Universal Agent for Any Screen
Press Space to continue
## the confession I have a confession to make. for the past year I've been quietly working on something that I think will make most of what you do on your phone... unnecessary. not unnecessary like "oh that's a nice optimization." unnecessary like "why would I ever do that myself again." let me explain. ## the app industrial complex here's what happened to software. somewhere along the way we convinced ourselves that the best human-computer interaction was... tapping on little colored squares. downloading a different one every time we wanted to do something new. want to order food? use this app. want to book a flight? download this other app. want to check if your friend is free? another colored square. we have literally trained an entire generation to be professional app navigators. and here's the thing that kills me: every single one of these apps does the same thing at the core. they take your intent and translate it into actions. that's it. that's the whole game. "I want thai food delivered" → tap tap scroll tap confirm tap tap done. "I want to fly to tokyo in march" → tap tap search tap filter tap compare tap tap tap tap book. you're the middleware, man. you're the translation layer between what you want and what happens. you're doing the computer's job. ## the vertical agent trap so then the AI boom happens and everyone says "agents! agents will fix this!" and they're right. kind of. what we got instead was... vertical agents. AI for legal docs, for customer support, for sales emails. but here's my problem. my mom isn't a lawyer. my cousin isn't a developer. my friend who works at the grocery store doesn't need an AI for enterprise sales workflows. what do they need? they need something that can just... use their phone in daily life. like they do. but better. ## the realization I remember the exact moment this clicked for me. I was watching my grandad try to book a doctor's appointment online. he's not tech illiterate (he was an engineer) but he's 67 and the healthcare portal was clearly designed by someone who hates elderly people. click here. scroll there. wait that logged you out. try again. wrong date format. oh now you need to verify your email. where did that email go. spam folder. click the link. link expired. start over. twenty minutes. for a doctor's appointment. and I thought: why is my grandad the one adapting to this interface? why is there an interface at all between what he wants and what he gets? he knows what he wants. "book appointment with dr. chen, next week, morning works better.” ## so we built it so we built an AI that can control any screen. any app. any website. any interface. not through special APIs or integrations. just... looking at pixels and clicking and typing. like you do. but faster. and it doesn't get confused or tired, and it doesn't miss the tiny X button that closes the popup. we call it a universal agent. because it's not an agent for lawyers or an agent for coders or an agent for salespeople. it's just... an agent. for anyone. for anything with a screen. ## what this actually means the phone as “a thing you interact with manually for hours every day” is ending. maybe not tomorrow. but soon. the era where you had to learn how to use technology is ending. my grandad shouldn't have to learn how to navigate a healthcare portal. he should just get his doctor's appointment. your mom shouldn't have to figure out which app to download for which airline. she should just get her flight booked. your coworker shouldn't spend their lunch time navigating a reimbursement form. they should just get paid back. ## why we're doing this look. I could be building another vertical enterprise agent. but I didn't get into this to make software for software people. I got into this because I think there's something deeply broken about how most humans interact with technology, and we finally have the tools to fix it. for literally everyone with a phone. that's 5 billion people who could have their daily friction reduced by 10%, 20%, 50%. that's a lot of hours given back to humanity, a lot of frustration eliminated, and a lot of capability unlocked for people who never had access to it. ## the death of the interface so yeah. we’re killing the interface. or at least, we’re making it optional. you can still tap your little colored squares if you want to. some people like driving manual transmission too. but for everyone else - the people who just want the thing to happen without having to learn the interface for the thing - we're building something different. the universal agent. the thing that makes apps obsolete. the end of humans as middleware. it's here. and it works. agi.app