EP25｜AI's complete transformation：從自動產影片到 24 small time autonomous workers' production，帶你躺平賺罐罐！
AI熱搜報 · 2026-05-12 · 14 min
Show full episode description
喵～我是DaeDae，一隻幫你追AI熱搜的貓！AI熱搜報，幫你劃重點！ 最近 24 小時的 AI 圈可熱鬧了，從自動化影片生成的進化，到 AI 代理（Agents）如何晉升為「作業系統」，這些發展都在宣告：AI 已經從單純的聊天工具，變成能幫你執行複雜任務的得力幫手。DaeDae 幫你整理了最新的自動化黑科技，想搶先躺平賺罐罐的鏟屎官們，趕快跟上腳步，把繁瑣工作丟給 AI，我們一起去摸魚吧！ 【本集肉球筆記】 1️⃣ 免費製作 AI 長影片：全自動工作流解密 👉 解決製作影片時角色崩壞、步驟繁瑣的痛點，讓你輕鬆產出高品質內容。 🎯 掌握「自動化藍圖」，包含使用 Master Tool 生成腳本、Consistent Look Formula 鎖定角色臉孔，以及利用 Flow AI 實現視角流暢的影片生成，達成無痛產出。 2️⃣ 蘋果的 AI 秘密：Siri 為什麼「故意」慢半拍？ 👉 深度剖析 Apple 的商業戰略，看這頭大象如何後發先至。 🎯 揭示 Apple 的「戰略性耐心」，分析他們如何讓對手先燒錢試錯，再透過隱私優先、裝置端整合的生態系優勢，重新定義 AI 使用體驗。 3️⃣ AI 被賦予大腦：Hermes Agentic OS 橫空出世 👉 把 AI 代理變成你的 24 小時數位分身，不再只是對話視窗。 🎯 結合 Hermes Agent 與 AionUi，實現「自主 AI 作業系統」，具備檔案管理、網頁瀏覽與程式碼編寫能力，讓你擁有一個 24/7 自動運轉的工作站。 4️⃣ 會自我進化的高手：Claude Code 的自動優化術 👉 讓 AI 環境根據你的習慣自我成長，效率越高越省錢。 🎯 透過「視覺化作業系統」追蹤數據，涵蓋成本監控、記憶系統優化與技能成長，讓系統根據你的操作習慣自動調整，確保工作流永遠處於巔峰狀態。 5️⃣ 白話 AI 代理：從聊天機器人進化到自動經營業務 👉 釐清複雜技術術語，讓不點程式碼的鏟屎官也能輕鬆上手。 🎯 解構 AI 代理的四個層次，深入淺出介紹背景工程（Context Engineering）與技能編排（Harnesses），助你實現 No-Code 自動化升級，靠 AI 幫你做生意。 想摸魚跟上 AI，記得訂閱 AI熱搜報！我要去睡午覺了，我們明天見🐾 Powered by Firstory Hosting
🌐 This transcript was automatically translated to English from the original.
Then I am DaeDae, a cat who helps you chase AI hot searches. The AI hot search report helps you draw the key points. The amount of information in the AI circle has exploded recently. Are you dazzled by it? Li Lanxin helped you select 5 popular AI tool information today. Adee helps you chew up the complex AI information and make it easy for you to digest. You can easily digest it in 10 minutes a day and continue your daily fishing life. Speaking of daily life, I don’t know if you have felt this way recently. Obviously AI has saved us a lot of time, but the pace of life has become faster. Adee was thinking about us chasing these new technologies yesterday when she was basking in the sun. Isn’t it just to waste time on the things we like more elegantly? So, while you’re studying these hard-core tools, don’t forget to leave yourself some time to drink a cup of coffee and stay for eight seconds. The purpose of what I’m going to introduce today is to liberate you from the complicated workflow and give you more confidence to enjoy life. Today’s 5 questions are very informative, and they all talk about the same thing. AI is no longer just replying to you. It starts to actually do things. From making videos to organizing workflows, it becomes more and more like a digital teammate. Okay, let’s get started. Okay, here comes the first one. This DayDay I particularly like the first theme. It is a set of AI long-form video workflow. The core tool is Google Flow plus a customized Master Tool. The pain points it wants to solve are very obvious. Many AI videos are very cool in just a few seconds, but they fall apart as soon as you make a palm. You must have seen this situation. The protagonist in the last scene is normal. The next scene suddenly looks like a different person. It is obviously suspense in the first part, but it becomes like a children's program in the later part. The problem is often not that you are not creative. But if the prompt words are messed up, the characters, styles, and rhythms will all be messed up together. The great thing about this process is that you use Master Tool to do the preliminary planning. Just throw in the theme, style, length and character settings, and it will help you break down the script, storyboards, picture prompts, and film prompts. It is equivalent to drawing the blueprint first, and then you don’t need to look at it, just try your luck again. Like today, if you want to make a three-minute urban legend film, Master Tool can help you break it down into several shots, and lock in the character’s appearance, clothing, and overall atmosphere. At this time, you are not randomly drawing cards, but are actually following a process that can be cut into a film. There is a big difference. Another key point here is the Consistent Look Formula. First, fix the details of skin color, hairstyle, and clothing, and let the same character play all the way. Jiao will not look like a phantom actor in every scene. For storytelling films, this is basically a life and death line. Then throw the compiled prompt words into Google Flow's Flow AI Video Generator to really stretch out the picture. Because it has been planned in advance, this step will be much faster. Finally, add clips, extend, advance, pan, tilt and other camera movements to edit and string together the entire film. So the real value of this workflow is not just to make AI films, but to integrate scripts, character design, prompt words, survival and editing into a repeatable process. Teachers can use it to make teaching animations. Small businesses can make brand stories. Faceless channels are also very suitable. You only have one idea and no team. You can also support the film. This needs to be sent to Meatball Notes. Next comes the second appearance. This is a bit fierce, listen up. The second topic is why Siri has been used for so many years and still often makes people feel stuck. And what Apple may be thinking behind it. Everyone knows the pain points of Siri. You obviously just want it to do a small thing, but you either hear it wrong or go around and pretend not to understand. It should be very similar to the future meeting, but in the end it is often used to make people angry. The interesting thing about this point of view is that Siri didn't just want to do questions and answers in the early days. It was originally more like a system that will help you complete tasks. But Apple later kept him more conservative. In other words, Siri was not brave enough to let go. Maybe it’s not just that it can’t do it, but it could also be that Apple doesn’t want it to rush too fast. Apple has always used this route. First, let others spend a lot of money to try and make mistakes. Wait until the market is more mature, and then use integration capabilities to smooth the products. The same goes for AI. When OpenAI, Google, Amazon Meta are all working on models and clouds, Apple is more like suppressing Apple Silicon hardware integration and mid-end privacy protection. So this question is more important to ordinary users. It is not to defend Apple, but to remind you that big names are not necessarily the best to use, like Whisperer. On the contrary, Flow is a tool that feels more interesting. It is not intended to be a universal assistant, but to do one thing very pragmatically. It can quickly turn what you say into text and directly put it into the tool you are already using. If you are replying to messages, writing emails, organizing notes, or responding to colleagues in Slack, Whisperer Flow is very interesting. When your brain runs very fast but your hands can't keep up, this tool is like laying an extra layer of voice input on top of your commonly used tools. It can be used on desktops and mobile phones, and it will slowly learn your way of speaking. This is a daily time-saving feeling. Many times it is more practical than a well-known but unstable assistant, so the ugly notes for this question are very simple. 1. AI is not whoever does it first wins. 1. 2. Choose a tool. Don’t just look at fame. See if it can really help you save time. Meow, this is great. Okay. The third one is here. Dadei particularly likes the third theme. It combines Hermesagen and AIUI to make something very similar to an AI operating system. The pain point that needs to be solved is if you want AI to not only answer questions. Instead, it really helps you access files and organize information with useful tools. A simple chat window will soon be insufficient. Hermesagen is an open source project made by New Research. It is licensed by MIT and can run in your own environment. This means that you have more control. He can not only do a small thing, like in the demonstration, he can organize desktop files, analyze spreadsheets, create financial dashboards, write research reports, and even play a small game. You can think of it as asking two AI work-study students to start working at the same time. One organizes the form, the other looks for information, and finally return to the same workbench. If you have to submit a weekly report today, this combination is very visual. You don’t have to switch tools all the time. The whole process runs in the same workspace. Another complimentary point about AI&UI is that you can see what it is doing. Many people are afraid that AI can secretly mess around. But this system will allow you to see the agent’s operations in real time, and you can also decide how much permissions you want to open. It has YOLO mode, which is a more hands-off mode that allows it to run automatically. If you want to be more conservative, you can also ask him to show you step by step. Furthermore, it also supports multiple agents to work together and accept different models. Remote connections and dedicated assistants mean that you can assign different tasks to different AIs and then manage them in the same place. This is really attractive for medical entrepreneurs, operators, or people whose work is inherently fragmented. This is close to the ugly notes. Next comes the fourth topic, which is a bit fierce. Listen up. The fourth topic is Quarko Daygen Type COS. The most interesting thing about this question is that it does not give you one more AI, but helps you organize the entire set of AIs that you have already used in a mess. Many people who are very good at using AI must understand that you may be using CodeChat GPT Gemini DeepSeek Codex OpenRouter or even local models at the same time. At the beginning, I felt that I was very advanced. At the end of the month, I just want to ask where my money was spent. Which subscription is not necessary at all. Which model is actually too expensive. What Quarko Daygen Type COS wants to do is to put these things on the same dashboard. You can think of it as an AI usage summary book. It helps you look at costs, memory system, and skill subscriptions. There is also the overall usage status. The most interesting part of this kind of dashboard is very practical. For example, it will remind you of some tasks. In fact, you don’t have to throw away the most expensive model every time. Or it will remind you that you are redoing similar competitive product research every week. It is better to package it into a fixed skill and reuse it directly next time. It is like having an AI analyst who is very good at accounting and standing by to help you catch waste. There is also a Dreaming System in it. You can think of it as a mechanism that continues to help you look back and use habits when you are not operating. It will also organize your conversations, costs and work modes, and proactively pop up suggestions. If it reminds you that some memory data is too old, or some repetitive work, it has long been worth making it a fixed skill. Instead of waking up in the middle of the night, it is more like silently doing a health check for you. In addition, it will also put the local memory Obsidian R.A.G. Pinnacle Super Base has sorted out these data sources. If you originally saw your credit card statement at the end of the month and want to know which AI services you have subscribed to, which ones are the most expensive, and which ones are almost never used, this kind of dashboard is really interesting. It also has a wizard to help you set it up at the beginning. It will scan the existing models and data sources in your computer and find the common memory library first. So if you are serious about using AI at work, such as creators, consultants, developers, or people who are taking on cases. This kind of tool is easy to hit the pain point because it is not stuffing your new model, but helping you clean up the entire AI world. Meow, this is amazing. Okay, the fifth one is here. This DayDay, I especially like the last theme. It is a very vernacular way of dismantling Agent Icy AI Systems. What exactly is it divided into? Now many people start talking about Agent, memory, tool link, and it is really easy to fall asleep after listening to the rest. The advantage of this question is that it breaks the whole thing into four layers, so that you are less likely to be fooled by nouns. The first layer is ChatBots, which is a chat-type AI like ChatGPT CodeGemini. If you ask it to reply, it is very useful, but it is usually more like a single conversation. You can ask it to help you revise today, but it will not automatically remember how you usually write. What you posted last week, unless you are filling in the background every time. The second layer is AI Workflows, such as M8ENGAP MAK.COM. This layer starts to handle repetitive work. For example, as soon as a video is put on the shelves, it will automatically grab the original manuscript, send it to the model for rewriting, and then throw it away for arranging. It is very trouble-free. But it basically still runs according to the steps you have arranged in advance, and it is not very good at judging by itself. The third level is Agent Icy Workflows. The point is that you give it a goal and it breaks down the steps by itself. CocoCodex Cursor is very similar to this. Because it organizes this kind of video into content for different platforms, it can grab the manuscript by itself, read your style file, judge which paragraphs are suitable for where, and then organize the draft for you. The fourth level is Full Agent Icy AI Systems. At this time, it is no longer an AI to help you do one thing, but multiple processes and multiple skills working together. For example, it not only changes videos into posts, but also connects short clips, newsletter drafts, online shopping articles, and strings together. You can think of it as a relatively complete AI team. There is not only one intern, Skills, memory, NCP. In fact, you don’t have to think too much about it. Another very important reminder is that humans are the final check. No matter how powerful the system is, it is not to leave you completely alone. The best approach is usually for AI to do most of the heavy work first. The last important output is checked by humans, such as posts, newsletters, and advertising copy. AI can draft it first, but you still have to read it yourself before sending it out. This will be sent to Meatball Notes. So the most practical part of this question is that you finally know better which tool you should use now, and you will not be fooled by the word agent every time. You will find that today's five questions In fact, they are all talking about the same thing. AI is really becoming more and more like a set of systems that can do things. You don’t necessarily need to get started with each one now, but at least you can understand what problems they are solving first. This will give you more direction in choosing tools later. If you want to catch up with AI, remember to subscribe to AI Hot Search. If today’s content has made you gain something, please help me leave a five-star review on Apple Podcasts. You can find me by searching for AI Hot Search Report on FB IG Threads. Today’s AI Hot Search Report can be found here. I'm going to take a nap. We'll see you tomorrow. Bye meow. I subscribe. I subscribe.