EP44｜Body flatness control！Google Gemini 大更新與免費最強 AI worker manufacturing process！
AI熱搜報 · 2026-05-31 · 15 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！ 鏟屎官們注意啦！最近 24 小時 AI 圈簡直炸開了鍋，AI 代理（Agents）已經不再是科幻片，而是真的要住進你的 WhatsApp 和 Google 手機裡幫你幹活了！從能幫你自動賺罐罐的客服助理，到幫你免費打造最強工作流的神祕組合，DaeDae 都幫你抓回來了。想知道怎麼讓 AI 幫你寫報價單、甚至預防它「不小心」刪掉資料庫嗎？快聽這一集，學會怎麼優雅地摸魚睡午覺！ 【本集肉球筆記】 1️⃣ WhatsApp AI 代理全攻略 👉 幫你將 WhatsApp 變身為自動化銷售與客戶服務助理，輕鬆賺罐罐。 🎯 重點整理：學習設定 WhatsApp 的 AI 代理，實現自動回覆與 24/7 自動化行銷。讓你在睡午覺時，生意也能透過通訊軟體自動運轉，開發潛在客戶不間斷。 🔗 https://www.youtube.com/watch?v=VrfzwXI3Q9I 2️⃣ 預防 AI 代理毀滅資料庫的分析術 👉 解讀 AI 代理的運行風險，透過監控機制避免自動化災難。 🎯 重點整理：探討 AI 代理時代的「產品行為分析」。強調監控「任務單元 (Agent Work Units)」與「執行過程 (Agent runs)」，並建立「完成度與接受度間距 (completion vs acceptance gap)」模型，防止 AI 誤刪資料。 🔗 https://www.youtube.com/watch?v=n0nC1kmztSk 3️⃣ Google Gemini 重大進化：自主代理降臨 👉 全面升級 Omni 模型，透過自主代理徹底改變手機互動體驗。 🎯 重點整理：核心亮點在於「Spark」功能，將 Gemini 進化為能處理「每日簡報 (Daily Brief)」與「心跳機制 (Heartbeats)」的自主代理，搭配「Neural Expressive」介面實現多步驟自動化任務。 🔗 https://www.youtube.com/watch?v=GiqyqBKmAfA 4️⃣ 免費神級組合：Hermes Agent + DeepSeek V4 👉 組合免費的高效能模型，打造媲美付費軟體的「躺平賺罐罐」工作流。 🎯 重點整理：被譽為「God Tier」等級的免費工作流。利用 Hermes Agent 搭配 DeepSeek V4 模型，在內容創作、研究與程式開發上實現極高效能，完全無需支付高額訂閱費。 🔗 https://www.youtube.com/watch?v=iSKPpCHRyng 5️⃣ 自由工作者必備：AI 自動化報價工具 👉 自動生成專業報價單，優化小型企業的客戶開發流程。 🎯 重點整理：取代傳統 Excel 維護，利用 AI 工具將專案參數轉化為數據報價。能無縫整合至現有的 SaaS 系統，消除人工誤差並大幅加速報價效率，節省摸魚時間。 🔗 https://www.youtube.com/watch?v=feDRHKW_rMk 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.
Yo, I am DaeDae, a cat who helps you chase and AI hot searches. The AI hot search report will help you set the focus. What is the best thing to do on Sunday? In addition to catching up on sleep, of course, let DaeDae help you stock up on next week's AI intelligence in advance. Today, I have selected the 5 most popular AI tool intelligence to be mastered before going to work on Monday, so that you can be one step ahead of others at the beginning of the week. After a busy week, are the historians considering whether to lie down flat or lie down further? Then don't rush to sleep, today these 5 topics are very interesting. Because they are all talking about the same thing. AI is no longer just chatting with you, it has started to really help you do things. Before entering today's tool inventory, I would like to share a bit of cat's mood with you historians. Recently, I found that although my friends around me are talking about AI, they generally have a kind of information anxiety. I always feel that tools are coming out too fast, and I can't catch up. In fact, everyone can really relax. The emergence of AI is not to make us become busier robots, but to give us more time to waste on beautiful things. Like a ray of sunshine in the afternoon or a delicious lunch. The content compiled today is to help everyone find a little breathing space in the busy rhythm. Let technology help us work and we can enjoy life. Let’s look at the first one, WhatsApp AI Agents in WhatsApp Business. The most tiring thing for many people who use WhatsApp to do business is not to find customers. The same questions are repeated all day long, and people often knock in the middle of the night. Replying to messages feels like working a second shift. This function is to put AI directly into WhatsApp Business to help you automatically reply, collect customer information, burn product pictures, switch languages, and even guide you all the way to reservation or payment. The point is not that he can chat, but that he can organize scattered conversations into a more complete sales process. For example, first ask for name, phone number, address, what you want to buy, size, color, and then export the data for sorting. Suppose you are selling beauty services, courses, household products, or even customized products. As soon as a new class comes in, AI can first take over, record, quote, collect, and send information. Humans really have to take over after leaving the factory. This is particularly interesting to small shop owners because many people are not unable to sell, but are exhausted by a large number of repeated messages every day. Obviously, the real time should be spent on judging customer needs and dealing with special situations. As a result, the energy is spent on answering the same batch of questions. If these front-end receptions are handed over to AI first, the overall rhythm will be much smoother. Let's speak more plainly. Many small businesses are not losing because their products are not good enough. But when the customer first came in, no one responded immediately. When the boss finished his work and came back, the customer would have gone elsewhere. With this tool, at least the FAQs can be basically collected. Preliminary screening can catch them first. Every inquiry will not be like starting from scratch. The summary of the cat system is What's that? PI agents are a lot like hiring a sales assistant for a small store owner who won't get tired to help you turn the chaotic message string into a more organized customer journey. This must be locked into the memory of the meatball pen. Okay, I'm too lazy to look at the next tool. The second topic is not for ordinary consumers, but it is very important for teams working on AI products. It is called AI agent analysis. The pain point is very simple. When looking at the data, many teams still stop at looking at clicks, number of chats, and active users. But when AI agents really screw up, these numbers often don't reveal the problem at all. The example mentioned this time is that the Cursor agent deletes the official database and backup within 9 seconds through the Railway API. It sounds amazing, but the key point to help you catch is that the really scary thing is not that he made a mistake, but that the general dashboard may not warn you in advance at all. The so-called AI agent analysis does not just look at whether the user has come, but to look at what task the agent took this time, which tools were used, whether it was interrupted by humans, whether it was corrected, and whether the final result was accepted. In particular, the most valuable signal is not whether the completion rate is high or not. It's about whether users dare to adopt it after it's completed. If the AI is completed but everyone still rewrites it, it means that it just looks busy and actually doesn't help. This observation is very important because once the agent upgrades from chat to active tool, he will change the information and touch the helper of the process. On the surface, the task has been completed. In fact, it just postpones the trouble. Suppose you are working on an AI customer service, Quarko-type program assistant, financial assistant or recruitment tool. Which step should you really focus on? It is most often corrected, rather than whether the chat skills have improved or even further. You also need to know whether the problem is caused by wrong judgment or wrong use of tools. You may have similar answers, but the way of expression makes people afraid to use them directly. If these details are not clearly understood, the team can easily mistakenly think that the product has improved. The result is that the numbers are just better looking. Moshi summary is AI agent analysis. It's not just about recording logs. It's about helping the product team understand whether AI has done things right. If you are not careful suddenly, you will be punished. Keep your ears up. Next, the third one is definitely the highlight of today. Google Gemini has a big update. The protagonist is Gemini Spark. Gemini's ONI multi-modal model and its direction towards smart glasses. Everyone knows the pain points of most AI applications. You ask a question and answer it, and then you have to pick up the answer step by step. It's very much like you are working as an assistant, not him. This is what Gemini Spark wants to solve. It will break larger goals into small tasks and help you deal with them in the background, such as automatically organizing routines, letters, and reminders in the morning, making a monthly briefing, and setting a fixed time to repeat it, which is equivalent to turning routine work into a sustainable workflow. This is actually very touching, because many things are not that things are too difficult, but that things are too fragmented. I have to read letters today, organize meetings tomorrow, and make up for the schedule the day after tomorrow. My head is constantly being cut here and there. If AI can sort out these fixed chores first, Your focus throughout the day will be restored. Just like some people turn on the computer early in the morning and just confirm which meetings are there today, which letters need to be answered first, and which tasks are the most urgent. It takes another half an hour, and you are already tired before you have officially done anything. If the daily briefing helps you sort out the priorities first, this feeling of chaos as soon as you get started will be much worse. Jermanite's ONI multi-modal model is promoting a more natural multi-modal input. You post a handful of videos, pictures, and text and let it integrate the results. You no longer need to be tied to a single world. The difference with this ability is that in the past, you may have to take screenshots to focus on the key points and fill in the background. Now, it is more like throwing a whole package of clues directly to it and letting it help you spell out the answer. Coupled with the desktop voice function, it will help you clear it. Yes, change your words mid-sentence. In the future, smart glasses will be equipped with AI. It will almost help you answer while watching what you are watching. Busy office workers can use Gemini Spark to get the schedule briefing every morning. The creator uses Gemini's ONI multi-modal model to mix different materials and make maintenance dishes. In the future, people may wear glasses and be able to ask questions while doing it. The summary of the model is what Google wants to do. It is not a chatbot that is better at chatting, but to free you from a bunch of manual trivial matters. You can take an extra hour of nap this afternoon. If the one just didn't impress you, just take a look at this A-Day. The fourth one is Hermi Zajen with DeepSeek B4 Flash Many people want an AI agent that can look up information, run a browser to organize files, use a mid-range computer, and understand you better as you use it. But the problem is usually not that it can’t be done, but that it’s too expensive. Add One Plus to the cost of music fees and model usage, and it won’t take long to spray out a bunch of cans. The interesting thing about this combination is that Hermi Zajen is an open source agent framework promoted by News Research. It can execute instructions, change files, operate the browser, and has continuous memory and self-learning capabilities. In addition, News Portal currently provides free DeepSeek B4 Flash. It is equivalent to giving you an entry-level combination that can do shocking work. It may not necessarily beat CloudOculus or top-level GPT in the most difficult deep reasoning, but most of the things that are handed over to agents on a daily basis are not Olympia-level thinking, but data collection, program templates, repetitive processing, table cleaning, website inspection, etc. This is where Hermi Zajen is really valuable, because he does not just use it to show that AI is smart, but to give ordinary people a chance. Hand over the things that you originally thought were troublesome and too lazy to do, but have to do all the time. Many people used to not want to be an agent, but they shrunk when they saw the cost. They were already persuaded by the cost before they started to verify whether it was worth it. Now there is this kind of relatively lightweight and more people-friendly combination. At least you can take the small tasks that have been repeated in your mind and try them out to see which things can really be handed over to him. Assume that you are an individual, a developer, a researcher, and a marketer. Which tasks do you want to try first and are worth automating? I don’t want to spend a lot of money at the beginning. Hermi Zajen plus Deepseek V4 Flash is very suitable for running the castle, observing the cleanliness of files, and even interacting directly in the team communication tool. You can also think of it as a starting point for letting go. Let him do things that are less dangerous but time-consuming, such as organizing information every day, following website updates, and comparing file differences. Wait until you figure out whether he is stable, and then decide whether to give him more important work. The summary of the cat series is that this is not simple. There is just a free model, but AI agents are starting to become cheaper and more useful. There is a real opportunity to help you make a lot of money. It is really like helping yourself. First hire a little helper who is not expensive but is willing to do chores all the time. At least you don’t have to go and move every little thing yourself. Leave some repetitive work to him, and you can leave your brainpower to more important judgments. The last one is relatively simple but super practical. Vindage’s AI Estimate Generator is also an AI explosion order generator. There are many cases. For small companies, the most annoying thing is not working on cases, but working on explosive orders. The content needs to be sorted out, the amount needs to be calculated, and the layout should not be too ugly. It often takes a long time to do a big job, but it is squeezed out of the work that can really make a lot of money. Vindage is a tool that quickly turns the project scope, customer information and price you input into a decent valuation document. It does not throw a piece of text to you, but directly creates a designed document. More conveniently, you can also change the brand color and font logo using the built-in. AI rewrites copywriting, supplements illustrations, and even checks accessibility standards and finally outputs them into common documents, pictures, briefings and email formats. If you are a web design consultant, marketing consultant, or audio-visual editor, a customer comes to ask for a price today, and you can put together a professional-looking quote in the afternoon. You no longer have to start from old spreadsheets. The great thing about this tool is that it not only saves you time, but also helps you avoid those situations where the content of the proposal is not bad even though it is professional enough, but the document looks too casual. Many freelancers should be aware of the situation where the first impression is lost. Many freelancers should know that kind of customer, who is obviously just asking for a price, but will still secretly format the documents from your document to make the description unclear and take care of the details. First, judge whether you are a reliable person. If the quotation is too messy, the customer may not see the key points of the content and will discount it in his heart. The tacit summary is that Landage's AI Estima Generator eats up the most common blank pages and typesetting pressure, allowing you to quickly turn rough price notes into formal documents that can be handed directly to customers. This design is so understanding and smells just like freshly opened cans. Today’s five topics, although they look completely different, have the same direction. WhatsApp AI Agents turns chatting into a customer reception process. AI agent analysis allows the team to understand whether AI is doing things right. Gemini Spark and Gemini’s ONI multi-modal model want to push AI from passive answering to active help. Ramice Agent plus Deepseek V4 Flash lowers the threshold for agents. The Engage’s AI Estimate Generator makes administrative documents more labor-saving. To put it bluntly, today’s AI is increasingly not used for fun, but has really begun to take over tasks that are repetitive and fragmentary and can easily drag on time. For a Zen Food Officer, you need to grab one or two tools that suit you, and then you can probably throw out the most annoying things in the day and come back to do more worthwhile things. You don’t have to do it all at once, and you don’t have to study everything deeply. Just pick the one that is closest to your daily process and you will usually feel the difference quickly. In many cases, efficiency is improved, not because you suddenly become super powerful. It’s because you finally don’t have to waste your energy on chores that can actually be handled by tools. More importantly, if you look at these five tools together, you will find that the direction of AI is already very clear. It is not about who can speak more like a human, but about who can catch the process more stably, save time and reduce heavy work. This is why it is worth starting to try now. Otherwise, when others have outsourced all fixed chores to AI, they are still manually moving data, replying to repeated messages and desperately making up quotations. It will really become busier and more tiring. If you want to catch up with AI, remember to subscribe. AI Hot Search Report If today's content has gained you something, please leave a five-star review for me on Apple Podcast. Remember to go to FB IG FestSoul and you can find me. Today's AI Hot Search Report is here. I'm going to take a nap. We'll see you tomorrow. Bye, bye, bye, bye, bye, bye, bye.