazing contributors! 🚀 #OpenSource #AI #GitHub #Release https://lnkd.in/ekNqF5AH Our CEO Ang Li is invited to speak at Princeton University this Thursday (November 21st) about Simular's vision and recent progress. Tune in to learn more! https://lnkd.in/dQwz4bfw Agent S just passed 500 GitHub stars. Try it out and start using your computers in a “simular” way. Was working on a deck and summarized the progress of operating system level agents over the past year. Been gazing at this figure for long time. It made me feel so exciting to see what's upcoming in the next few months. Who tops off the benchmark does not seem to matter any more, what matters is that our society starts unlock the new potential in computing, where the boundaries of what’s possible are being rewritten. Howdy fellas! 👋🏼 Spark and Trouble are at it again, connecting the dots between human-like thinking and machine-powered action. Is it possible for computers to think, act, and explore like one of us? The answer might be just a few clicks away in this edition's puzzle! Here’s a sneak peek into today’s edition 👀 🔥 “Agent S” Can Perform Your Desktop Tasks—How Far Can It Go? 🔥 Want to Create a Webinar That Wows? Start with today’s Fresh Prompt! 🔥 5 AI Tools that you just can’t miss out on 🔥 A student’s AI creation is turning heads when it comes to real-time call screening Time to jump in!😄 https://lnkd.in/g2DfX-6n Simular | Saaket Agashe | Jiuzhou Han | @Shuyu Gan | Jiachen Yang | Ang Li | Xin Eric Wang Tezan Sahu | Sandra Anil #agents #simular #airesearch #aisgents #guiagent #opensource #github #ainewsletter Over the last few weeks, you've probably seen me cover a bunch of stuff about how AI agents are transforming specialized domains, like research, CLI commands, and more. But today, Sandra & I have the perfect blend of something new – Meet 𝐀𝐠𝐞𝐧𝐭 𝐒, an open multimodal LLM-based agentic framework developed by the folks at Simular, designed to interact with computers through a Graphical User Interface (GUI). 🤯 In simpler terms? It's an AI that can use your computer just like you do – clicking, typing, and navigating its way through tasks with surprising efficiency. So, what's new? In this post, we'll break down: 🕵️ The 3 key innovations that power 𝐀𝐠𝐞𝐧𝐭 𝐒 🕵️ How 𝐀𝐠𝐞𝐧𝐭 𝐒 outperforms existing agents in system-level and desktop tasks 🕵️ The exciting potential applications of 𝐀𝐠𝐞𝐧𝐭 𝐒 across industries To dive deeper into the details of 𝐀𝐠𝐞𝐧𝐭 𝐒, check out the full article here:  https://lnkd.in/gcdSqAE3 Simular | Saaket Agashe | Jiuzhou Han | @Shuyu Gan | Jiachen Yang | Ang Li | Xin Eric Wang And to stay on top of your AI game, subscribe to The Vision Debugged - AI Newsletter - your 2x per week dose of AI breakthroughs: #agents #simular #airesearch #guiagent #opensource #ainewsletter 🚨Paper Alert 🚨 ➡️Paper Title: Agent S: An Open Agentic Framework that Uses Computers Like a Human 🌟Few pointers from the paper 🎯Authors of this paper presented “Agent S”, an open agentic framework that enables autonomous interaction with computers through a Graphical User Interface (GUI), aimed at transforming human-computer interaction by automating complex, multi-step tasks. 🎯 Agent S aims to address three key challenges in automating computer tasks: acquiring domain-specific knowledge, planning over long task horizons, and handling dynamic, non-uniform interfaces. 🎯To this end, Agent S introduces experience-augmented hierarchical planning, which learns from external knowledge search and internal experience retrieval at multiple levels, facilitating efficient task planning and subtask execution. 🎯In addition, it employs an Agent-Computer Interface (ACI) to better elicit the reasoning and control capabilities of GUI agents based on Multimodal Large Language Models (MLLMs). 🎯Evaluation on the OSWorld benchmark shows that Agent S outperforms the baseline by 9.37% on success rate (an 83.6% relative improvement) and achieves a new state-of-the-art. 🎯Comprehensive analysis highlights the effectiveness of individual components and provides insights for future improvements. Furthermore, Agent S demonstrates broad generalizability to different operating systems on a newly-released WindowsAgentArena benchmark. 🏢Organization: Simular AI 🧙Paper Authors: Saaket Agashe, Jiuzhou Han, Shuyu Gan, Jiachen Yang, Ang Li, Xin Eric Wang 1️⃣Read the Full Paper here: https://lnkd.in/guqbUpt8 2️⃣Project Page: https://lnkd.in/ggjV-vf5 3️⃣Code: https://lnkd.in/gJFGTKvM 🎥 Be sure to watch the attached Demo Video - Sound on 🔊🔊 Find this Valuable 💎 ? ♻️REPOST and teach your network something new Follow me 👣, Naveen Manwani, for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements. The Top ML Papers of the Week (Oct 14 - Oct 20): - Agent S - CoTracker3 - Thinking LLMs - Model Swarms - Introspection in LLMs - On the Planning Abilities of OpenAI's o1 Read on for more: 😲 Agent-S is a new framework that uses multi-modal agents to handle computer GUI tasks, like completing tasks on the internet. Super impressive! 💡 Agent S is an open agentic framework designed to interact autonomously with computers through a Graphical User Interface (GUI). It aims to automate complex, multi-step tasks. Example tasks: ⛳ Remove an email account: Agent S can navigate through the settings of an email client to locate and remove a specified email account. ⛳ Calculate total sales, average monthly sales, and generate visualizations: Agent S can interact with spreadsheet software like LibreOffice Calc to perform calculations on sales data and create charts for visualization. 📖 Agent S automates computer tasks using three main strategies (I've oversimplified it here for brevity). ⛳ Hierarchical Planning: The Manager breaks down tasks into subtasks by using web searches and past experiences. The Worker executes these subtasks with reflective guidance, signaling completion or failure. ⛳ Continual Learning: Agent S learns from successes and failures by continuously updating its memory with new task experiences, improving over time. ⛳ Agent-Computer Interface (ACI): ACI allows precise interactions with the computer's GUI using a combination of image input, OCR, and predefined actions like clicks and typing. Link: https://lnkd.in/ebc-ipGN Simular Research Introduces Agent S: An Open-Source AI Framework Designed to Interact Autonomously with Computers through a Graphical User Interface Simular Research introduces Agent S, an open agentic framework designed to use computers like a human, specifically through autonomous interaction with GUIs. This framework aims to transform human-computer interaction by enabling AI agents to use the mouse and keyboard as humans would to complete complex tasks. Unlike conventional methods that require specialized scripts or APIs, Agent S focuses on interaction with the GUI itself, providing flexibility across different systems and applications. The core novelty of Agent S lies in its use of experience-augmented hierarchical planning, allowing it to learn from both internal memory and online external knowledge to decompose large tasks into subtasks. An advanced Agent-Computer Interface (ACI) facilitates efficient interactions by using multimodal inputs. The structure of Agent S is composed of several interconnected modules working in unison. At the heart of Agent S is the Manager module, which combines information from online searches and past task experiences to devise comprehensive plans for completing a given task. This hierarchical planning strategy allows the breakdown of a large, complex task into smaller, manageable subtasks. To execute these plans, the Worker module uses episodic memory to retrieve relevant experiences for each subtask. A self-evaluator component is also employed, summarizing successful task completions into narrative and episodic memories, allowing Agent S to continuously learn and adapt. The integration of an advanced ACI further facilitates interactions by providing the agent with a dual-input mechanism: visual information for understanding context and an accessibility tree for grounding its actions to specific GUI elements.... Read full article here: https://lnkd.in/ga5QT94r Paper: https://lnkd.in/gEXSqF8y GitHub: https://lnkd.in/dpsCmgWR Simular Jiachen Yang Ang Li #ai Agents are not just about foundation models. Our research team showed the equivalent importance in organizing memories and acquiring new knowledge in this work. We also achieved significantly better results compared to pre-existing methods on this agent benchmark. It feels great to continue contributing to the open research community as a startup and to be at the forefront of shaping the next generation of human-computer interaction, powered by general agents. Simular’s AI employees sitting on top of each other, layered like a lasagna of productivity. Introducing Automated File Organization with Simular. In just a few clicks, your agent immediately transforms your messy desktop into a streamlined, productive workspace. Spend less time searching and more time creating!💡 🔍 Join the beta program and discover more actions in our community channel! 🔗 Learn more: https://www.simular.ai/ 🔗 Join the waitlist for Simular MacOS: https://lnkd.in/gy7GJMhP #Simular #SimularAI #AIAgent #ComputerAutomation #WeeklyDemo Chatting with AI as if you were talking with a real person is no longer a dream. Today, we are pushing things even further. Introducing Simular, the AI Agent that goes beyond conversation. Now, you can prompt your AI Agent to perform digital tasks for you on your personal computer, all through natural language. Check out the demo below!✨ Stay tuned, as the future is happening. 🚀 🔗 Learn more: https://www.simular.ai/ 🔗 Join the waitlist for Simular MacOS: https://lnkd.in/gy7GJMhP #Simular #SimularAI #AIAgent #ComputerAutomation #WeeklyDemo Is your Chrome tab a chaotic mess?🤯 Discover the game-changing “Organize Chrome Tabs” auto action with Simular. Just one click and watch your tabs transform—like magic! ✨Check out our quick demo video to see how effortlessly the Simular agent can handle this. Got ideas for other automated actions? Drop a comment below! 🔗 Learn more: https://www.simular.ai/ 🔗 Join the waitlist for Simular MacOS: https://lnkd.in/gy7GJMhP #Simular #AIAgent #OrganizeChromeTabs #ComputerAutomation #WeeklyDemo Frustrated with repetitive forms and constant data entry? There’s a smarter solution. Introducing Simular, your new AI assistant. With just a click, Simular automatically fills out forms using your securely stored information. Discover more with Simular’s action feature. Share which automated digital tasks could make your life easier—leave a comment and let us know! 🔗 Learn more: https://www.simular.ai/ 🔗 Join the waitlist for Simular MacOS: https://lnkd.in/gy7GJMhP #Simular #AIAgent #FillForm #ComputerAutomation #WeeklyDemo Come build cool stuff at our Hackathon at AGI House! https://lnkd.in/euaz_qpj There will be an awesome line up of speakers including Andrew Ng @ DeepLearning.AI, Harrison Chase @ LangChain, and Jose M. @ Groq You'll also get a chance to experiment with some of the newest agent tools with workshops hosted by Simular (Ang Li & Jiachen Yang) and Invariant Labs (Marc Fischer & Mislav Balunović & Luca Beurer-Kellner)! Wondering what an agent can do beyond searching and chatting? Join us next Saturday, 8/24, for a hands-on demo session of our Simular Agent at the AGI House Open-Source Agent Hackathon! 🤖 📅 Aug 24, 2024 🕙 11:00 AM PST 📍 AGI House 🔗 RSVP: https://lnkd.in/ejwz9dwa #SimularAI #AIAgent #Hackathon #Livedemo It’s been an incredible journey filled with innovative solutions, milestone moments, and fantastic community growth. Reflecting on our first year, we are getting closer to creating an AI agent that can truly impact and improve our work and lives. But this is just the beginning—stay tuned for more exciting demos and updates! 🚀 #SimularAI #AIAgent #AIStartup #Oneyearanniversary #Milestones I was asked what I was doing one year ago. I was literally just editing a white background slide deck with apparently no theme at all and debugging a buggy hacky demo code. It's been such an incredible journey so far and grateful to have everyone who helped, criticized, brainstormed alongside and whoever makes this journey more fun and more rewarding. Excited to contribute to this continual learning framework which reflects a very simple idea: a neural network should know how an old self would represent the data. And we showed such information can be used to avoid catastrophic forgetting. I believe this idea is very powerful and hope our work could spark more interesting works to enable neural nets to learn continuously in a better way. Totally unexpected audience flow at #CVPR Super excited to welcome Xin Eric Wang joining Simular! Xin is an extraordinary mission driven researcher who specializes in multimodal and embodied agents. It’s a great pleasure to work alongside with him to create personal agents in the digital world! Our team is heading to CVPR 2024 from June 19-21 at the Seattle Convention Center! Visit us at Booth 1743 to experience our latest progress and see how your personal AI companion can change the way you interact with your computer. Plus, we have an open position for Ph.D. Research Intern. If you're interested, drop by and connect with our team for a chat at CVPR! See you there!🚀 Start a Zoom meeting in a "Simular" way! Eric S. Yuan (he / him / his) This is so true A Googler told us they don’t use Google Chrome. That was quite surprising. What browser do you use then? Arc. And now, Googlers can use Simular to perform actions on Arc as well. Expect more to come! Two years ago, I couldn't have fathomed the ease of developing an app with Flutter from scratch as I can today. The advancements in AI have been staggering, making the process much more efficient. Despite the challenges of limited network access while traveling by plane or train, this app has been a lifesaver. The speed and accuracy of local AI today is truly mind-blowing. Amazing job Simular ! #AIRevolution #FlutterDevelopment #TechAdvancements #genai #offlinetraining #aiapplications #aiapps Simular was born three months ago. It has been an incredible journey, and I want to take a moment to express my gratitude to all Simular friends for joining us on this remarkable journey. As we are at the conclusion of 2023, I would like to give a recap about our achievements and outline our objectives for the upcoming year, 2024. In September, Jiachen Yang and I established our company and joined South Park Commons, an amazing community for founders. At the first week, we launched a MacOS chatbot prototype that could utilize any selected text as context. Many friends who knew us expressed their confusion as to why we just built a wrapper of GPT. We literally told them “just be patient”. In October, we launched our first API-based tool that integrated with Google Calendar. And by November, we had achieved our first pixel-level computer control, allowing us to physically manipulate the mouse and perform tasks. By then, Simular basically acquired a wide spectrum of abilities in the digital world. We had the opportunity to showcase our agent's capabilities at SPC's demo faire and were excited to receive a lot of positive feedback. https://lnkd.in/gwbBdUGn It’s been a tremendous journey for both of us. It’s like every morning when we wake up, we feel more excited about what’s next about the agent. We call the past three months the inception phase of our company, as we built out the basic capabilities of the agent. Simular's abilities have been steadily increasing. However, we are not content with the current speed. This leads to our next phase, scaling. Scaling will be the primary theme for the year 2024. Our objective is to significantly expand the range of supported actions in an exponential manner. As many of you may know, this is a very challenging problem, and we invite you to stay tuned on our progress. I still vividly remember the first thing Aditya Agarwal taught us in SPC is the dilemma faced by founders, which is finding the balance between focusing on the long term vision and attending to the details of the current product. We’ve been executing our plans consistently under this philosophy. If you have been following us from day 1, you probably already notice how much of a difference we have made so far. By saying that, I also want to share a preview for our upcoming launch in the next few days: 1. Introducing a fully private local mode which piggyback on the awesome community of llama.cpp and open source LLM models. 2. Improved handling of context selection, you can select any application, any window, any text or any image you want to start your task with. 3. More and better cross-app action controls, like creating google/zoom meetings and sending messages in different messaging apps, reading an image and setting calendar events, etc. And of course, a better streamlined user interface for a better user experience. Please apply to our beta program at simular.ai Wishing you all a happy and prosperous New Year! Just a while back the idea of something like Simular felt like a pure fantasy. And now, it’s becoming a reality for everyone to witness. Simular wasn’t born from data-driven calculations or deep market research. It sparked from a single flame of curiosity and discovery. We’ve spent decades researching about artificial intelligence. Trying to find an application that would truly impact our lives. And then when it clicked! Simular is consistent with many values, one of which is improving work and life. This demo did a great job of showcasing multiple brainstorming sessions and restless days of ideation and coding behind it. -- To present an AI that perceives, reasons, and takes action. The idea behind Simular is simple -- To make space for better and greater in life and work. So as you look at the cursor seamlessly gliding across the screen to get these jobs done. You can be productive without them holding you back. This was just a trailer, you can watch the full video here: https://lnkd.in/gNeT4hmK Join our waitlist to see Simular in action yourself. https://lnkd.in/gy7GJMhP #Simular #AGI #AI #Assistant Demos > slides! We take the name of SPC Demo Faire seriously. Every presenter at our December 2023 event showed off live demos and products. Follow the builders and check out our recap here: https://lnkd.in/gtgJ9hSj Spotted two passionate students in NeurIPS wearing our newly designed t-shirts for "Backdoor in deep learning" workshop. If you get the joke on their backs and find it funny, attend the workshop on Friday Dec 15 to receive a complimentary t-shirt. Regardless of whether you get the joke or not, you can still join the workshop where incredible speakers and panelists are going to seriously share their insights: Bo Li (University of Illinois Urbana-Champaign), Ruoxi Jia (Virginia Tech), Adam Dziedzic (CISPA), Florian Tramèr (ETH Zürich), Yiming Li (Zhejiang University), Baoyuan Wu (The Chinese University of Hong Kong), Dawn Song (University of California, Berkeley), Franziska Boenisch (CISPA). And of course, thanks to our awesome organizers who finally made this happen: Khoa Doan , Aniruddha Saha, Anh Tran, Yingjie Lao, Kok-Seng Wong, Ang Li, Haripriya Harikumar, Eugene Bagdasaryan , Micah Goldblum, Tom Goldstein. Workshop website: https://lnkd.in/gcw_TSVK Live site: https://lnkd.in/ghsN9yi9 #NeurIPS #DeepLearning #AI #MachineLearning #Workshop #Conference #Research #Technology #DataScience #aisafety #backdoor The responsible development of AI requires a deep understanding of current AI systems' weaknesses. You are invited to join the "backdoors in deep learning" BUGS workshop at NeurIPS this Friday. We will have a great lineup of speakers and panelists to discuss the "bugs" of AI. Link: https://lnkd.in/ghsN9yi9 Workshop: https://lnkd.in/gcw_TSVK #AIbugs #DeepLearningBugs #AIWeaknesses #NeurIPSWorkshop #BugHunting #ResponsibleAI #AIResearch #AIWorkshop #AICommunity #EthicalAI Simular's cofounder Jiachen Yang was featured in the first photo, presenting our latest product and technology in front of hundreds of audience, and of course, with a newly designed Simular t-shirt :) Wanna see some of Simular's latest technologies? Join SPC demo faire on December 6! It's exciting to see AI enthusiasts joining the force. We are hiring, btw. If you are interested and want to skip the long wait for interview, we can collaborate starting today. Let’s meet in the code! Congratulations to my wife Grace who has recently been featured by artnet for her dedication to promoting women artists globally. She organizes annual exhibitions at the Google campus and also runs her own gallery Insel Island Art Community with an art-inspired vacation house in Sea Ranch, California. https://lnkd.in/gG5wd8Wz #WomenInArt #Inclusivity #SeaRanchCalifornia #InselIslandArtCommunity Talking is easy. Building something that actually works is 10x harder but 100x more important. OpenAI folks, I feel you. I understand the challenges that come with a leadership change, often stemming from a misalignment of the mission. We are a boutique AGI shop coming out of DeepMind, with a very special vision on product and research. Come talk to us about the potential to co-invent the frontier of future technology and make a positive impact on the world. Together! #AGI #FutureTech #PositiveImpact #Innovation #DeepMind #BoutiqueShop #ResearchVision #GameChanger #TechRevolution #Collaborate