Hello, I'm the host of The Better Talk Talk. Today, we have with us Yoo-han and Seung-hoon. We will be discussing the topic of Cursor AI. Specifically, I would like to introduce the Cursor AI, which has been affectionately referred to by various names, including Father Cursor and Seung-bo. I will be leading this session with Seung-hoon.

Could you please introduce yourselves? Let's start with you, Yoo-han. 

Yes, I'm Yoo-han. I was introduced to Cursor AI by Seung-hoon, and I presented my own version with examples during the last session. It’s actually a really interesting tool that Seung-hoon introduced to me, and even though I’m not very skilled in development, I’ve been using it enthusiastically.

I have also made the payment, so I learned a lot. I will study and explore first to share good examples and cases. I am really looking forward to today's content. Thank you very much. Hello, I am Kim Seung-woo. I am part of the Jupyter operating team, and I manage a community of about 2,000 people focused on AI and non-violence. Recently, thanks to the support from the Better community, I started my YouTube channel, Seung-woo. I posted a video comfortably, believing in the power of the audience. I made a video about napkins, and I felt that the Better audience recognizes the value of this content, which allowed me to create videos more comfortably and convey my message with authenticity. Thanks to that, I...

It's been about two weeks, and currently, the number of subscribers has reached around 1,700. I believe this is possible because I focused on the essence of it. The content I've prepared this time can be seen as the second part of my series. Last time, after watching the real-time integration cases of the works of the artist and the cursor, I felt that I was at the center of a world that is changing, and it was quite astonishing. After that, I thought that I should conduct another talk, and not long after that, I prepared and executed it. It looked really impressive, and I wanted to take on a challenge that I could approach lightly, which is why I participated in this event. Now, getting to the main point, hello everyone, I am Kim Seung-ho. Yes, that's right. The hot topic is the video that has over 2 million views.

I have confirmed it, thank you. So, these days, starting from the end of August to September 2024, innovations in AI are really accelerating. Just in the two weeks since I witnessed the innovations with the ChatGPT model, the GPT-5 preview model was released. So today, I think it will be a time where you can effectively grasp the value of the Obsidian GPT-5 and the significance of these AI tools. Let's start lightly. I'm really looking forward to this. You are preparing the screen, right? Yes, that's correct. The thought process is truly remarkable. Two weeks ago, before the release of ChatGPT-5, while you're preparing the screen, I'll keep talking. Is the screen ready to share? Yes, it is. Okay, I have shared the screen. Let's take a look. Now, regarding recent updates, OpenAI has introduced the ChatGPT model.

OA released the AI 51 model. Now, its mathematical abilities are outstanding. It seems like this is an early version of AGI, similar to the existing Strawberry Latte. There’s AI that thinks before answering, but I always have some fears regarding this. Surely, in about two or three years, AI will become smarter than humans and will pose a significant threat to human jobs and roles. So, what can we prepare for before that happens? I think the enjoyment of human achievement might only last for two or three more years. If technology continues to improve and AI becomes smarter, there may be scenarios where humans can still play a role. I’m not trying to say anything definitive, but that’s a possibility, even if it might not happen.

There are situations that seem impossible, so I wondered what I could do if this is the last opportunity for achievement. If this is the last chance to challenge myself, what would be the best use of my time? What should I do to grow the most? I thought about this. 

Now, regarding the role of AI, the roles that people used to perform have changed because of computers. Simple calculations that people used to do with an abacus are no longer done by humans. Jobs and roles can shift, and I pondered what the essence of being human should focus on in this changing landscape. 

In that process, I came up with two things that would be good to prepare for. I have thought about this for a long time. The first is branding. The reason I considered this is that branding leaves a lasting impression on people's minds.

Anyway, I think that as long as devices like Neuralink are not implanted in human brains, people will consume brands that have been ingrained in them. Even if AGI comes out and technology is applied to everyone at a low cost, if a company has already secured its brand, then AGI can utilize its capabilities more effectively and apply them better. That's the feeling I get. So the first point is the brand, and the second is fandom. This is because it is fair; it is provided to everyone. Whether someone uses prompts well or not, if you ask it, it will do it for you. Even if you ask it to use prompts well, it will do it for you. So I thought about what aspects can bring about essential differentiation. I realized that perhaps an individual could be stronger than a corporation.

I thought it was possible for a company to collaborate with influencers to enhance branding value. However, if influencers can create their own products and deliver them to consumers using their influence without needing to collaborate with companies, then companies may become unnecessary. This means that the power of individuals becomes more important. So, what can we do to strengthen individual power? I thought that building a fandom could be a good approach. Now, returning to the main topic, we use Obsidian to reflect on our personal content, organize our knowledge, and outline our areas of interest. We make notes about our experiences, including failures, successes, and our unique stories. We can use those notes with AI when creating the introduction for the content we want to produce.

The introduction needs to be crafted into good content that resonates with people and opens their ears. The important point is that we are not creating an AI video script from start to finish. We should incorporate existing experiences or personal stories. Yes, a strong introduction is crucial. By focusing on this part, we can create a memorable icebreaker for the audience. When we share personal stories, it makes the narrative engaging. There are moments when we attend lectures and our ears are closed, but then something clicks, and information starts flowing in. This could be one of those moments. Once that icebreaker hits, our minds become clearer and more receptive.

When you become interested and start listening, if you can create a strong introduction, then you can develop the AI and naturally express what you want to say. The first 5 to 30 seconds of a YouTube video are crucial. In the first 5 seconds, viewers decide whether to continue watching after clicking on the thumbnail. If the first 5 seconds aren't compelling, they won't watch the next 30 seconds. However, if you can capture their attention in the first 5 seconds and then present the content you want to discuss in the following 30 seconds, viewers are likely to have the patience to watch for about 3 minutes. People generally have less patience than you might think, so creating a strong introductory segment is essential. Just looking at the title alone makes it hard to understand, so let's compare it this way. We have a new AI tool, for example, a visualization tool like Neakin AI.

I create YouTube content that reviews various topics. In this context, some YouTube channels simply showcase features like AI tools, explaining how to use them, how to log in, how to sign up, their pricing, and what functionalities they offer. Unfortunately, most YouTube channels tend to focus solely on this type of content. However, when more engaging elements are included, people tend to enjoy it more. This is because YouTube is fundamentally a platform for leisure rather than education; it's not just about following along but rather about watching casually. People watch videos while lying down before bed or during their downtime. Therefore, I believe that YouTube content can simply be entertaining, even if it is presented as a variety show.

If you just change the recognition of cognitive transformation, then the tutorial will find everything on its own and learn by itself. So, I think it would be good to have entertainment elements in the videos that facilitate recognition transformation. Looking at the two types of content, the latter contains the creator's experiences, values, and special stories, including success and failure stories. For example, I can present such cases and visualize them. For instance, I tried to explain a concept to someone by talking for two to three hours, but it didn't get through at all. However, when I used a napkin to create a diagram and talked while showing a picture, the concept was conveyed in just five minutes. This shows that rather than lengthy explanations, people...

I realized that I understand things better when I can see them. By doing it this way, it becomes a more special content, and since I am sharing my story, the audience feels a differentiation from other content. That's right; it's about individual experiences and storytelling. Many people are aware of how important storytelling is. When you search for storytelling on Google, you'll find many books on the topic. This shows that the technique of storytelling resonates with people. So, what are we going to do? We will take a look at three tools today: OpenAI's model, Cursor, and Obsidian. By using these tools, we can receive AI assistance in building fandom, which could be really important in the age of AGI and for creating content that fosters fandom.

Let's talk about strategy. We'll start now. First, we will use three tools. The first is that we will create a file structure in the Obsidian repository to store individual experiences. I'll show you how it looks. This way, we can document career experiences, what kind of job one currently has, what skills they possess, what communities they are involved in, who they meet here, what activities they have participated in, and what their hobbies are. For example, one might have a hobby of playing the piano. Additionally, we can include what types of writings they enjoy. This information can be generated using the first tool. However, today, we will not only create the structure but also fill in the content. We will imagine a hypothetical mission and start from there.

Imagining that I am not yet prepared for a personalized experience, I would like to show this. Then, using Cursor AI, I will play the role of a connector. I will create a video explaining the napkin AI tool, referencing my own experience. In the Ohan repository, you will understand this image a bit better. My values and experiences are stored in documents within the Obsidian repository. It's not just one document; it's well-organized and structured. 

Now, by utilizing AI 2 and Cursor AI with Obsidian, I can create a foundational framework for content that reflects my own experiences. This is necessary because, when we shoot videos, if we do it somewhat spontaneously, there are often parts we miss during editing. In fact, I wish I had included certain parts after the fact.

In the case of the door, I had some regrets. People often ask questions like, I created a visualization with this napkin, but is there a way to export it as a PPT or PNG file? I noticed that this is a common inquiry. I thought it would have been good to include this content, so I considered how to ensure that we don't miss such points. AI can help us address these gaps. 

It can elevate our understanding and help us keep track of what we might overlook. Reflecting on our unique experiences, we can't always instantly recall everything from our brains. While some people have that ability, it’s not common. 

Every time I go to Obsidian, I've been waiting for this day. It’s wonderful. When Claude was released, I thought it might take a while before we reached this level of capability, but here we are now.

I think it's become possible. Thank you. So, we're going to create this step by step in order. Let's take a look. So, starting with this output, when we make a YouTube video, what can we gain? It creates content that can boost the algorithm and build a fandom. Let's go ahead and start this. If anyone is able to use Obsidian, Cursor AI, and ChatGPT-5, feel free to follow along. I will share the prompts I use so you can follow along easily.

First of all, those of you who do not have an Obsidian repository don't need to worry, because Uwon will create everything for you. I will use a pre-prepared prompt to try out the Uwon preview model. From now on, up until now, has been the introduction.

This is a practice session, so I would appreciate it if you could focus. Now, I will try to incorporate this prompt, which is important. You can change this part. Imagine a fictional person who has put a lot of thought into visualization, and create a structure based on their personal experiences, including successes, failures, and interesting experiences. Fill in the content as well. The experience part is crucial. This person has had experiences related to visualization in the Better Growth Community, enjoys playing the piano as a hobby, and is a 30-something office worker. By the way, each experience needs to be divided into separate files. Let's start with this. After going through this thought process, we have generated this file. The remaining content is being created like this. Now, I will transfer this to Obsidian. However, it's a bit...

It's not very specific. It's about 20 seconds fast, and I'm around 19 seconds. However, I'm not satisfied with the generated result, so I'll try generating it again. While generating, what should I use as a standard for what I like and dislike? If I show you the previous conversation, it would be helpful. Let's first create the file structure. I have something I want to show you. I think it would be good to try the prompt again with this. Watching the generation process is also quite interesting. This is how it turns out; it's what I wanted. If I ask to think of the file structure first, it comes out like this. Change the name to Korean and fill in the content. I'm continuously reasoning through various steps to reach the goal. I'm thinking about how to fill the content.

We create a structure like this, and for each content, we develop the main text. Then, a virtual experience for this person is created. How would this person explain the AI? Let's take a look at that. However, there is this file structure. For example, I will delete this. Now, if I create a folder called Experience and then create another folder called Success, it would take too long. We don't have to do this manually. First, I will copy this and move it to the GPT-4 model. I will paste it here to generate a naming convention for creating this file structure. This will create the command. I will copy it and go to the terminal below here. I will paste it. Please pay attention to the left screen. When we run it, it appears like this. Yes, that's right.

All the files have come in, and now we can copy them one by one. This can be done in bulk. If we utilize AI well, it should be fine. Now, let's move on and try copying one by one. I wanted something lengthy, but for now, this is not something AI can create, so I’m showing it quickly as a virtual creation. You should write it down by hand yourself; that would be good. Now, I will paste one by one for a successful visualization project. This needs to be done manually. I successfully performed a piano piece, and it would be good if you remember the persona part: a man in his 30s, working in a company, participating in a better growth community, has experience related to visualization, enjoys playing the piano, and works at a company. Also, while playing the piano and looking at the sheet music, he has related experiences in visual expression.

You can think of it as having this domain. There are experiences of success, and now we have a tool for visualizing failure. Now, you will fill it with your content. There were frustrations in practicing the piano. I couldn't secure my skills during the practice, and my progress stagnated. I kept making mistakes in the pieces I was playing, and I felt a sense of regression. I decided to reorganize my time management and commit to practicing diligently again. That's right. We often organize the knowledge we've gained about such learning experiences. By utilizing that, you can start creating persuasive YouTube content that helps in building a fandom. Participating in community workshops can be a fun experience. I found interesting elements in the sheet music. The visualization of music and the ability to sing by sight—does the structure differ a bit? One by one, I lost my focus. You shouldn't feel disheartened yet; we haven't even started. I'm just struggling to find this file.

I will try to explain it in my own way. Learning new technologies from colleagues is important in our daily lives. Here, we can see the significance of visual perception. In the report on music and visual recognition, we have transformed traditional text-based reports into visual reports using graphs and infographics. This has greatly enhanced the effectiveness of information delivery. It's a clear proposition, indeed. I have saved everything, and now, based on this experience, let's explore how we can create an introduction section for YouTube using AI. First, we need to go to the Kin AI site. I will provide the address here. On this main page, all the content is available. Let's copy everything from here. Now, I will create a folder here, which I will name the output folder. This is because, if you look here, it is currently the storage location.

I'm creating content that reflects my own experienced experiences using AI 2. So, I'll be making an introduction here. I will label it as Rapteen AI Project MD. When I click the command here, the chat window appears on the right. Let's give it a try. I'll do this separately as well. I can attach the Rapin AI description MD file, so I'll paste it here. 

Now, let's create an introduction for YouTube. Before that, let's take a look at what happens when we use the existing ChatGPT. I think we need a comparison. I'll copy this and ask it to write a YouTube script. Just by giving a brief explanation, it can generate something. However, the point is that anyone can do this; it should not be a differentiating factor. You all know how important it is to visually express your ideas, right? It's amazing.

도구에게 지시를 필요 없이 텍스트만 붙여 놓으면 된다. 이런 맥락은 다 되고요. 이제 여기에 나의 경험을 넣어 보겠습니다. 제가 준비한 프롬프터는 이거고, 커서에서 유튜브 서론 창을 열어 놓고, 여기서 골뱅이를 해서 잠시만요, 폴더를 넣어 줄게요. 이렇게 하나의 폴더로 묶었고, 여기서 볼트라는 폴더를 멘션하고요. 어떤 툴인지 이제는 여기서 레킨 파일스, 냅킨 AI 설명. 이렇게 드래그해서 넣을 수 있네요. 이렇게 해서 사용한 기존 운선 제목을 괄호로 언급해서 연결해 주라고 했어요. 이렇게 해서 만 어떻게 만들 수 있는지 한번 보도록 하겠습니다. 이렇게 모든 폴더를 전송하고, 자, 이런 식으로 만들어졌는데, 맥락 파악을 잘 못한 것 같아서, 더 보고 차례를 볼게요.

I will try to change the model a bit. It seems to be somewhat related to Seo-won. I’m currently reading a specific tool introduction video. The text file on the Dekin AI homepage is a bit long, so let’s take a look at it. It seems like the curse of demonstration has occurred. Let’s check it out. I will reset it now. Yes, that’s right. It comes out like this, but I will request the existing tools based on the contents of the folder and generate it like this. Since it’s continuously applied, I will approach it in a different way. I want to shoot YouTube content about Napkin AI, so please find something from my experience that could help with ice-breaking. It came out like this, and it found related content. Now, let’s start with this. I guess I need to approach it this way. The copying didn’t go through properly.

I won't do it manually, so please input it for me. Now, um, just a moment. Ah, yes, yes, yes. Here, now, I need to mix the content above and the file content to create an introduction for YouTube. I'm looking into whether I should do this through various conversations. I want it to be made in this way. Since the context of the conversation is above, I should incorporate my experiences. So, I'll request a modification like this. Here in the latter part, personal experiences came up, and I've found this part to add. The green part is the newly added section. Now that this part has been added, I'll keep only the 30-second segment from this content for the YouTube script, focusing on experiences and connecting existing experiences with the strengths of AI. So, I've written it like that. I will press the fly button and set it up. Ah, yes, looking at this, I've talked about my experiences. This part, um, let's see, my passion has grown, and is it visually changing? It's complex.

It allows us to convey complex ideas easily. It generates appropriate visual materials. However, since the cursor is specialized in coding, there seem to be some random issues where it works sometimes and doesn't work at other times. Additionally, there may be problems with file names being in Korean. It would be good to identify these issues. 

If we use our tool to categorize materials, for instance, by storing my success stories, failure experiences, and unique narratives, we can not only focus on creating YouTube content but also plan lectures. For example, if I conduct a lecture offline, I can connect the lecture content with my experiences to break the ice with the audience and warm their hearts.

I think it's possible to create a natural connection and transition into the main lecture content. So today, I demonstrated this approach. However, I found that Obsidian has some shortcomings in the AI aspect, which I believe requires further research. If the functionality of the cursor improves, or if we can identify issues with prompts, that would be beneficial. 

To summarize, we discussed strategies for building fandom content with the help of AI. First, I think it would be good to use the 5W model to explore some questions. Then, regarding the structure of Obsidian, let's take a look at generating its structure using the one-preview model. I will write this out. 

Now, using RAG, I will plan YouTube content that incorporates my personal experiences effectively.

I want to organize my experiences, my know-how, my success stories, my failures, special stories, and interesting anecdotes. From a content planning perspective, I think about what topics I should note down. I want to write this and use AI to automatically connect my intended topics with my experiences. Ultimately, I would like to consider what the structure of an Obsidian vault file should look like. 

This is something I can change, and now I can start gathering data on what I want. Designing this framework can be quite unfamiliar and difficult at first. After going through the thought process, I can figure out what content would be beneficial to include, such as the reasons and situations behind my failures, the lessons learned, and the methods for improvement. 

I would like to create an Obsidian file structure and file format that can encompass all these topics.

Let's create a format where we can answer questions instead of directly creating content ourselves. We will set it up so that we only need to respond to the questions that AI generates. Would it be better to use the Wooon model for this? If so, I will give that a try. It seems like the context size might be an issue; the model's performance fluctuates depending on its condition. I remember it worked well when I was speaking. 

In this project, you will write about your experiences regarding success, but I will generate all the questions for you. The questions will include when, where, and under what circumstances you had these experiences, what emotions you felt, and what you learned. This way, you will receive direct answers to these questions. You can then duplicate this format to fill in the document. This is how we can create such prompts.

If you want to plan a business, for example, and change your business model, you don't necessarily have to change everything. If you rephrase it to How to effectively plan a business model, it gathers everything you need. When you want to explore something in depth, referring to the 5W model seems beneficial. It checks various possibilities according to the goals you want to achieve. It’s good to write down insights, ideas, and insights about industries or markets. It helps create structures as well. I believe the 5W model is especially useful when you want to solve complex problems. This approach can also help you figure out how to structure a lecture. Let's take a look at the 5W model.

I’m not sure if I can use the 5W model. I’ve made this first, and I learned about context text summarization. Thank you. I will try again. I think I can easily express complex ideas using this tool, and I just need to write the prompts well in this part. Why is this section empty? I only made corrections in this part, but I will try changing the model again. The 5W preview model is not usable, and the 5W mini model takes a long time. It seems to be doing some inference. The results are definitely coming out more concretely. Thank you. As you mentioned, I will try the prompts again. Please mention the content specifically, like this. The pro model is 5W mini, and it's free. Thank you. I can use it ten times a day.

It's free to use ten times a month, which is nice, but I feel like it's a bit sufficient. If this works well, we can move on to the next topic. I hope we can stay motivated. I think there might be an issue with the prompt I wrote. Now, this has been saved, and if I open Obsidian, let's take a look. I will click to open it. In the graph view, this part connects to the YouTube introduction, but it seems that the connection is incorrect. If this part had been connected correctly, it would have linked smoothly. I believe there is potential to use AI explanations in this way. Let's try some different functions and see how it goes. I'll try a pose; it seems a bit lower when composing.

It seems that there are specific examples being discussed. For instance, important data analyses conducted by the company have emerged. Let's apply the importance of visualization in everyday life. Finally, we have drafted this document. Let's see what content is included. According to the napkin AI explanatory document, it generates the most suitable visual materials based on text. By using napkin AI, complex data from important data analysis projects conducted by the company was expressed in an easily understandable manner. Thanks to the visualized materials in this document, decision-making during management presentations was expedited, and communication became quicker. Through this experience, I once again realized the powerful impact of visualization. Now, regarding the team project, there is a discussion on the correlation between music and visual recognition. This part is somewhat...

Because of that, I chose to ask for more specific examples and metaphors. If you press Command L, you can mention it. This part needed a more detailed explanation. The relationship between music and visual perception was somewhat weak in the document, and since it wasn't specific enough, I think this is why it was approached this way. When I asked for more specificity, it wasn't about speaking in abstract terms but rather about providing concrete instances, such as experiencing certain colors like blue or red when hearing specific sounds, or recalling natural landscapes or abstract images when listening to melodies. This is how the text gets revised to be more specific. Now, if you think of this mentioned data as your own data, it connects in this way.

Now, with the advancement of AI, there is potential for music and visual recognition to come together in everyday life, allowing these experiences to automatically gather around products. This possibility is currently opening up, and I think we can view it as a significant opportunity. I believe that structuring and storing these personal experiences and connections will become a valuable asset in the future. This is because when we think about what experiences to share with AI, it’s not something we can easily recall from our minds. Therefore, I think this technology could help alleviate that challenge. I demonstrated this in real-time, and I must say that the demonstration was quite risky. Despite some variables, it still presents good potential.

I hope you have discovered it. So now we have The Better Talk Talk. Then, why is the session called Talk Talk? The reason is that The Better gathers truly wonderful audiences. I thought it would be a place where instead of one-sided knowledge delivery, we could have conversations like, Oh, I want to do this, and I want to do that. So I think it would be great if you all could share your thoughts as well. For example, when planning a lecture, how can I utilize the experience and data I have? AI can automatically find ways to utilize it, so it would be good to discuss this. That's all I have prepared. Thank you. Yes, thank you. Thank you, Suhun. I believe this is my first time doing a live session like this, except for when you gave lectures in the past. It feels like it's been a long time since then.

Is that right? Ah, yes, that's correct. Usually, the risks are high, so we don't often do demonstrations. That's right. Yes, during fan zone sessions, when we do these live sessions, it always seems to go strangely well. Even though we're using the same prompt and the same examples, sometimes the results come out unexpectedly. Yes, that's right. It's the magic of live events. Nevertheless, those three thoughts seem to create a loop of ideas, and the process of building those thought loops is really interesting. I think it's great. So, I realized that utilizing AI is somewhat like this, and I felt that a lot. Also, after trying the composer function at the end, it seems that if we utilize the composer function well, it could work effectively. And if we enrich the data further, it will definitely improve.

It seems that, in fact, for beginners like me, the way to utilize AI is simply to ask questions that I am curious about. When I receive answers, I think, Oh, that's interesting, or That's quite smart, and I often just use that response. However, I realized that by engaging in deeper thinking together, I can enrich and refine my strategies and the content I create. Thank you. 

It seems this is developing. So, I am currently trying to create a total of 12 examples of how AI has been used for business and industrial innovation. I have discovered about two so far. The first one is creating slides, and the second is editing documents with a cursor. Once I complete this, it will be the third example. I will make sure to finish it well.

By that time, I thought we might need to invite another Tober Talk again. If you have any questions, feel free to ask in the chat or turn on your microphone. Do you have any questions? I'll count to five: 5, 4, 3... It seems there are no questions. The lecture was excellent, so it looks like there are no inquiries. Thank you so much for staying with us until this late hour. That's right; you really need to try it yourself to understand. If you don't try, you won't really know, and if you don't give it a shot, it won't stick in your mind. Later, you might think, What was it again? and when you search for it on YouTube, there won't be anyone to ask. That's right. If you come to the Tober room, Seung-hoon is always there, so it would be good for you all to ask questions as you progress later on.

Yes, Mr. Sim asked this question about how to handle a situation where there are too many reference materials. How should I respond? The concept of a cursor was originally created to manage programming code, so it selectively uses files. First, it searches for file names and then selectively retrieves the appropriate files, which means it doesn't require all files to be included. Therefore, as long as the length within a file isn't excessively long, having a lot of materials shouldn't be a problem. It seems that processing is possible. Regarding token limits, it should be manageable within those constraints. Yes, the limitations on tokens can sometimes be frustrating, but they can also be beneficial, and there shouldn't be significant issues as improvements are likely to come. Indeed, the future seems promising.

I find myself waiting and becoming curious again. I am grateful for those who have spent time with me. It seems that Lee Sung is asking a question, and you are providing an answer. If there are no additional questions, we were planning to conclude here. However, Young-in has submitted another question about the structure of the YouTube algorithm. I will answer that for you. 

As for linking documents in Obsidian, I understand that it is done with curly braces. That is correct. One thing I would like to mention is that when using the cursor in Obsidian, mentions can only link to files and folders. The important thing is that for files, the title must start in English for the search function to work. So, there are some issues with that. Regarding folders, should they also be in English? My note titles are in English.

It reads well; otherwise, I have to write it in manually. So, automatic completion doesn't work very well. When I input similar files in bulk, I start the file names with an emoji, which I showed you back then. If my format starts with an emoji, it tends to work quite well. However, it seems that Korean text is not read very well. So, I have been starting all the long and short texts I've written with the same emoticon. I have the ability to do that, and I thought I should do more work on gathering similar texts into folders to manage them better for bulk mentions. Recently, I had a meeting with the Seoul National University startup team about creating a note, and I was discussing it with a team from Release AI.

I was thinking about whether to use the Obsidian Ball plugin and asking for opinions on it. I shared a few experiences with using Cursor AI, and it seems like they might create something interesting, so it might be worth waiting for that. Oh, that's great! So, it's starting to come out now. Yes, it seems like there are possibilities with Cursor, especially since it's good for coding. I think we can also expect it to be useful for document writing, as you mentioned. Thank you! 

By the way, when you create a folder, you have to move things back later, right? That can be a bit annoying. What I do is set the shortcut for Rein Finder in Obsidian. So when I press the shortcut, the corresponding file pops up in Finder. Then, I can just drag the items from there and move them to Cursor.

Everything opens up, and by the way, you can also include files from other vaults. This isn't just my vault; it's possible to include files from different folders, meaning you can reference sources from other vaults as well. I think it wouldn't be a bad idea to manage these reference sources in a separate vault. So, aside from my knowledge management vault, I can keep templates or other reference materials separately. This is how my folder and vault are structured. If I have a bit of time, it would be good to look at this together. Recently, I briefly conducted a lecture on the concept of Uwon. It was really short, and I was wondering if we could take a look at it. How does that sound? I would really welcome that. Great, thank you! First of all, I believe it's important for everyone to have a good first impression of the Uwon preview. So, the meaning of the Uwon model is...

What it is, is that there has been this huge, mysterious model from OpenAI since about a year ago. Because of that model, Sam Altman and the other directors were somehow ousted from OpenAI after some disputes. But then they came back. The name of that model can be referred to as Strawberry Latte, which can perform reasoning and various tasks. If you look at it, it seems like the preview would be better, but in terms of mathematical performance, Uwon Mini can be used 50 times a day, while this one can be used 50 times a week, increasing to 70 times. Uwon Mini is also quite good, but it can also perform many tasks. The downside is that it has a bit of a weakness in broad legal knowledge. However, the preview has strong broad knowledge. Uwon Mini excels mainly in development and mathematics, so when it comes to coding or similar tasks, using Uwon Mini is better than the preview.

Alright, so 5W1H has been trained based on COT data. What this means is that it has learned the step-by-step reasoning methods of humans. For example, it understands the logical thinking processes that people follow, such as steps 1, 2, 3, 4, and 5. It has learned the entire text surrounding these processes. Therefore, when a question is asked, it doesn't just provide an immediate answer. Instead, it first reviews the data regarding the procedures humans typically go through to answer that question, retrieves the reasoning process, and then generates a response based on that reasoning. So, you can think of it in that way. Personally, I see this model as an agent. A novel came to mind, although I can't recall its title. The story involves a small, atomic-sized material that introduces an alien substance into the world.

I was looking for the name of a novel in which Earth is threatened, and I had a clue that it was a science fiction novel. I asked the 5-watt mini model about it, and it took 4 seconds to think. Then, it began to execute a step-by-step reasoning process. I was trying to find the name of a science fiction novel where Earth is threatened by tiny, atomic-sized substances. I thought it would be interesting to deduce the book title from there. Then, I delved deeper into what could be the minuscule substances threatening Earth in science fiction. Some works came to mind, but it turned out that The Play was not the right one. So, I said no. However, I remembered a clue; the book title might have been Astroneer. But when I searched for Astroneer, it turned out to be a game, not the novel I was looking for.

When you give a clue like this, I thought about the 158th free, 158th for a long time. If you look at this alone, you might miss this part if you don't see it together. What I'm showing you now is based on the clues from Astroneer. I'm thinking of words that start with astro, like starmaker or astros. I think a lot about this. Sometimes, I might mishear astro, so I'm constantly trying various approaches to see if I'm right or not. If I'm correct, I respond; if not, I think of something else. In this way, the need to ask questions actually diminishes. People would usually say, Think about this perspective and respond, but this process happens automatically. It's a fun aspect, like solving a puzzle in memory.

I am connecting various words while doing this. You know, a real person can't do this like an astrobiologist. I am exploring various aspects, so there is something different about the phenomenon of extraterrestrial substances flowing to Earth. The six letters refer to the name of that extraterrestrial substance. If you scroll down, you can see that I can answer this question in a lengthy manner. However, I couldn't find the answer here. It might be that I have the book title wrong, or the name of the substance might not be the title of a book. As you can see, there are various terms like Astro-X, Astro-Man, Astro-M, and Astro-Gel, as well as Astro-Gini and Astro-Bits. So I provided a prompt. Astro seems to be correct, but I can't remember what comes after that. Astro? Question mark, question mark. In this part, something decisive comes out. What I mean is that I am brainstorming various names related to Astro, like Astro Worm.

There are names like Astro Walker, and when you see this, various names come to mind. Astro, nano, nanotech, Astro Nest, Net, Natori, Nature, and Neve. It seems like everything related is being mentioned. Reflecting on this, I realized that when we want to discover how to create a new drug, we think about finding a method for treatment. We keep trying various approaches until we succeed. It seems there is a certain perseverance in this. I felt that possibility, and so here, I also provided additional context. In the story, I recalled that there is content about using alien materials as fuel for spaceships. So, I thought about it for 42 seconds.

So, this is how it turned out. After going through various books and thought processes, this part is indeed the answer. Among the various books, Project Hail Mary seems the most interesting. However, Astrobiology exceeds the six-letter limit. Now, I was asking about Astrobiology in Korean, but this friend thinks it has gone beyond six letters. Here, it mentions that they are considering the combination of Astro and Biology. Thinking about it, Astrobiology could also be formed, but Project Hail Mary seems to be the most suitable. Astrobiology deals with extraterrestrial substances used as fuel and scenarios that threaten Earth. This is how it comes out. So, after completing the thought process, I concluded that what you were looking for is likely Project Hail Mary.

When you need to find a single solution based on specific clues, I think the model can lead to long contemplation. In this context, if I get caught up in this, it can feel like a burden. Therefore, it might be beneficial to consider when to use the model and when not to. Additionally, we should think about what kind of questions can stimulate the model. The part about the model being strong in certain aspects came to mind. Currently, the role of humans is to utilize intuition to define problems. Once the problem is defined, we can ask the model for solutions, explore those solutions, and implement them. I believe this is the most effective approach. This is something we can think about further. So, there was this discussion from a developer at Google DeepMind.

However, if COT is used correctly, like the 5-W model, it is mathematically proven that we can solve any problem in this world. So, with the necessary improvements in hardware, such as the current transformer models like ChatGPT, and with better power efficiency and more graphics cards, we will be able to process more context. If we can generate as much as we need, we can solve any problem in this world. The performance of LM inference is singular; it means there is no singularity. I hope this gives you some reference on what meaning the circle might have and how we can use it. With that, I will conclude. This was a brief sharing of my thoughts on the 5-W model. Thank you. As you mentioned, it provides good insights.

Thank you very much. I really appreciate it. I feel like this presentation, in terms of content, might actually be less than what we've discussed so far. Yes, that's right. I felt a bit disappointed, so I wanted to add a few more comments. I enjoyed both parts, and I feel lucky to have experienced this. To truly grasp the value of this, we need to watch it together in real-time. Thank you so much for your precious time. I hope there are even more takeaways from this session. I am confident that there will be many valuable insights. Thank you again for preparing this The Better Talk and for generously sharing your knowledge. I really appreciate it, and thank you to our wonderful speaker, Seung-hoon. Thank you, everyone.

The YouTube class channel is already loved by many, but we ask for even more love. Yes, this has been The Baro Talk Talk. Thank you, thank you!

클승우님의 나만의 경험을 반영한 콘텐츠 만드는 방법(feat. 옵시디언, Cursor AI)

"어제보다 더 나은 우리"
성장에 진심인 사람들만이 모여 다 함께 성장할 수 있는 세상을 꿈꿉니다.

커뮤니티 및 사이트: thebetter.today
뉴스레터: thebetter.stibee.com
--
#성장, #지식관리, #커뮤니티, #자기계발, #생산성, #자기관리 



클승우님의 나만의 경험을 반영한 콘텐츠 만드는 방법(feat. 옵시디언, Cursor AI)

스크립트