Eugenia Kuyda @ ekuyda, the founder of @ wabi, who previously raised $20 million to build a personal software, made a popular "7 predictions for consumer AI". Let's take a look at how she predicted: 1. Screensless artificial intelligence devices will fail -We mainly use mobile phones to passively obtain information: opening information streams, scrolling browsing, and watching videos. -Voice is great as an auxiliary interface, but terrible as the primary interface: -It's almost impossible to use it unless you're alone. -You must initiate every interaction proactively (otherwise it will appear too offensive). -You can't just glance at your phone and read messages or updates. -There are more -The functions of mobile phones are too powerful. It's difficult for us to get rid of our dependence on information flow, let alone charging another device. Devices that always listen cannot succeed. -In fact, our lives are not interesting enough and not worth constantly recording - and for online/work meetings, you may prefer to go back to the Granola you already have (without the need for another device). -Most things are not important. And the truly important things are precisely what you dare not record - important days, crucial job interviews, long conversations with friends, psychological therapy. -In addition, a Gmail account allows you to learn more about a person (what they bought, where they went, what they subscribed to), rather than listening to their out of context chat records all day long. -When we shift from audio to video, the situation may change - but that way, the issue of social accessibility may become more complex. 3. The mini program will unlock user generated content (UGC) personal software -Mini programs are to software what short videos are to content. -Developing a complete application is too difficult: authentication, integration, database, and so on. But almost anyone can describe an idea, a process, a screen. -Nowadays, some applications take more than 10 minutes to complete the registration process and attract users to subscribe - in this case, developing your own mini program is actually faster. -This will unlock user generated content (UGC) software - and thus give birth to the first truly consumer super application in the United States. By 2030, two large general-purpose artificial intelligence chatbots will emerge. Nowadays, we have ChatGPT (and similar chatbots) and a large number of chatbots targeting specific fields. By 2030, we will have two large general-purpose chatbots: ChatGPT like assistant: focuses on knowledge retrieval, search, and tasks. An AI friend: Committed to helping you live the best life possible. They cannot coexist in the same product: one must be predictable and always respond to your queries; Another should have autonomy (to help you thrive), bring you surprises, and build relationships with you. People must be reliable. Another person must feel the existence of life. 5. The effectiveness of application marketing is dead Nowadays, too many application studios focusing on effect marketing in Eastern Europe, Türkiye, Israel and other places are conducting the same CAC/LTV arbitrage: pay close attention to any application that gets attention on the App Store, develop a copycat version within a few days, and then test it with hundreds of creative and marketing funnels. The product itself is no longer important - nowadays people even purchase subscription services before seeing the actual product. This makes all profit margins zero. You can use paid customer acquisition to increase sales, but it cannot be used as a business model. The consumer product with the fastest annual recurring revenue reaching $1 billion may be an AI webcam girl Once the cost of real-time video generation drops to a level sufficient to charge users, it will become the fastest-growing consumer product in history. With unlimited personalized customization and 24/7 chat functionality, OnlyFans will appear insignificant in the future. Whoever solves AI discovery first wins the final victory. Ordinary people don't know how to handle blank text input boxes besides chatting and searching - these are all functions provided by the command line. These application scenarios are very large, but there are hundreds of other application scenarios, and they are all locked behind the command-line interface, which is currently the main AI interface. Whoever can find a way to unlock all these application scenarios beyond search and chat - showcasing what artificial intelligence can do instead of prompting users - will become the next major consumer platform. I quite agree with her views on 4, 3, 6, and 7 mentioned above Note: @ wabi AI Application World's YouTube "or" App version of YouTube "is a personal software platform. It allows anyone (even without programming skills) to quickly generate, discover, remix, and share personalized mini apps through natural language prompts Zero code creation: Simply enter a description in the chat box (such as "Create an app for reminding a pixel style cat to drink water"), and AI can instantly generate a complete mini program with UI, interaction, and data. Social+Remix Culture: Like YouTube or TikTok, you can browse other people's apps, like, comment, and directly "remix" other people's works (check the prompt logic and modify it into your own version), promoting creative dissemination. Highly personalized: solving the problem of traditional apps being "one size fits all", the generated software is completely customized around your life and needs (such as habit tracking, personal CRM, mini games, fitness tools, etc.).