EP17｜AI body flatness and real battle：Vibe Coding 席捲opening space，一鍵生成長影音的摸魚神器！

I am DaeDae, the cat who has been chasing AI hot searches for you. The AI hot search report will help you identify the key points. I am tired from going to work every day. How can I have time to test so many new AI tools? Don’t worry. Today I have selected the five most popular AI tool information for you. Let AI enter your daily life for ten minutes a day, and you will become a fishing master without knowing it. Today’s episode is very suitable for people who are commuting while thinking about what AI has to do with me. Because these five topics are not novel and will disappear after reading them, but they will really affect how you work, how to create content, and even how to turn ideas into finished products. Okay, let’s get straight to it. Before getting into today’s tool post, DaeDae would like to share some life observations with you. Recently, a friend complained to me, saying that he felt that the pace of life today is so fast that it almost makes people dizzy. The amount of information received every day is like a waterfall. Even I feel the same way. Although I am a cat, I am in the era of the explosion of AI. Information anxiety is really a disease of modern civilization. But I also found that when we learn to hand over repetitive and trivial things to AI, we can use the remaining time to daze, drink a cup of coffee, or even just look out the window and relax. The feeling of regaining control of life is really great. I hope today’s sharing can help everyone save more time to embrace real life. Let’s take a look at the first one, Five Coding. When you hear this word in turn, you may think, what is this? To put it bluntly, it means that you don’t have to use a bunch of programming grammar first. Instead, clarify the requirements first and let AI help you build the program all the way. So it became popular not because everyone suddenly became engineers, but because many people who were stuck outside the door are now finally able to get in. For example, many people now use Quad, Codex, Gemini, Daha, Copilot, Clocko to directly describe the requirements in natural language. Ask AI to help you write the program code, modify the program code, and repair it all the way until it runs. What does this feel like? Much like when you wanted to make something in the past, you had to first learn to hold a hammer, sentences, and food. Now you speak out the image in your mind first. Suddenly there is an assistant beside you who will not be tiring to help you set the stage. If you are a small businessman, a creator, or just want to make your own gadget, this change is really impressive. In the past, it was often either too expensive or too troublesome and finally gave up. It is different now. You speak human language first. AI helps you make a prototype first. The threshold is much lower. There is an example in the manuscript and it is very touching. Because the warehouse owner did not buy that kind of expensive and hard warehousing system. Instead, he used AI to help him create a delivery process automation tool. For example, batch processing orders, inventory shortages, etc. can all be handled according to his own process. This is very different because many small companies do not want to systematize the process. It is a template on the market, which does not look like they are working at all. Another example is a designer with no technical background at all. They relied on Vibe Coding to create a complete application and finally put it on the App Store. In the past, to get to this point, you usually had to find an engineer, find an outsourcer, or study hard on your own for a long time. Now that distance has been greatly shortened. No wonder many people feel like "Hey, I really made it" when using it for the first time. But don't rush too fast. Google Cloud's Adios Money also reminded that Vibe Coding is very suitable for exploratory ideas to make the first version. But actually making a formal product is another matter. You need to test, confirm the needs, check the quality, and be responsible for the results. Especially if this thing is not just for you to play with, but for customers, companies, and even many users to use it together. Stability is much more important than speed. So in the final analysis, AI does not make engineers less important. Instead, it is more like changing the content of the work. Strong engineers from outside are not necessarily the ones with the fastest hand skills, but the ones who are best at determining requirements, saving results, and grasping problems. AI can quickly release the first version, but to become a product that can be officially launched, it still has to rely on the original finishing. This needs to be sent to Meatball Notes. Again, this question is very practical. It is the five ways to make money with AI in 2026. Let me condense it into one sentence for you. Don’t keep thinking about whether AI can make money. It’s more important to choose a path first. The first is to make AI personal brands and monetize through sponsorship and cooperation. This is very suitable for people who like to organize information and focus on key points. You don’t need to invent models or start your own company. You can use Quad to help you organize updates and break down the key points, and then make short videos or social posts. When you slowly gather a group of people who are interested in AI, AI startups may come to you for cooperation. The second option is faceless content, which is the AI Avatar mentioned here. AI avatar accounts are like helping TikTok Shop to make product videos, operating AI Avatar accounts or directly helping the boss to create digital avatar content with 11labs voice copy. This path is very suitable for people who don’t want to show their faces but are willing to create content. But a reminder, don’t imagine full automation at the beginning. First manually test whether the content has the original view, and then add automation. This is faster. The third one is the company’s Cloud corporate training, that is, you teach the company how to really use Cloud in work. I think this is very pragmatic. It’s not that many companies don’t want to use AI. Everyone knows that they need to use it, but they don’t know how to actually put it into their daily work. If you can teach them, you can teach them how to use Call code, design project settings with prompt words, and support collaboration and scheduling of tasks. As long as you can really save the team a few hours every week, the company will be more willing to pay. The fourth item is to use Vibe Coding to do a single function. Note that the key is not to make it bigger, but to only do one function. Like Call AI, take a photo of food and directly help you estimate calories. This product is easy to understand and easy to spread. It is also easier to create the first batch of users. Many people fail not because they are not skilled enough, but because they do everything they want in the first version and are unable to explain anything clearly. The fifth article is AI marketing automation. The tools named here include Go High Level ManyChat, ExpandEye, FentonBuster, and Lesshark. You can think of it as helping others to get missed customers back. Like someone leaving a keyword on IG, you will automatically reply to his form. After filling out the form, push the appointment link. Those who did not reply will be contacted again after three days. This process is connected one after another. The boss usually understands it as soon as he hears it, because one less missed list will lead to more business, so you don’t need to memorize a lot of big principles for this question. You just need to remember one thing, choose one and do it for a longer time. If you are better at talking, go for content or teaching. If you are better at arranging processes, do automation. If you like to make products, go for a single-function app. I also dipped into each one. The last one is not raised. Next, if you often feel that chat-type AI is not enough, you should be aware of this question. The third topic is Hermes Agent from News Research. You can first think of it as more than just a chatbot. It is an AI assistant that will always be online, will remember things, and can run tasks by itself. It is usually like ChairGPT or Cloud. In many cases, it is more like opening a window to ask a question and getting the answer. Then it will be like answering the call again when you come back next time. But Hermes Agent takes a different approach. It can be installed on a VPS such as Hostinger, which is to put it in your own server environment. There is a 24-hour online AI in it. You can even interact with it through Telegram. To put it simply, in the past, you had to find an AI. The face-saving AI can stay there to help you take care of things. There are two things that are most impressive about it. The first is scheduling. For example, it helps you organize Hacker News every morning and make it into a PDF and send it directly to your mobile phone. This kind of repetitive work is very suitable for it to be handed over to it. The second is memory. Hermes Agent will keep what it has learned before, including your preferences, working methods, and common processes, so that you can do similar tasks next time. It is not starting from scratch. It will be more like the kind of colleague who gets more familiar with you. If you have to check the same data and sort out the same information every day, it is actually like a research assistant who never leaves work. This is where it is different from OpenCode or general chat tools. The focus is not just which model is smarter, but whether the entire working method is connected. Mes Agent itself supports many models and can also connect to Code through API, so you get not only model capabilities, but also resident memory scheduling and the flexibility to control the data by yourself. Several demonstrations in the original manuscript are very graphic. For example, it can automatically capture data, make clean products and organize PDFs. Keep an eye on the product changes of OpenAI and Farkick Gemini Groke every day. It can also help creators remember the tone of the channel and analyze past content. The more you write, the more you write. For people who check information, organize information, and do the same thing repeatedly every day, this is really different. Of course, Hermes Agent is not the kind of toy that can be understood instantly when you click it. You have to accept one thing a little bit. You are not playing with a chat window. You are raising an AI system that will continue to work. But if you have always wanted an assistant that will do things silently in the background, then it is really worth watching. If you are working on content, you can really pay attention to this question. The fourth theme is Wary AI. What it wants to do is very simple. It is to put the tools that creators usually know about one thing into the same place as much as possible. It is very interesting to win. Use one platform to think about titles at night, use another to make pictures at noon, edit videos at night, change subtitles, supplement translations, and jump to other places. The browser has so many paginations that it’s as messy as overturning the feed. Just copying and pasting is enough. What Wary AI wants to solve is the problem that this tool is too fragmented and the work is constantly interrupted. The first case in the original manuscript is to use the old title to redo it. This is very similar to what happens in the real world. Many creators do not start from scratch, but have an old thing on hand that looks a bit outdated and want to make it cleaner and more professional. You throw the original materials into Wary AI and add the style you want with reference pictures. You can fix it one by one. You can always go back and forth to find someone to correct it. It is much faster to try it yourself. The second way to use it is to use it as a YouTube title assistant. It does not just throw ten titles to you, but also tells you which one is more attractive and which one is easier to find. Why are they arranged like this? I think this is very important because many tools will only accompany your brainstorming but will not help you push forward. The third is to make thumbnails. You upload the image, specify the size, clarify the feeling you want, and then slowly adjust it to a more impactful version. This process is suitable for people who create pictures, because the thumbnails look like a trivial matter, and it really takes a lot of time to do. Finally, there is another function that I think is low-key but very practical, which is to organize a large paragraph of text into a visual summary that is easier to understand. This is very convenient for proposals and planning. So the attraction of Wary AI is not just a single function, but that you can connect ideas, vision, and optimization in the same place. For the parliamentary team, there are fewer platforms to jump to, and the pace of work is much different. Of course, Wary AI will not win over specialized tools in every aspect. Some models cost more points, and whether the prompt words are written well or not will also affect the results. If you are originally a member of a team or a small team, and you want to continue writing, making pictures, and changing content all the way without interruption, then its sense of integration is very attractive. It is not that each item is the strongest, but the overall execution is smoother. The last one is more like the milk source savior. That is VocalML. What does it focus on? In a word, it helps you compress the long process of making long videos into a process that is easier to get started. If you run your own channel, you will definitely shake the tone. It’s not that you don’t have ideas. From the script to the finished film, the dozen or so workers will really make you roll your eyes. Pictures, AI dubbing, scenes, and edited subtitles can all be exported in the end. You can also choose different models, such as LTXR and ELEVENLAPS, and you can also adjust settings such as video ratio, length, transitions, text overlays, and subtitles. For novices, the most impressive thing is that you don’t have to rethink every scene. For example, if you want to make an 8-minute product explanation video today, it can help you pull out the picture and narration skeleton first, and then slowly modify it later. This difference is really big because you don’t have to start from a blank piece of paper, and it is not locked after completion. After it is generated, you can go back and modify the clips, modify the sound, and adjust the content, because if a fully automatic video cannot be repaired at all, there will usually be a canning position soon. VocalML at least leaves a space for you to modify things to be more like your own film. In addition, it has a free quota. It is friendly to students, small creators, and small businesses. You don’t have to pay a heavy cost at the beginning. You can first try to see if the long video process can run. To put it simply, it may not guarantee that you will make a masterpiece in one piece, but if you really save a lot of time, it will be very impressive. Oh, this is amazing. Looking back, these five topics today are actually talking about the same thing. Many times, what really makes people feel is not that AI can talk about it, but whether it can help you connect the troublesome steps. Five ways to make money from Vibe Coding to AI. From assistants like Hermes agent that can do things by themselves to tools like Wary AI and VocalML that put the creative process into the same place. To put it bluntly, it is really easier to see if you have used it up. So if you take away only one important point today, I would say don’t rush to chase everything. Pick a direction that you can really use and are willing to keep trying, and slowly turn it into your work habit. This way, it will be less popular in three minutes, and it will be easier to really see the results. If you want to catch up with AI, remember to subscribe to the AI hot search report. If today's content has gained you something, then help me leave a five-star review on Apple Podcasts. FB IG Threads So I can be found in today's AI hot search report. I'm going to take a nap. See you tomorrow, bye meow.

我是 DaeDae,一直幫你追AI熱搜的貓, AI熱搜報幫你劃重點 每天上班都已經夠累了,哪有時間去測那麼多AI新工具,別擔心 今天幫你精選了五個最熱門的AI工具情報 每天十分鐘讓AI走進你的日常生活,讓你不知不覺成為摸魚大師 今天這集很適合那種一邊通勤,一邊想著AI到底跟我有什麼關係的人 因為這五個主題不是看完覺得新奇就沒了,而是真的會影響你怎麼工作 怎麼做內容,甚至怎麼把點子變成成品 好,我們直接開抓 在進入今天的工具大補帖之前, DaeDae想先跟大家分享一點生活小觀察 最近我有個朋友跟我抱怨,說他覺得 現在的生活節奏快到快要把人轉暈了 每天接收的資訊多到像瀑布一樣 即使我也深有同感,雖然我是隻貓,但身處在AI爆發的時代 資訊焦慮真的是一種現代文明病 不過我也發現,當我們學會把重複、瑣碎的事情交給AI處理後 剩下的時間拿來發呆,喝杯咖啡,甚至只是看著窗外放空 那種重新拿回生活主導權的感覺真的很棒 希望今天的分享能幫大家省下更多時間去擁抱真實的生活 先來看第一個,Five Coding 依次聽到這詞你可能會想,這是在幹嘛 白話講,它就是你不用先揮一堆程式語法 而是先把需求講清楚 讓AI幫你一路把程式做出來 所以它紅不是因為大家突然都變工程師了 而是因為很多本來卡在門外的人 現在終於進得去了 像現在很多人會用Quad, Codex, Gemini,Daha,Copilot,Clocko 直接用自然語言描述需求 請AI幫你寫程式碼,改程式碼,一路修到能跑 這種感覺很像什麼? 很像你以前想做東西,得先學會拿錘子、句子、糧食 現在是你先把腦中的樣子講出來 旁邊突然多了一個不會含累的助手 先幫你搭第一板 你如果是小商家、創作者 或者只是想做個自己的小工具 這種變化真的很有感 以前常常不是太貴,就是太麻煩 最後乾脆放棄 現在不一樣了,你先講人話 AI先幫你做出雛形 門檻一下就低很多 原稿裡有個例子就很有感 因為倉庫老闆沒有去買那種又貴又硬的倉儲系統 反而用AI幫自己做出貨流程自動化工具 像是批次處理訂單 缺庫存這些都能照自己流程來 這個差很多欸 因為很多小公司不是不想把流程系統化 是市面上的範本,根本不像他們在工作 另一個例子是完全沒有技術背景的設計師 靠Vibe Coding做出完整應用程式 最後還上架到App Store 以前你要走到這一步,通常得找工程師 找外包或者自己苦學很久 現在那段距離被大幅縮短 難怪很多人第一次用會有那種 欸,我真的做出來了的感覺 喲,這個厲害了 不過先別衝太快 Google Cloud的Adios Money也提醒過 Vibe Coding很適合拿來探索式想法做第一版 但真正要做正式產品還是另一回事 你要測試,要確認需求,要把關品質,要對結果負責 尤其如果這東西不是你自己玩玩而已 而是要給客戶、公司、甚至很多使用者一起用 那穩不穩比快不快重要太多 所以說到底,AI沒有讓工程師便不重要 反而更像是把工作內容改掉了 外來厲害的工程師不一定是手打最快的人 而是最會下需求、省結果、抓問題 走品質的人 AI可以很快伸出第一版 但要變成能正式推出的產品 還是得靠原來收尾 這個要寄進肉球筆記 再來這題很實際 就是2026用AI賺罐罐的五條路 這對我幫你濃縮成一句 不要一直想,AI到底能不能賺錢 先選一條你做得下去的路比較重要 第一條是做AI個人品牌靠贊助合作變現 這條很適合本來就喜歡整理資訊講重點的人 也不用發明模型 也不用自己開公司 你可以用Quad幫你整理更新 拆重點 再做成短影音或社群貼文 等你慢慢聚到一群對AI有興趣的人 AI新創就有可能來找你合作 第二條是不露臉內容 這裡講的AI Avatar也就是 AI分身帳號像幫TikTok Shop做商品影片 經營AI Avatar帳號或者直接幫老闆做數位分身內容 配上11labs的聲音複製 這條路很適合不想露臉但願意做內容的人 不過提醒一下 不要一開始就幻想全自動 先手動測內容有沒有原看 再來加自動化 這樣比較快啦 第三條是企業的Cloud企業培訓 也就是你去教公司怎麼真的把Cloud用進工作裡 這個我自己覺得很務實 現在很多公司不是不想用 AI是大家都知道要用 可是不知道怎麼真的放進每天工作 你如果會教就可以教他們怎麼用Call code 把提示詞設計專案設定 反對協作排成任務這些東西帶進去 只要你能讓團隊每周真的省下幾小時 企業就會比較願意買單 第四條是用Vibe Coding做單一功能 注意哦關鍵不是做大 而是只做一個功能 像Call AI這種拍一張食物照 就直接幫你估熱量 這種產品好懂好傳 也比較容易做出第一批使用者 很多人失敗不是技術不夠 是第一版就想什麼都做 結果什麼都講不清楚 第五條是AI行銷自動化 這裡點名的工具有Go High Level ManyChat,ExpandEye,FentonBuster,Lesshark 你可以想像成幫別人 把漏掉的客戶解回來 像有人在IG留關鍵字 你就自動回他表單 填完表單再推預約連結 沒回的人三天後再追一次 這種流程一接起來 老闆通常一聽就懂 因為少漏一個名單 就可能多一筆生意 所以這一題不用記一堆大道理 你只要記住一件事選一條做久一點 你比較會說就走內容或教學 你比較會排流程就做自動化 喜歡做產品就去做單一功能 App也每條都沾一下 最後一條都沒揚起來 下一個 如果你常常覺得聊天型 AI不夠用 這題你應該會有感 第三個主題是Hermes Agent 來自News Research 你可以先把它想成 不只是聊天機器人 而是一個會一直在線 會記得事情 還能自己跑任務的 AI助手 一般像ChairGPT或Cloud 很多時候比較像 你打開視窗 問一題拿答案關掉 下次回來又像重新接話 但Hermes Agent走的是另一條路 它可以自己架在 Hostinger這類VPS上 也就是放在你自己的伺服器環境裡 面存一個24小時在線的AI 你甚至可以透過Telegram跟它互動 簡單講 從以前你要去找AI 面存AI可以待在那邊幫你顧事情 它最有感的地方有兩個 第一個是排程 像每天早上固定幫你整理Hacker News 做成PDF直接傳到手機 這種重複工作很適合交給他 第二個是記憶 Hermes Agent會把之前學到的東西留下來 包含你的偏好 工作方式 常用流程 下次再做類似任務 它不是從零開始 會比較像那種 越做越熟的同事 如果你每天都要查差不多的資料 整理差不多的資訊 那它其實很像一個不會下班的研究助理 這也是它跟OpenCode或一般聊天型工具 不太一樣的地方 重點不只是哪個模型比較聰明 而是整套工作方式有沒有接起來 而Mes Agent本身支援很多模型 也可以透過API接Code 所以你拿到的不只是模型能力 還多了常駐記憶排程 還有自己掌控資料的彈性 原稿裡幾個示範都很有畫面 像它可以自動抓資料 做淨品整理PDF 天天盯OpenAI andFarkick Gemini Groke的產品變化 也可以幫創作者 記住頻道語氣 分析過去內容 越寫越像你 對那種每天都在查資料 整理資訊 重複做一樣事情的人來說 這個真的有差 當然Hermes Agent 不是那種 點開就秒懂的玩具 你要稍微接受一件事 你不是在跟一個聊天視窗玩 你是在養一個會持續工作的AI系統 但如果你一直很想要一個 會在背景默默做事的助手 那它確實很值得看 這題如果你有在做內容 真的可以注意一下 第四個主題是Wary AI 它想做的是很簡單 就是把創作者 平常懂一個西一個的工具 儘量收進同一個地方 贏得很有感吧 晚上想標題 用一個平台 中午做圖用另一個 晚上剪片 改字幕 補翻譯 又跳到別的地方 游覽器分頁多到像 打翻飼料一樣亂 光是複製貼上就飽了 Wary AI想處理的 就是這種工具太碎 做事一直被打斷的問題 原稿裡第一個案例 是拿舊片頭去重做 這超像真實世界會發生的事 很多創作者不是從零開始 而是手上有一個舊東西 看起來有點過時 想把它改得更乾淨更專業 你把原素材丟進Wary AI 再補上你想要的風格 跟參考圖 就能一板一板修 你一起一直來回找人改 自己先試速度快很多 第二個用法 是拿它當YouTube標題助手 它不是只丟十個標題給你就算了 還會順便告訴你 哪個比較有吸引力 哪個比較容易被搜到 為什麼這樣排 這點我覺得蠻重要的 因為很多工具 只會陪你腦力激盪 但不會幫你往前推一步 第三個是做縮圖 你上傳圖片 指定尺寸 講清楚想要的感覺 再慢慢調成更有衝擊力的版本 這種流程對擀片的人使用 因為縮圖看起來小事 真的做起來超吃時間 最後還有一種我覺得低調 但很實用的功能 就是把一大段文字 整理成比較好懂的視覺摘要 這對提案 整理企劃都很方便 所以Wary AI吸引人的地方 不只是單一功能 而是你可以在同一個地方 把發想、視覺、優化一路接下去 因為議員團隊來說 少跳幾個平台 做事的節奏就差很多 當然Wary AI 也不是每一項都會贏過專門工具 有些模型比較吃點數 提示詞寫得好不好也會影響結果 那如果你本來就是 一員團隊或小團隊 要的是從寫、做圖 改內容一路接下去 不要斷 那它的整合感就很有吸引力 不是每一項都最強 而是整體做起來比較順 最後一個 比較像奶源救星 那是VocalML 它主打的是什麼? 一句話就是 幫你把長影片製作這條抽長工序 壓進一個比較好上手的流程裡 如果你自己經營頻道 你一定抖那個tone 不是沒點子 是從腳本到成片中間 那十幾道工真的會做到想翻白眼 你要想結構做畫面 配旁白接轉場補字幕 常常一支片還沒做完 原先累了 VocalML想做的就是你丟一個想法 或者直接貼一份腳本進去 它就一路幫你伸出影片結構 畫面、AI配音、場景、剪輯字幕 最後還能匯出 它裡面也能選不同模型 像LTXR、ELEVENLAPS還能調整影片比例、長度、轉場、文字覆蓋、字幕這些設定 對新手來說 最有感的是不用每一幕都自己重想一次 像如果你今天要做一支8分鐘的產品解說片 它就能先幫你把畫面跟旁白骨架拉出來 後面再慢慢修 這種差別真的很大 因為你不用從一張白紙開始硬撐 而且它不是做完就鎖死 生成之後 你還能回頭改片段修聲音、調內容 因為全自動影片如果完全不能修 通常很快就會有罐頭位 VocalML至少還留了一個空間 讓你把東西修得更像自己的片 再加上它有免費額度 對學生、小創作者、小商家都比較友善 你不用一開始就被很重的成本 就能先試試看長影片流程能不能跑起來 簡單講它不一定保證你一件就做出神片 但你又完如果真的省下很多時間 那就很有感了 呦這個厲害了 回頭看今天這五個主題其實都在講同一件事 很多時候真正讓人有感的 不是AI都會講 而是他有沒有幫你把麻煩的步驟接起來 從Vibe Coding到AI賺罐罐的五條路 從Hermes agent這種會自己做事的助手 到Wary AI、VocalML這種 把創作流程收進同一個地方的工具 說穿的就是看你用完有沒有真的比較省事 所以今天如果你只帶走一個重點 我會說先別急著什麼都追 挑一個你真的用得到也願意持續試的方向 慢慢把它變成你的工作習慣 這樣比較不會三分鐘熱度 也比較容易真的看到成果 想摸魚跟上AI記得訂閱AI熱搜報 如果今天的內容讓你有點收穫 那就幫我到Apple Podcast 留個五星好評吧 FB IG Threads So AI熱搜報就找得到我了 今天的AI熱搜報就報到這 我要去睡午覺了 我們明天見 掰啦喵