00:01-00:21 - Researcher: Hey there! I'm Trooly, and I'm really excited to chat with you today. This interview will take about 30 minutes, and I'm here to learn from your experiences. Before we get started, could you please give me a brief self-introduction? I'd love to get to know you a bit better. For example, you could tell me your age, occupation, and hobbies — or anything else you think would help me understand you better. - 嘿，你好！我是 Trooly，非常高兴今天能和你聊天。这次访谈大约需要 30 分钟，我是来向你的经验学习的。 在我们开始之前，你能简单做个自我介绍吗？我想多了解你一点。比如，你可以告诉我你的年龄、职业和爱好——或者任何你认为能帮我更好了解你的信息。

00:29-01:11 - Marcus: I My name is Marcus. I'm twenty years old. I'm a I'm a current student at Apex. I'm pursuing a degree in computer science, hopefully. I'm graduating this year. Really hoping I am. You know, I'm really into coding. I've been doing this since I was about around ten years old. My dad kinda introduced me to this. You know, he's technologically versed. First thing he taught me how to use was Python. You know, I I learned how to create like, simple, like, drawing applications and, like, how to make shapes, like, animate them and make them move. And I went from doing that you know, developing, like, you know, like, tablets, you know, websites, softwares, you know, it's just It's it's really fun. I like doing it. - 我，我叫 Marcus。我今年二十岁。我现在是 Apex 的学生。我正在攻读计算机科学学位，希望如此。我今年毕业。真的希望我能毕业。你知道的，我非常喜欢编程。我从十岁左右就开始做这个了。我爸爸算是我在这方面的启蒙。你知道，他精通技术。他教我使用的第一个东西是 Python。你知道，我学会了如何创建简单的绘图应用程序，比如，如何制作形状，让它们动起来，让它们移动。然后我从做那个发展到，你知道，开发类似，平板电脑应用、网站、软件之类的，你知道，这真的很有趣。我喜欢做这个。

01:13-01:28 - Researcher: That's amazing, Marcus! Starting at ten years old and now graduating with a CS degree — that's quite a journey. I love that your dad got you hooked with Python animations. So you've been coding for a decade now. What kind of projects are you working on these days, either for school or just for fun? - 太棒了，Marcus！从十岁开始，现在就要拿到计算机科学学位毕业了——这是一段了不起的历程。我喜欢你爸爸用 Python 动画让你着迷的故事。 所以你已经编程十年了。最近你在做些什么项目，无论是学校的还是纯属好玩的？

01:32-02:36 - Marcus: Well, I would one of the best ways that I I like to learn I like teaching myself how to make certain things, you know, like, for games using languages that I know. So, like, right now, you know, I do Minecraft plugin development so I make websites for all of the servers, you know, full stack websites for it. Like, it'll integrate a website, which will talk to an API, which will talk to a Discord bot, which will also talk to stuff in game. You know, they all communicate with each other through that API. I developed plugins for Minecraft, so, you know, like, things that add, like, special commands for, like, moderation or any of the ease of use access that they might need. I also develop mods. I design and develop mods, you know, like, I I can make custom weapons, you know, models, all of that stuff. I've also been trying start my own kinda, like, contracting company in a way. Like, I've been trying to build my own brand, you know, got my own website, Marcus Tech dot com, you know, really nice So, you know, got my own website going, you know, trying to showcase my skills and also, like, increase my reach and also increase some other clients I have to, you know, make more money, get my name out there. - 嗯，我喜欢的一种学习方式是教自己如何制作某些东西，比如，用我掌握的语言制作游戏插件。所以，现在，你知道，我做 Minecraft（我的世界）插件开发，我为所有的服务器制作网站，你知道，是全栈网站。比如，它会集成一个网站，与 API 通信，API 再与 Discord 机器人通信，机器人还会与游戏里的内容通信。你知道，它们都通过那个 API 互相交流。我为 Minecraft 开发插件，所以，你知道，比如，添加特殊指令的东西，用于管理或他们可能需要的任何易用性访问功能。我也开发模组（mods）。我设计并开发模组，你知道，比如，我可以制作自定义武器、模型，所有那些东西。我也一直在尝试创办自己的那种……算是一家外包公司吧。比如，我一直在尝试建立自己的品牌，你知道，有了我自己的网站，Marcus Tech dot com，你知道，很不错。所以，你知道，我有了自己的网站，试图展示我的技能，同时也想扩大我的影响力，增加一些其他客户，你知道，多赚点钱，让我的名字响亮起来。

02:38-02:55 - Researcher: Wow, that's a serious tech stack you're managing — full integration across web, Discord, and in-game systems. And launching your own contracting brand while still in school? That's ambitious. You mentioned trying to get more clients and make more money. How's that going so far? Like, have you landed paying clients yet, or is it still in the early stages? - 哇，你管理的这套技术栈相当厉害——涵盖了网页、Discord 和游戏系统的全面集成。而且还在上学就推出了自己的外包品牌？很有抱负。 你提到过尝试获取更多客户多赚钱。目前进展如何？比如，你已经接到付费客户了吗，还是仍处于早期阶段？

03:59-04:07 - Marcus: I will admit it's kinda hard. That's why I'm also glad I get like, a video game development. There's a server, uh, called Dev Connect. I use that frequently. And when I go in there, you know, I'm able to, like, stake out clients. People will post out that they wanna develop it to do x y z. And I'm like, hey. I can do it. Just let me know. And then, boom, they hit me up. I complete the work. I get paid for it. I'm also on this platform called Upwork, as you know, is for freelancers. Uh, that's how I'm primarily doing my freelancing now. So, um, when I find, um, anyone who wants to hire me outside of, you know, Upwork, I talk to them, then I do direct them to Upwork. You know? Our conversations on Upwork. That way, I can get paid through Upwork safely. That way, you know, if they pay me through something like PayPal, I won't just immediately, like, have to lose it and, like, I end up losing a hundred dollars. You know? I I don't wanna experience that. But you know, I I feel like it's starting off pretty good. But my I've only made I mean, like, I just started this this week, but I've only made like thirty dollars so far. But I have a hundred and fifty dollars, uh, in the works in progress for a contract, and I have another contract pay me seventy-five dollars an hour. So, you know, it's it's starting to pick up. - 我承认这挺难的。这也是为什么我也很高兴能参与类似……视频游戏开发的工作。有一个服务器叫 Dev Connect。我经常用它。当我进去时，你知道，我就能守候客户。人们会发帖说他们想开发实现某某功能的项目。我就说，嘿，我能做。跟我联系就行。然后，砰，他们就找上我了。我完成工作，拿到报酬。我也在 Upwork 这个平台上，你知道，那是给自由职业者用的。呃，那是目前我从事自由职业的主要方式。所以，嗯，当我发现任何想在 Upwork 之外雇佣我的人，我会先跟他们谈，然后把他们引导到 Upwork 上。你知道吗？我们在 Upwork 上沟通。那样的话，我可以安全地通过 Upwork 拿到钱。那样，你知道，如果他们通过像 PayPal 之类的东西付钱给我，我就不会因为某些原因立刻损失掉，比如最后丢了一百美元。你知道吗？我不想经历那种事。但你知道，我觉得起步还算不错。但我目前只赚了……我是说，我这周才刚开始，但到目前为止只赚了大约三十美元。但我有一个正在进行中的合同，有一百五十美元，还有另一个合同每小时付我七十五美元。所以，你知道，情况开始好转了。

04:09-04:28 - Researcher: That's a solid start for just one week in! The seventy-five dollars an hour rate is impressive. Sounds like you're being smart about payment protection too. So between school, freelancing, and your own projects, you're juggling a lot. Do you have any other income right now, or is it mainly the freelance work? I'm curious about your overall financial situation as a student. - 对于刚做一周的人来说，这是一个扎实的开始！每小时 75 美元的费率令人印象深刻。听起来你在支付保护方面也很聪明。 所以在学校、自由职业和你自己的项目之间，你忙得不可开交。你现在还有其他收入吗，还是主要靠自由职业？我很好奇你作为学生的整体财务状况。

04:36-05:28 - Marcus: Honestly, I will admit, like, it it's it's a struggle. I am I am trying to make three grand because that's just rough around how much we have to pay for school next semester after, like, all the scholarships and all that because I got all the stuff from FAFSA or whatever. But, you know, you know, it's still quite a bit of money that I still have to pay. It is currently or as of now, it's my current income only, but when I say I do freelancing, it's not just like me, like, going on a contract to work. I also do, like, AI annotation, which, you know, counts as freelancing, so that that's cool. And it gives me a pretty good hourly rate for working with them. And then, um, I'm also, um, I'm also I used to do Uber Eats, like, as a little, like, food delivery person on my bike, but I got in an accident on it. So, um, I haven't been able to do that recently, which is kinda why I'm working so I can you know, I can build me a nice powerful bike. And once I get that powerful bike, I'm going right back to doing Uber Eats too for good money. - 老实说，我承认，这很不容易。我正在努力赚三千美元，因为那是我们在除去所有奖学金之类后的下学期学费大概额度，因为我拿到了 FAFSA（联邦学生资助）之类的东西。但，你知道，那仍是一笔我不小的开销。目前或者说现在，这（自由职业）是我唯一的收入来源，但当我说做自由职业时，不只是我去接合同干活。我也做，比如，AI 标注，你知道，那也算作自由职业，所以那挺酷的。而且和他们合作给我提供了相当不错的时薪。然后，嗯，我也……我以前做过 Uber Eats（优食），就在我的自行车上当个送餐员，但我出了一次事故。所以，嗯，我最近没法做那个了，这也是为什么我拼命工作，这样我就能，你知道，给自己组装一辆性能强劲的自行车。一旦我有了那辆强劲的自行车，我会马上重新回去做 Uber Eats，赚一大笔钱。

05:31-05:48 - Researcher: So you're hustling across multiple streams — freelancing, AI annotation, and planning to get back to Uber Eats once you rebuild. That three grand for next semester is a real target you're working toward. When you mention AI annotation, what exactly are you doing there? And what's the hourly rate like compared to your coding freelance work? - 所以你在多条战线上奔波——自由职业、AI 标注，还计划在修好车后回归 Uber Eats。下学期的那三千美元是你努力奋斗的真实目标。 当你提到 AI 标注时，你具体在那儿做什么？时薪和你的编程自由职业相比怎么样？

06:02-06:32 - Marcus: Honestly, the the the pay is nowhere near. I mean, like, I used to do it for TaskFlow for where I made, like, sixteen. And then after that, I I work for CampusJobs now, and I make eighteen. So it was a two dollar pay increase. But, you know, it could be worse versus, like, freelancing. It's like, depends, like, me and, like, how I find my clients. Like, usually, when I find a client, you know, the the money that I make doing the work that they want, like, doesn't really take long. So, you know, it's short. I mean, like, someone literally paid me money just, like, creating an account on their Discord bot or game. You know, I I got paid for that, and then, like, someone just paid me to, like, not even, like, really, like, beta test their app, just to have it installed for three days, and they paid for that too. So, you know, it's it's pretty nice. - 老实说，报酬完全没法比。我是说，我以前在 TaskFlow 做，赚十六美元左右。在那之后，我现在在 CampusJobs 工作，我赚十八美元。所以涨了两美元。但，你知道，比起……像自由职业这种，情况可能会更糟。这取决于我，以及我如何寻找客户。比如，通常当我找到一个客户时，你知道，我做他们要求的工作赚的钱，其实花不了多长时间。所以，那是短期的。我的意思是，有人甚至纯粹为了在他们的 Discord 机器人或游戏上创建一个账户就付钱给我。你知道吗？我为此拿到了钱，然后，比如，有人付钱给我甚至不是为了……真正的测试他们的应用程序，只是为了把它安装三天，他们也为此付了钱。所以，你知道，这挺不错的。

06:34-06:49 - Researcher: Wait, so people are paying you just to install an app for three days or create a Discord account? That's wild. How did you even find those gigs? And with the AI annotation at eighteen an hour — what kind of tasks are you actually doing? Like, labeling data, reviewing AI outputs, something else? - 等等，竟然有人付钱让你安装一个应用三天或者创建一个 Discord 账号？这太神奇了。你到底是怎么找到这些活儿的？ 还有那个时薪 18 美元的 AI 标注——你实际上在做哪种任务？比如数据标记、审核 AI 输出，还是别的什么？

06:53-07:28 - Marcus: So for those gigs, those are on Upwork. You know, people post, like, QA testing and stuff like that or, like, need, blah blah blah, quick test, make money, and, like, I applied to a few, then they got back to me. I was like, sure. Why not? You know, they all seem to be pretty pretty cool people. For in terms of annotation, though, I don't annotate anything for you. Just, like, I haven't started anything for the CampusJobs stuff. Because I literally like, I just got on board a few days ago. But from what it described, it was just, like, labeling, transcription, you know, stuff like that. - 至于那些活儿，是在 Upwork 上的。你知道，人们会发布类似 QA 测试之类的内容，或者说需要某某某、快速测试、赚钱，然后我申请了一些，他们就回复我了。我想着，当然可以，为什么不呢？你知道，他们看起来都是挺酷的人。不过就标注而言，我还没为你标注任何东西。只是，比如，我还没开始 CampusJobs 的任何工作。因为我真的是……几天前才入职。但根据描述，就是标记、转录，你知道，诸如此类的事情。

07:30-07:46 - Researcher: Got it, so you're just getting started with the campusjobs annotation work. Makes sense. Let me shift gears a bit — since you're deep into coding and working with AI annotation, I'm curious about the AI tools you actually use for your own work. In the past three months, how many AI products have you actually paid for yourself? - 明白了，所以你才刚刚开始在 CampusJobs 的标注工作。理解。 让我换个话题——既然你深耕编程并从事 AI 标注工作，我很好奇你在自己的工作中实际使用的 AI 工具。在过去的三个月里，你实际自费购买了多少款 AI 产品？

07:48-07:51 - Marcus: I'd say one or two. - 我会说一到两款。

07:54-08:01 - Researcher: Okay, one or two AI products in the past three months. Let's pin that down — is it one or two? And which products are we talking about? - 好的，过去三个月有一到两款 AI 产品。让我们确定一下——是一款还是两款？我们谈论的是哪些产品？

08:05-08:07 - Marcus: Give me one second. I can check. - 给我拿点时间，我可以查一下。

08:09-08:10 - Researcher: Take your time! - 没关系，你查吧！

08:17-08:23 - Marcus: Okay. So I pay for two right now. I think I used to pay for, like, three or four but right now, I pay for two. - 好的。我现在付钱买了两款。我想我以前付过三四款的钱，但现在，我付两款的钱。

08:25-08:31 - Researcher: Got it, so you're paying for two right now, but you used to pay for three or four. Which two are you currently paying for? - 明白了，所以你现在正为两款产品付费，但过去曾付过三四款的钱。你目前正在为哪两款付费？

08:38-08:44 - Marcus: ChatGPT and Copilot. I'm also looking into, like, paying for one of those those little, like, AI video ones. I'm also looking into that. - ChatGPT 和 Copilot。我也在考虑，比如，给那些类似 AI 视频工具中的一个付钱。我也在考虑那个。

08:47-09:00 - Researcher: ChatGPT and Copilot — makes sense for your coding work. And you're eyeing an AI video tool too. So with ChatGPT and Copilot, is this your first time paying for them, or have you renewed a few times already? Like, how many billing cycles have you been through? - ChatGPT 和 Copilot——对于你的编程工作来说确实合情合理。而且你也在关注一款 AI 视频工具。 那么对于 ChatGPT 和 Copilot，这是你第一次付钱吗，还是已经续费几次了？比如，你经历了多少个计费周期？

09:12-09:14 - Marcus: For ChatGPT, I'd say about like, I've been paying for it to close to a year. But it was kinda, like, off and on with ChatGPT with Copilot. I I just now started paying for it. - 至于 ChatGPT，我会说大约……我已经付了快一年的钱了。但在 ChatGPT 上是断断续续的。对于 Copilot，我才刚开始付费。

09:16-09:28 - Researcher: Interesting — so ChatGPT has been off and on for close to a year, and Copilot is brand new. When you say "off and on" with ChatGPT, what made you cancel and then come back? Like, what triggers you to turn it off versus turn it back on? - 有意思——所以 ChatGPT 断断续续用了一年，而 Copilot 是全新的。当你说 ChatGPT “断断续续”时，是什么让你取消后又回来？比如，什么会触发你关掉服务，又是什么让你重新开启？

09:31-11:41 - Marcus: It starters, it was the price and also because at the time, I was primarily using ChatGPT for my education work. You know, I know it's bad, you know, when have an assignment, you're just using AI to cheat on it. But you know, sometimes most of the times, actually, when I don't use it, it would be for something that's, like, unimportant. I'm like, oh, I already know this. Like, I'd rather it just, like, do it and just, like, show all the work and then, like, you know, like, maybe, like, one or two questions. I actually try to make sure that I know I can do it. But then, otherwise, you know, I just use the AI. That way I can be like, okay. Yeah. Like, I know what I'm doing, but, like, you know, AI is faster. I send in the piece of the work, and I just copy everything down. But, like, mainly for me, it was just the fact that, like, when I was using it for my educational work, like, I didn't need it as much, and I was, like, paying, I think, like, from a seventy to twenty dollars for it. So I was like, yeah. Like, let me let me stop paying for it for a little bit. And, like, when I needed to get for school, like, when I needed it for work, or, like, when I was, like, oh oh, crud. I'm I might need a little help with, like, setting up this coding application and you know, the the AI has limits on how much, like, screenshots you can send and then how long you can use GPT-5 without paying. And I noticed that when I would use a lower language model like GPT-4 on, um, ChatGPT, you know, when I would, like, use and send some of my code to it. Like, I because, like, when I send my code, like, I know what is giving me back. Like, I'm not just, like, sending it and being, oh, okay. Like, let's just paste it. I don't know what it does. But, like, sometimes when it was sending back, like, I would see, like, outdated it's like that for, for example, for JavaScript. Right? When I was coding I think it was when I was setting up my It was when I setting up my load balancing system using JavaScript. It would, like, recommend out of date packages to do certain functions or out of date or, like, complex ways to do certain things. So I'm just, like, why would I do it this way if I could just simply do it this way? Or, like, why are you making me duplicate the same three functions but, like, making it worse when you can just call this file which has this function? Why are you making me store fifteen different values of the same thing when I can just set up a simple JSON file? Which store is it? Or is it a simple dictionary on JSON that stores it. Or sorry. JS JS. - 首先是价格，也因为当时我主要把 ChatGPT 用于我的学业。你知道，我知道这不好，你知道，当有一个作业时，你只是用 AI 去作弊。但你知道，有时甚至大多数时候，实际上，当我不使用它时，是因为那是一些不重要的事。我想，噢，我已经知道这个了。比如，我宁愿让它去完成并展示所有的过程，然后，比如，可能一两个问题。我实际上会尝试确保我知道自己能做。但除此之外，你知道，我就用 AI。那样我可以觉得，好吧，是的。比如，我知道我在做什么，但是，比如，AI 更快。我发送一部分作业，然后我就把所有内容抄下来。但，对我来说，主要是因为，比如当我在学业中使用它时，我不需要那么频繁，而且我支付的费用，我想，大概是 17 到 20 美元。所以我想，是的。不如让我停止付费一段时间。然后，比如当我需要在学校使用它时，或者在工作需要它时，或者当我，噢，糟糕。我可能在搭建这个编程应用程序时需要一点帮助，你知道，AI 在你能发送多少截图，以及如果不付钱能使用 GPT-5 多长时间方面是有限制的。我注意到，当我在 ChatGPT 上使用像 GPT-4 这种较低层级的语言模型时，你知道，当我把一些代码发给它时，因为，当我发送代码时，我知道它返回给我的是什么。我并不是那种，噢，发过去，好吧，粘贴过来，完全不知道它是干嘛的。但，有时它发回来的内容，我能看出那是过时的。例如 JavaScript。对吧？当我编程时……我想是在我搭建我的……是在我用 JavaScript 搭建负载平衡系统的时候。它会推荐一些过时的包来写某些函数，或者推荐一些过时或复杂的写法来实现某些功能。我就想，如果我能简单地用这种方式实现，为什么要用那种方式？或者，你为什么要让我复制同样的三个函数还要写得更烂，明明你可以直接调用包含这个函数的文件？为什么要让我存储 15 个相同事物的不同值，明明我可以设置一个简单的 JSON 文件？哪个更好存？或者是一个简单的 JSON 字典。或者不好意思，是 JS，JS。

11:43-11:59 - Researcher: So you were bouncing between models and noticed GPT-4 was giving you outdated or overcomplicated code suggestions. That's frustrating when you actually know what you're looking at. When you resubscribed to ChatGPT, was it specifically to get access to GPT-4o or the newer models? And what made that worth the twenty dollars again? - 所以你在不同模型之间切换，并注意到 GPT-4 会给你过时或过于复杂的代码建议。当你其实很懂行的时候，这确实让人沮丧。 当你重新订阅 ChatGPT 时，是为了专门获得 GPT-4o 或更新模型的访问权限吗？是什么让它重新值得那 20 美元？

12:12-13:09 - Marcus: Well, I think they're on I'm pretty sure they're on GPT-5. Now GPT-4 is out of dated, but yeah, GPT-5 was worth it because like I said, when the difference between the five and the four was that the five was, like, way more to date and way more accurate. Like, I could confidently say if I was, like, hey, ChatGPT. You know, I'm having this issue with my load balancing system for where when my system is posting or, I'm sorry, when my server is posting to my API that it's requiring three more servers my load balancing system is pulling from the API. But then, like, for some reason, like, it's not scaling the three servers, and I send it, like, the the response that it gets when pinging the API from your services required, and it'll also paste the file that, like, you know, like, does the main server main well well, you just be specific. My backend handler dot js, I sent it to it. And it's like, oh, hey. You know? Is what's going on. You know? It's it's a capitalization error. You you didn't normalize the values here. It's like the small things, but it points it out. And then I know that, like, whatever encoded gives me return, like, it'll work, and it'll be more modern. - 嗯，我想他们现在应该有 GPT-5 了。现在的 GPT-4 已经过时了，但没错，GPT-5 是值得的，因为就像我说的，5 和 4 之间的区别在于，5 要跟得上时代得多，也准确得多。比如，我可以很有信心地说，如果我说，嘿，ChatGPT。你知道，我的负载平衡系统遇到了这个麻烦，当我的系统正在发布，或者，不好意思，当我的服务器向我的 API 发布信息说它需要增加三台服务器时，我的负载平衡系统正在从 API 获取数据。但是由于某种原因，它没有扩展那三台服务器，我发送了在向 API 发送请求时收到的响应，它还会粘贴那个运行主服务器处理程序的文件。好吧……具体点说，是我的 backend handler dot js，我发给它。它就会说，噢，嘿。你知道发生什么了吗。是大小写错误。你这里没有对数值进行归一化。这都是些小事，但它能指出来。然后我就知道，无论它返回什么代码，都会管用，而且更现代。

13:11-13:26 - Researcher: Ah, so GPT-5 (or the latest model) catches those tiny bugs and gives you modern, working code. That's a concrete value you can feel immediately. You're paying monthly for ChatGPT at twenty dollars, right? And what about Copilot — what's the payment setup there? Monthly, yearly, or something else? - 啊，所以 GPT-5（或最新模型）能抓住那些微小的 bug，并给你现代、可用的代码。这是你能立即感受到的具体价值。 你每月为 ChatGPT 支付 20 美元，对吧？那 Copilot 呢——那里的支付设置是什么样的？按月、按年，还是别的？

13:27-13:33 - Marcus: Um, it's it's the same. I have it for a month, and it's nineteen a month, nineteen dollars. - 嗯，是一样的。我用按月的，每月 19 美元。

13:35-13:51 - Researcher: So you're paying nineteen dollars a month for Copilot and twenty for ChatGPT — that's thirty-nine dollars total each month for AI tools. If you could choose any payment model for these tools, what would be ideal for you? Like, would you prefer monthly, yearly, pay-per-use, credits, or something completely different? - 所以你每月为 Copilot 付 19 美元，为 ChatGPT 付 20 美元——每个月在 AI 工具上总共花 39 美元。 如果你可以为这些工具选择任何支付模式，对你来说最理想的是什么？比如，你更喜欢按月、按年、按次付费、积分制，还是完全不同的模式？

14:03-14:58 - Marcus: Honestly, I prefer yearly. Like, I mean, the only reason I really pay for Microsoft 365 is that sometimes it's it's better because, you know, when I pay for Microsoft 365, it includes you know, Copilot, but then, like, a whole bunch of other stuff too, you know, Office and all. And, like, sometimes that's, useful for, like, when I'm doing, like, school and work. But, you know, like, I I prefer, like, yearly usually because I noticed that when people set up a yearly system, it's cheaper. It's just that sometimes, you know, like, I'm like, I I probably won't need this for a year. I don't feel inclined to, like, buy it. But, like, then sometimes, like, the monthly jump crazy. Like, some I was looking at something else. I forget what it was, but I was looking at subscribing to something, and it was, when you buy it for yearly, it was, like, the equivalent of, like, three dollars a month. But if you paid monthly, it was fifteen dollars. Like, sometimes, like, I might not like, let's say I have fifteen dollars, but I don't have the thirty You know? Like, I wouldn't be incentivized to pay or wait for the thirty if I need it now. Which, like, has me hooked up monthly payments. - 老实说，我更喜欢按年。我是说，我付 Microsoft 365 费用的唯一原因是有时它更好，因为，你知道，当我买 Microsoft 365 时，它包含了 Copilot，但也有一大堆别的东西，你知道，Office 之类的。而且，有时候那很有用，比如当我在处理学业和工作时。但，你知道，我通常更喜欢按年，因为我注意到当人们设置按年系统时，它会更便宜。只是有时，你知道，我会觉得，我可能一年都用不到这个。我没觉得有必要去买。但，有时候按月的涨幅太疯狂了。比如我之前看过的别的东西。我忘了是什么了，但我当时正考虑订阅某个东西，如果你按年买，相当于每月三美元。但如果你按月付，就是十五美元。比如，有时候我可能没有……比方说我有十五美元，但我没有三十美元。你知道吗？如果我现在就需要，我也没动力去为了付这一笔钱而等到凑够三十美元。这让我陷入了按月支付。

15:01-15:33 - Researcher: That makes total sense — yearly is cheaper per month, but if you don't have the lump sum upfront and you need it now, monthly wins even though it costs more overall. Cash flow matters when you're hustling to make three grand for tuition. So thinking about your overall attitude toward paying for AI products, which of these best describes you: 1) AI is a must-have for lifework, willing to pay continuously 2) It's very valuable, happy to pay as long as it's useful 3) Paying is acceptable, but I'm cautious about it 4) Not really willing to pay unless absolutely necessary or super cheap 5) Would never consider paying for AI Which one fits you best? - 这完全说得通——按年平摊到每月更便宜，但如果你手头没有预付那一整笔钱而且现在就要用，那么按月支付就赢了，尽管总成本更高。当你正为学费筹集 3000 美元时，现金流非常重要。 那么考虑到你对支付 AI 产品的整体态度，下列哪项最能描述你： 1) AI 是生活工作的必需品，愿意持续付费 2) 它非常有价值，只要有用就乐意付费 3) 付费是可以接受的，但我对此比较谨慎 4) 除非绝对必要或超级便宜，否则不太愿意付钱 5) 绝不考虑为 AI 付费 哪一个最适合你？

15:33-15:34 - Marcus: Definitely two. - 绝对是 2。

15:39-16:21 - Researcher: Cool, so you see real value and you're happy to pay when it's useful. That tracks with how you're using ChatGPT and Copilot for actual work that makes you money. Now, thinking about all the different things AI can do, I want to know what you'd actually pay for. From this list, pick your top THREE in order of priority — what you'd most willingly pay for: - Content creation (writing, rewriting, translation, copywriting) - Information organizing & summarizing (long docs, meetings, materials) - Learning & knowledge help (explaining concepts, learning paths, Q&A) - Work efficiency (automation, templates) - Data analysis & insights - Creativity & inspiration - Decision support (comparing options, risk analysis) - Programmingtechnical support - Personal assistant stuff (planning, reminders, task organizing) What are your top three? - 酷，所以你看到了真正的价值，并且在有用时乐于付费。这与你将 ChatGPT 和 Copilot 用于能帮你赚钱的实际工作非常契合。 现在，考虑到 AI 能做的所有不同事情，我想知道你实际上愿意为什么付费。从这个列表中，按优先级选出你的前三名——你最愿意为什么付费： - 内容创作（写作、改写、翻译、文案排版） - 信息整理与总结（长文档、会议、材料） - 学习与知识辅助（解释概念、学习路径、问答） - 工作效率（自动化、模板） - 数据分析与见解 - 创意与灵感 - 决策支持（比较选项、风险分析） - 编程/技术支持 - 个人助理事务（规划、提醒、任务组织） 你的前三名是什么？

16:22-16:30 - Marcus: So the first one would probably be information organizing, then work efficiency. And then programming slash technical support. - 第一名大概是信息整理，然后是工作效率。然后是编程/技术支持。

16:32-16:46 - Researcher: Interesting order! So information organizing is number one, then work efficiency, then programming support. I'm curious — why is information organizing the top priority for you? I would've guessed programming support might be first given all the coding work you do. - 有意思的顺序！所以信息整理排第一，接着是工作效率，然后是编程支持。 我很好奇——为什么信息整理是首选？我本以为考虑到你做的所有编程工作，编程支持会排在第一。

17:32-18:06 - Marcus: Yeah. And, you know, you'd be right to think so. But for me, it's also really important that when I ask my model anything, first of all, like, the information provided to me is accurate. And then on top of that, the information is provided to me in a way that is easy to read, and I can just get the information. Because sometimes it's annoying if I'm asking a programming question, and, like, yeah, it gives me the answer, which is why I'm like, yeah. I why it'd be important where you think it'd be important, but it's also annoying when I'm, like, I have to read through, like, a paragraph of it going, like, yeah. So JavaScript, which was made in blah blah blah blah, you know, telling me, like, a whole bunch of details and other stuff. And I'm just, like, I'm looking for the fix. Like, I'd prefer to see the fix or, like, what would directly help me to fix it. But then instead, like, it's I'm just getting, like, all this other feedback, which isn't necessary. It's like, hey, Marcus. Thank you for asking me how I could fix this. So upon your request, it's just giving, like, a whole bunch of this, and I just prefer it being, so in order to fix it, and then it shows me the fix, then afterwards, it's explaining, and it's like, yeah. So you actually did this wrong. We should do this and this and that, or you know, it's it's easier for me that way. Like, usually, when I ask my AI anything, I already, like the first thing in the conversation is I tell them how I want it, like, formatted. Because in my opinion, how the AI, AI in general is, like, the information is provided. It's just inefficient. Just my opinion. - 是的。你知道，你这么想也是对的。但对我来说，同样非常重要的一点是，当我问我的模型任何问题时，首先，提供给我的信息要是准确的。最重要的是，信息要以一种易于阅读的方式提供给我，让我能直接获取信息。因为有时候，如果我问一个编程问题，是的，它给了我答案，这也就是为什么我……为什么你会觉得它重要，但这也很烦人，比如我得读完一整段类似“是的，JavaScript，它是在某某年由某某某发明的……”的内容，你知道，告诉我一大堆细节和其他东西。我只想找修复方案。我更想看到修复方案，或者直接能帮我修复它的东西。但相反，我得到的全是这些没必要的反馈。就像“嘿，Marcus。感谢你询问我如何修复这个。根据你的请求……”它只是给了一大堆这种东西，我更希望它直接说“为了修复它，你应该……”，然后展示修复方法，然后再向我解释，比如“是的。你其实这里做错了。我们应该这样做……”或者你知道，对我来说那样更容易。通常，当我问我的 AI 任何事情时，我已经……对话的第一件事就是告诉它们我想要的格式。因为在我看来，AI 普遍提供信息的方式是低效的。纯属个人观点。

18:08-18:26 - Researcher: That's super insightful — you're basically saying the presentation of information matters as much as the accuracy. You don't want fluff or backstory, you want: here's the fix, here's why it was wrong, done. So you actually prompt it upfront with formatting instructions every time? What does that look like — like, what do you tell it at the start of a conversation? - 这非常深刻——你基本上是说信息的呈现方式和准确性一样重要。你不需要废话或背景故事，你想要的是：这是修复方案，这是为什么原来错了，搞定。 所以你每次都会预先向它输入格式指令吗？那是什么样的——比如，你在对话开始时会告诉它什么？

18:29-19:05 - Marcus: It depends. My most recent example would be, like, when I was getting really agitated because it was doing what I was, you know, telling you before. Was giving me a whole bunch of fluff. I was like, hey. So whenever I ask you a question, answer it. Just give me the answer and nothing else, and it'll just be like, okay. The answer choice is a. Oh, you need to do this. And then, like, in the case where, like, I want an explanation, I'm like, tell me the answer choice, nothing else. And then after the answer choice, provide me an explanation on how you came up with the answer. That way, if it's something, like, important, like, I would need to verify that, like, it's right, then, like, I asked, like, that's some of the information that but I always ask for the answer first. - 看情况。我最近的一个例子是，当我变得非常烦躁的时候，因为它确实在做我之前告诉你的那种事。不停给我废话。我就说，嘿。无论我问你什么问题，直接回答。只给我答案，别的什么都不要。它就会说，好吧。答案选项是 a。噢，你需要做这个。然后在我想听解释的情况下，我会说，告诉我答案选项，别的什么都不要。然后在答案选项之后，向我说明你是如何得出这个答案的。那样的话，如果是类似重要的事，我需要核实它是对的，那我就会要求……那是其中的一些信息……但我总是要求先给答案。

19:07-19:24 - Researcher: So you've literally trained it to give you the answer first, then explain only if you ask. That's efficient. When you first started paying for Copilot — your most recent AI tool purchase — where did you first hear about it or decide to try it? Was it through social media, a friend, searching for solutions, or something else? - 所以你实际上训练它先给答案，只有在你要求时才解释。这很高效。 当你第一次开始为 Copilot 付费时——这是你最近一次购买 AI 工具——你最初是从哪里听说它或决定尝试它的？是通过社交媒体、朋友、搜索解决方案，还是别的？

20:05-20:22 - Marcus: Um, it was through social media. Like, Copilot, there was a lot of advertisements for it on stuff like Instagram. Um, it was also really annoying because, like, you know, I have a Windows PC, and so they just automatically preinstall it, especially when I updated it with Windows eleven. And it's, like, it's integrated in pretty much everything I use. You know? Like, when I'm on GitHub, they're, like, GitHub Copilot. You know? When I'm on Visual Studio code, which is one of the main things I use to code besides IntelliJ. But, you know, when I'm on Visual Studio, you know, it's code with Gemini code with Copilot. It was just all there, and I'm just like it just just gets annoying. I mean, they're kinda everywhere. So then I eventually I I tried it out, especially since it was, like, kinda easier to use since it's, like, already all like, on my PC, and I can just install it on my PC if it's already on there instead of me having to visit a website. So, you know, if I wanted to do anything, I just, like, touch something on the top or I'm sorry, the bottom right of my taskbar, and then boom. It just pops up. Yeah. - 嗯，是通过社交媒体。比如 Copilot，在 Instagram 之类的地方有很多它的广告。呃，它也挺烦人的，因为，你知道，我有一台 Windows 电脑，所以他们会自动预装它，尤其是在我更新 Windows 11 之后。而且它几乎集成在我使用的所有东西里。你知道吗？当我在 GitHub 上时，他们会说 GitHub Copilot。当我在 Visual Studio Code 上时，那是除了 IntelliJ 之外我主要用来写代码的东西。但你知道，当我在 Visual Studio 上时，它是“用 Gemini 编程”或“用 Copilot 编程”。它无处不在，我就觉得……就开始觉得烦了。我的意思是，它们简直无孔不入。所以后来我最终尝试了它，尤其是因为它用起来更简单，因为它已经在我的电脑上了，既然它已经在那儿了，我可以就在电脑上安装它，而不用去访问某个网站。所以，如果我想做任何事，我只需点一下任务栏右下角的东西，然后砰。它就弹出来了。是的。

20:24-20:58 - Researcher: So it was everywhere — ads, baked into Windows 11, GitHub, VS Code — basically impossible to ignore. And the convenience of it being right there on your taskbar made it easy to just try. When you actually decided to pay for Copilot, what was the main reason? Was it: 1) Unlocking features or usage limits that were blocked 2) Getting access to a stronger model or advanced capabilities 3) Hit a specific need or limitation while using it 4) A must-have for work or school 5) Needed it to make money or bring in revenue 6) Just thought it was good and wanted to supportuse long-term 7) There was a discount or trial ending 8) Curiosity or novelty Which one pushed you over the edge? - 所以它无处不在——广告、内置在 Windows 11 中、GitHub、VS Code——基本上不可能被忽视。而且它就在你任务栏上的便利性让你觉得很容易去尝试。 当你真正决定为 Copilot 付费时，主要原因是什么？是因为： 1) 解锁被限制的功能或使用限额 2) 获得更强大的模型或高级功能 3) 在使用过程中遇到了特定的需求或限制 4) 工作或学业的必需品 5) 需要它来赚钱或增加收入 6) 觉得它很好，想长期支持/使用 7) 有折扣或试用即将结束 8) 出于好奇或新鲜感 哪一个促使你最终付了钱？

21:03-21:04 - Marcus: One. - 第一点。

21:06-21:15 - Researcher: So you hit the usage limits or locked features and needed to unlock them. When that happened, did you hesitate at all before paying, or was it pretty much an instant decision? - 所以你遇到了使用限制或锁定的功能并需要解锁它们。当那种情况发生时，你在付钱前有过犹豫吗，还是说基本上是瞬间的决定？

21:33-21:51 - Marcus: I I I will admit I did hesitate because like I said, I was looking at other options. But then I, like, I looked it up and, you know, people said it was pretty good. You know, I was looking for some audio and video generation. And it turns out Copilot, it could do all that. And it had, like, experimental features that I could try too. I was like, pretty interested in. So, you know, like, after looking at that, and it was, like, me paying for something that I already use and just, like, adding additional features and models versus, like, me, like, paying for a whole new service and then me using three platforms instead of me just sticking at two, except I just pay for both of them now. So, yeah, it was - 我，我承认我确实犹豫过，因为就像我说的，我当时在看其他的选择。但随后我查了一下，你知道，人们说它挺不错的。你知道，当时我正在找一些音频和视频生成工具。结果发现 Copilot 也能做这些。它还有我可以尝试的实验性功能。我对此挺感兴趣的。所以，你知道，在看了那个之后，感觉这就像是为我已经在使用的东西付费，只是增加了额外的功能和模型，而不是我去购买一个全新的服务，那样我就得用三个平台，而不是继续留在两个平台上，只不过我现在要为这两个都付钱。所以，是的，就是这样。

21:56-21:57 - Researcher: So you weighed it against - 所以你在……之间权衡了。

21:57-21:59 - Marcus: was - 是的。

22:01-22:16 - Researcher: adding new tools versus just upgrading what you already had. And Copilot won because it bundled audio, video generation, and experimental features all in one place you were already using. What tipped the scales in the end — was it the bundled features, the convenience, or something else that made you finally commit? - 权衡了增加新工具还是直接升级已有的。最后 Copilot 赢了，因为它把你已经在用的各种功能——音频、视频生成和实验性功能都捆绑在了一起。 最后是什么起到了决定性作用——是捆绑功能、便利性，还是别的什么让你最终决定掏钱？

22:55-23:07 - Marcus: Oh, it was more so the effect. It's kinda like with ChatGPT. When you pay, I noticed that when you don't pay so let me word it this way. When you don't pay for their subscription service, everything is, like, loaded on your computer end, which is really annoying and on your computer when you're, like, doing stuff that's heavy. Like, if I'm coding, and I have to run five Java servers, right, Each Java server is gonna need a minimum of a gigabyte of RAM. Right, on top of that. Like, if I'm using the AI tool on my computer, you know, after the the conversation gets a little long and, you know, the model has to to remember, um, it starts to become that the model like, it starts, like, using a lot of resources on my computer. It turns out, like, when I use browsers like Opera or, like, Chrome, you know, it goes from one gigabyte to using, like, four or five. - 哦，更多是效果方面的原因。有点像 ChatGPT。当你付费时，我注意到——这么说吧。当你没付订阅费时，所有东西都像是加载在你的电脑端，这真的很烦人，尤其是当你在电脑上做一些繁重的任务时。比如，如果我在写代码，我需要运行五个 Java 服务器，对吧，每个 Java 服务器至少需要 1GB 的内存。对吧，除此之外，如果我在电脑上使用 AI 工具，你知道，对话变得有点长之后，模型需要去“记住”前面的内容，它就开始消耗我电脑的大量资源。结果就是，当我使用 Opera 或 Chrome 之类的浏览器时，内存占用会从 1GB 飙升到 4GB 或 5GB。

23:09-23:27 - Researcher: Ah, so the free version was running locally and eating up your RAM — which is a huge problem when you're already running multiple Java servers. Paying for it offloaded the processing to the cloud, so your machine could actually breathe. That's a really practical reason. So performance and resource management was the final push, not just the features themselves? - 啊，所以免费版是在本地运行并吃掉了你的内存——当你已经在运行多个 Java 服务器时，这是一个巨大的问题。付费后将处理过程卸载到了云端，这样你的机器就能喘口气了。 这是一个非常实际的理由。所以性能和资源管理是最后的推动力，而不仅仅是功能本身？

23:29-23:43 - Marcus: Yeah. Exactly. Like, the features were really nice, but the fact that it gave my computer a massive performance boost allowed me to do more. Also being able to use the AI and, like, me not having to worry about my computer liking. You know, it was it was just beneficial to me. - 是的。没错。功能确实很棒，但它给我的电脑带来了巨大的性能提升，这让我能做更多事情。而且能够使用 AI 却不用担心电脑卡顿，你知道，这对我来说非常有益。

23:46-24:13 - Researcher: Makes total sense — you literally couldn't afford the performance hit when you're running multiple servers and coding at the same time. Okay, last thing on this: when you think about what makes you willing to pay for an AI chat assistant, which of these factors matter most to you? Rank your top three: - Feature set - How often you use it - Price vs. value (bang for buck) - Ease of use - How well it actually works (effectiveness) - Uniqueness (can't get it elsewhere) - Quality of the output What's your top three in order? - 完全理解——当你在同时运行多个服务器和写代码时，你确实无法承受性能损耗。 好，关于这个的最后一件事：当你思考是什么让你愿意为 AI 聊天助手付费时，以下哪些因素对你最重要？请排列出前三名： - 功能集 - 使用频率 - 价格与价值（性价比） - 易用性 - 实际效果（有效性） - 唯一性（在别处买不到） - 输出质量 按顺序排列的前三名是什么？

24:13-24:20 - Marcus: Quality of the output is definitely first. And then it would be feature set, and then definitely price versus value. - 输出质量肯定排第一。然后是功能集，最后绝对是价格与价值。

24:22-24:37 - Researcher: Output quality, feature set, then price vs. value. That tracks perfectly with everything you've said — you need it to actually work and give you modern, accurate code first, then it needs the right tools, and then it has to be worth the money. Why is output quality so critical for you compared to everything else? - 输出质量、功能集，然后是价格与价值。这和你刚才说的完美契合——你首先需要它能真正起作用，给你提供现代、准确的代码，然后它需要有合适的工具，最后它必须物有所值。 为什么输出质量对你来说比其他一切都如此重要？

25:58-26:52 - Marcus: Because, like I said, it's kinda annoying when the information isn't formatted correctly. Like, imagine, right, you asked your AI a simple question, and it gives you seven paragraphs. And after you sift through all seven paragraphs, it turns out the answer that it gives you is wrong. That gets really annoying, and I've had it happen before. You know? Like, I've had it happen when I was coding. You know? I asked for help. It gave me, quote, unquote, help, and then, like, I pasted the code inside my, um, my IDE, and it had, like, seventeen new issues that weren't there before. And I was like, woah. And it wasn't just, like, simple, like, oh, like, missing packages or, like, it use a library that you don't have installed. No. It it wasn't anything like that. It just added things that were, so deprecated that I was like, what? I've I had don't even see people use it anymore. Like, nobody uses this. So, you know, like, it's it's the quality of work definitely matters because if it gives me bad quality and bad information, I have to take more time out of my day, which I'm I'm trying to save by using the system. I have to take more time out of my day to either find my own fixes the AI system couldn't or I need to find I need to take more time out of my day trying to fix whatever it decided to break. It's also, um, a pattern that I've noticed that, like I said, especially with Cody. Because, um, like, most of our perspective on this is, like, has to do with code because I use it a lot when I'm, um, developing stuff, but it's also a pattern I've noticed. Like, the AI assistant will just, like, it'll just, like, leave certain things out even when because like I said, I trained it myself beforehand. I'm like, give me all the code, nothing else, no comments. It's just the entire code. So I send my file. It gives it back. And then, like, core features are missing. And I don't notice until, like, I already build the jar and I put it in my server and I run the server, and I'm like, wait a minute. Where did like, how would these features go? And then, like, sometimes I don't notice and, like, I I'm, like, using AI for, like, multiple parts of the code, I start layering and layering. And I'm like, wait a minute. It removed the double-jump command, like, five versions ago. And, like, it was just hidden in all the stuff, and I didn't expect it. And I'm like, what? So then I have to, like, go back and, like, either, like, re add everything I did or have to reverse all my changes. You know, get all the double-jump system I set up and just, like, implement it into the most current version. You know, that just takes a little more time. - 因为就像我说的，当信息格式不正确时会很烦人。你看，想象一下，你问了 AI 一个简单的问题，它给了你七个段落。当你筛选完这七个段落后，发现它给你的答案是错的。这变得非常恼火，我以前就遇到过。你知道吗？我在写代码时遇到过这种情况。我寻求帮助，它给了我所谓的“帮助”，然后我把代码粘贴到我的 IDE 里，结果出现了 17 个之前没有的新问题。我当时就想，哇。而且这不仅仅是简单的，比如漏了包或者用了没安装的库。不，完全不是那样。它只是添加了一些非常陈旧、已经弃用的东西，以至于我心想：什么鬼？我甚至都没见过还有人在用那些东西。比如，没人用这个。所以，你知道，工作的质量绝对重要，因为如果它给我糟糕的质量和错误的信息，我就得从我原本想通过系统省下来的时间里抽出更多时间。我要么得花更多时间去找 AI 解决不了的修复方案，要么得花更多时间去修理它搞坏的地方。这也是我注意到的一种模式，就像我说的，特别是在 Cody 上。因为，我们的大部分观点都和代码有关，因为我在开发东西时经常用它，但这也是我注意到的一种模式。AI 助手会遗漏某些东西，甚至即便我事先训练过它。我会说：“给我所有的代码，不要别的，不要注释。只要完整的代码。”然后我把文件发过去，它传回来，结果核心功能缺失了。我直到编译成 jar 包并放入服务器运行后才注意到，我心想：“等等，怎么回事？这些功能跑哪去了？”有时我没注意到，而且我还在代码的多个部分使用 AI，开始一层又一层地叠加。然后我想：“等等，它在五个版本之前就把‘二段跳’命令给删了。”它就藏在所有东西里，我根本没预料到。我当时就懵了。然后我不得不回去，要么重新添加我做过的一切，要么回滚所有的更改。把我设置好的那个二段跳系统找出来，然后再重新实现在当前版本里。你知道，那就要多花不少时间。

26:55-27:16 - Researcher: Wow, so bad output doesn't just waste time — it creates cascading problems that you don't catch until way later, and then you're untangling layers of broken code. That's way worse than just getting no answer at all. Last question: you're using ChatGPT and Copilot together. Can you describe their relationship using a work or family metaphor? Like, who's the boss, who's the assistant, how do they work together in your system? - 哇，所以糟糕的输出不仅浪费时间——它还会产生连锁反应，你直到很久以后才会发现，然后你得去理顺那些破碎代码的层级。这比完全得不到答案还要糟糕得多。 最后一个问题：你同时在使用 ChatGPT 和 Copilot。你能用工作或家庭的比喻来描述它们的关系吗？比如，谁是老板，谁是助手，它们在你的系统中是如何协作的？

28:19-28:56 - Marcus: Oh, yeah. ChatGPT is definitely the goat. I I love ChatGPT. It's the best. You know, um, I use ChatGPT for pretty much everything. In my opinion, it's faster and more efficient than Copilot. But then, like, Copilot is kinda like that, like, one little brother or sister sibling in general that's kinda like annoying but you love so much, that's Copilot. Because, like, sometimes Copilot will mess up, and it'll be like, I'm not allowed to do this. And all I, like, like, all I do is, like, send a piece of code, and it's talking about a safety violation. Meanwhile, the code is just something simple like, oh, yeah. Ping the API server. Make sure it's running. I've done that before. I sent it to Copilot, and Copilot was like, I'm not allowed to perform any inappropriate content. And I'm like, what is inappropriate about it? Like, all it's doing is not even, like, you're sending, like, a dozen packets. Like, it's just, like, keep alive. Like, you just you would just ping it once. It was a singular ping. It pings it once, and then it's done for, like, an hour or two. You know, I was just trying to test something. So, like, yeah. And they work pretty well. I mean, like I said, Copilot's, like, integrated in my system. Like, I don't usually mind having a ChatGPT tab. I'm I'm not even sure ChatGPT has a PC app. They might by this point, but you know, I don't mind, um, having ChatGPT open in my web browser because I noticed that versus Copilot and that Copilot might take up, like, five, six hundred megabits on my computer, which isn't really that bad in the grand scheme of things versus, like, the four or five gigs it used to take up. But then when I, like, open a tab of, like, ChatGPT, it literally takes nothing at all. Like, absolutely nothing. It it it doesn't even take two hundred it's crazy how they do it. I don't know how. - 哦，是的。ChatGPT 绝对是 GOAT（史诗级最强）。我超级爱 ChatGPT。它是最棒的。你知道，我几乎所有事情都用 ChatGPT。在我看来，它比 Copilot 更快、更高效。而 Copilot 有点像那种烦人的小弟或小妹，虽然有点烦人，但你又非常爱它，这就是 Copilot。因为，有时 Copilot 会搞砸，它会说：“我不被允许做这个。”而我做的只是发送一段代码，它就说什么安全违规。实际上，代码只是很简单的东西，比如：“噢，Ping 一下 API 服务器，确保它在运行。”我以前做过这个。我发给 Copilot，Copilot 说：“我不被允许执行任何不当内容。”我就想：“这有什么不当的？”它甚至连一打数据包都没发。它只是，比如，保持连接。你只需要 Ping 一次。就一次 Ping，然后一两个小时内就完事了。你知道，我只是想测试些东西。所以，是的。它们协作得挺好。我是说，就像我说的，Copilot 集成在我的系统里。我通常不介意开一个 ChatGPT 面板。我甚至不确定 ChatGPT 是否有电脑版应用。现在可能有了，但你知道，我不介意在浏览器里开着 ChatGPT，因为我注意到，相比之下，Copilot 可能会占用我电脑五、六百兆内存，从大局来看这其实并不算糟，比起它以前占用的 4GB 或 5GB 来说。但当我打开一个 ChatGPT 标签页时，它几乎不占什么资源。绝对没什么占用。它甚至连 200MB 都不到，我不知道他们是怎么做到的，太疯狂了。

29:02-29:23 - Researcher: Ha! So ChatGPT is the GOAT — faster, more efficient, barely uses resources — and Copilot is the annoying little sibling that sometimes freaks out over nothing but you still keep it around because it's useful and already built into everything. Alright, shifting gears now. What's the maximum amount per month you'd be willing to pay for AI products total — like across all of them combined? What's your ceiling? - 哈哈！所以 ChatGPT 是 GOAT——更快、更高效、几乎不占资源；而 Copilot 是那个有时会无缘无故抓狂的烦人小老弟，但你还是留着它，因为它很有用，而且已经内置在所有地方了。 好，现在换个话题。你每月愿意为所有 AI 产品支付的最高总额是多少？你的上限是多少？

29:37-31:04 - Marcus: Honestly, believe it or not, mine would probably be around fifty dollars, but that's only if it has everything I'm asking for. Like, the whole reason to have all of them separate is because, like, some app like, sometimes ChatGPT does things like Copilot Copilot does things horrible. And I know ChatGPT does it, like, ten times better. So, you know, it it just depends. Like, if I can, like, set up a system on whatever AI I'm paying for, and I'm like, okay. Yeah. It's like I don't know. Like, maybe, like, groups or something. Like, if I, like, select, like, coding, like, the the chat about coding, then, you know, when I'm talking to the AI assistant, it knows that since it's all about coding, I want all the code and, like, everything I added should be in there, you know, and no comments or anything or no explanations. Just send me the code. You know, stuff like that would be nice. You know? And, also, like, the AI assistant could, like I don't know. Like, it generate photos for me, generate videos for me, generate audio files for me. You know? That's that's all nice. That's all things that I really need pretty much in, like, increased history because I noticed that on both AI models, they have a limit on history. I mean, obviously, like, that the AI can remember, and it kinda gets annoying. Like, I would pay more for, like, longer history because it it isn't it's not even, like, the amount of, like, conversations or messages that you sent with each other that's counted. It's the length of all the messages in the conversation. So, like, naturally, with coding, I mean, like, sometimes I send things that are, like, four hundred lines long. And then after, like, maybe, like, like, five conversations about it, like, the AI generally, like, forgets stuff already because, like like I said, like, the code is, like, really, really long and it gets annoying. - 老实说，信不信由你，我的上限可能在 50 美元左右，但前提是它得包含我要求的所有功能。之所以把它们分开，是因为有时 ChatGPT 能做到的，Copilot 做得极烂。而我知道 ChatGPT 能做得好上十倍。所以，这取决于情况。比如，如果我能在付过费的 AI 上设置一个系统，比如……我不知道，也许像分组之类的。如果我选择了“代码”，即关于代码的聊天，那么当我和 AI 助手交谈时，它知道既然全是关于代码的，那我就要所有的代码，我添加的每一个东西都应该在里面，你知道，不要注释，也不要解释。直接把代码发给我就好。你知道，如果有这种功能就太棒了。而且，AI 助手还可以，比如，我不知道，为我生成照片、生成视频、生成音频文件。你知道吗？这些都很棒。这些都是我非常需要的功能，还有增加的历史记录。因为我注意到两种 AI 模型在历史记录上都有上限。我是说，显然 AI 能记住的东西是有限的，这有点烦人。我愿意为更长的历史记录支付更多费用，因为计算的甚至不是你们之间对话或消息的数量，而是对话中所有消息的总长度。所以，很自然地，对于写代码来说，我有时发送的东西有 400 行长。在进行了大概 5 次对话后，AI 通常就已经开始遗忘东西了，因为就像我说的，代码非常非常长，这变得很烦人。

31:08-31:30 - Researcher: So fifty dollars is your ceiling, but only if it's an all-in-one that does everything — coding with your custom formatting rules, photovideoaudio generation, and way longer conversation memory so it doesn't forget your 400-line code files after five exchanges. That fifty dollars — how did you land on that number? Is it based on a percentage of your income, or comparing it to other subscriptions you pay for, or just what feels right? - 所以 50 美元是你的天花板，但前提是它是一个全能工具——能按你的自定义格式规则写代码、能生成图文音视，并且拥有长得多的对话记忆，这样它就不会在五轮交流后就忘掉你那 400 行的代码文件。 那 50 美元——你是怎么得出这个数字的？是基于你收入的百分比，还是参考了你支付的其他订阅服务，或者只是感觉合适？

32:03-32:05 - Marcus: It's kinda based on what feels right and also, like, it's just the I mean, like, what I'm asking for is pretty much a combination of what everybody does for me separately. Like, you know, like, that combined would just be, like, amazing. Like, I would love to be able to use Copilot without, like, it randomly flagging nothing as something, you know, like, causing an error, and I can't use it. You know? And if that wasn't an issue in a different service and it has literally everything that I was asking for both models, then, yeah, I'd pay for it. That's why I added the extra ten bucks because, like, I'd be paying for quality. - 主要是基于感觉，而且这也……我是说，我所要求的基本就是把别人分别帮我做的事情组合在一起。如果能结合起来就太神奇了。比如，我希望能用 Copilot 而不被它无缘无故地把正常内容标记为违规，导致报错甚至无法使用。你知道，如果另一个服务不存在这个问题，并且拥有我向两个模型要求的所有功能，那么，是的，我会付钱。这就是为什么我额外加了 10 美元，因为我是在为质量买单。

32:08-32:31 - Researcher: That makes sense — you're basically saying if one tool could replace both ChatGPT and Copilot and fix Copilot's annoying false flagging issues, you'd pay a premium (an extra ten bucks) for that convenience and quality. Now, you mentioned you're interested in an AI video tool but haven't subscribed yet. What price would make you actually pull the trigger and try it? Like, how cheap would it need to be or what discount would get you to subscribe? - 明白了——你基本上是说，如果一个工具能同时取代 ChatGPT 和 Copilot，并解决 Copilot 烦人的误报问题，你愿意为这种便利和质量支付溢价（额外的 10 美元）。 现在，你提到你对 AI 视频工具感兴趣，但还没订阅。什么样的价格会让你真正下单尝试？比如，它得便宜到什么程度，或者什么样的折扣会让你订阅？

32:33-33:27 - Marcus: I mean, I prefer to use stuff with a free trial, but I've noticed that most things with a free trial is, like, really limited on what I'm able to do. Like, I just wanna, like, be able to see, like, what I'm able to do with it, but it's not like but it's not like I desperately need one, but it would just be, like, used for, like, if I'm, like, doing something for my contracting company and, like, just want, like, a simple, like, a voice over video and stuff like that. So it just it wouldn't be anything crazy. It'd probably be something that, like, I use for, like, a month or two. But, like, you know, it's not something that's, like, very, very important to me, but it would be nice. And I don't know. Like, something cheap. Like, I prefer, like, if there is, like, a plan for, like, something small, and then, like, I'm gonna give you, like, five, ten dollars for it, and I can just get the simple stuff I need. It's not like I'm gonna need, like, crazy video generation for it. It's like, people running and stuff like that. Like, it might just be, like, simple, like, video voiceovers or, like, a simple, like, narrated video with just, like, text. You know? Like, nothing crazy. - 我的意思是，我更倾向于使用有免费试用的东西，但我注意到大多数有免费试用的东西在功能上都非常有限。我只是想看看我能用它做什么，并不是说我迫切需要一个，但如果我为我的承包公司做点事，只是想要一个简单的配音视频之类的，它就能派上用场。所以它不需要太离谱。它可能是我只用一两个月的东西。虽然它对我来说不是非常非常重要，但如果有的话会很棒。我不知道，便宜点的。比如，如果有一个针对小需求的方案，我付个 5 到 10 美元，就能得到我需要的简单功能。我并不需要那种疯狂的视频生成，比如人们在跑步之类的。可能只是简单的视频配音，或者只有文字的简单解说视频。你知道，不需要太夸张。

33:29-33:47 - Researcher: So you'd pay five to ten dollars for a basic tier that does simple stuff — voiceovers, text overlays, basic narration videos — nothing fancy. You don't need Hollywood-level generation, just practical tools for client work occasionally. Is there a specific video AI tool you've been eyeing, or is this more of a "when I need it, I'll find one" situation? - 所以你愿意花 5 到 10 美元买一个基础档位来做简单的事情——配音、文字叠加、基础解说视频——不需要花哨。你不需要好莱坞级别的生成效果，只需要偶尔用于客户工作的实用工具。 有没有某个你一直在关注的特定视频 AI 工具，还是说这更多是“等我需要时再去寻找”的情况？

33:50-33:55 - Marcus: Definitely, when I need it, I'll find one situation. I haven't really had, like, an immediate use to find one. - 绝对是“等我需要时再去寻找”。我目前还没有立即需要用到它的地方。

33:57-34:15 - Researcher: Got it — it's on the back burner until a client project actually needs it. Let's flip to the tools you're already paying for. Take ChatGPT, since that's your main one. If they raised the price, how much would it have to go up before you'd seriously consider canceling? Like, what's the breaking point — a specific dollar amount or percentage increase? - 明白了——这事儿在后备状态，直到某个客户项目真正需要它。 让我们看看你已经在付费的工具。以 ChatGPT 为例，因为它是你的主力。如果他们涨价，价格涨到多少你会认真考虑取消？你的临界点是多少——具体的金额或百分比？

34:17-34:31 - Marcus: I think the most I would go is, like, thirty five to forty. Like I said, ChatGPT, in my opinion, is fifty times more worth it than Copilot. And for Co Pilot, I say, like, my limit would be, like, twenty five dollars honestly. - 我觉得我最高能接受 35 到 40 美元。就像我说的，在我看来，ChatGPT 的价值比 Copilot 高 50 倍。对于 Copilot，我说实话，我的极限大概是 25 美元。

34:34-34:51 - Researcher: So ChatGPT could go up to $35-40month and you'd still keep it because it's that good, but Copilot maxes out at $25 before you'd drop it since it's the annoying little sibling. Right now ChatGPT is $20month, right? So they could raise it by $15-20 and you'd stick with it. What makes ChatGPT worth almost double what Copilot is to you? - 所以 ChatGPT 可以涨到每月 35-40 美元你仍然会保留它，因为它太好用了，而 Copilot 超过 25 美元你就会放弃它，因为它只是那个烦人的小老弟。 现在 ChatGPT 是每月 20 美元，对吧？所以他们可以涨 15-20 美元你也会坚持使用。是什么让 ChatGPT 在你心中的价值几乎是 Copilot 的两倍？

35:34-35:35 - Marcus: Like I said, the the less amount of lag that it uses and how it's super fast. It's like, I I ask it a question, and it'll be, like, thinking for, like, at most, like, five seconds, and, like, everything is just out there. And I'm like, woah. Like, it appears so quickly. Like, I don't my brain doesn't even process how fast it's getting out that information. It's just really nice. You know, it's really nice to be able to just, like, ask something and then know that, like, with, like, pretty much ninety five percent certainty that, like, ChatGPT is gonna get me right. Like, I'll be honest. I'll be honest. I I did use it for, you know, some educational things. And, like, when I did, like, I got a hundred on pretty much everything I touched. Like, it was it was amazing. - 正如我所说，它运行时占用的资源更少，而且超级快。就像我问它一个问题，它最多思考 5 秒钟，所有东西就都出来了。我当时就想，哇。它出现得太快了，我的大脑甚至都还没反应过来它是如何以这么快的速度输出信息的。这真的很棒。你知道，能够随口问点什么，并且有 95% 的把握确定 ChatGPT 会给我正确的答案，这感觉真的很好。说实话，我确实在一些教育相关的事情上用过它。当我用它时，我几乎在每一项任务上都能得满分。真的太不可思议了。

35:38-36:04 - Researcher: So it's the speed, the accuracy (95% certainty), and the fact that it just works — whether it's coding or schoolwork, it delivers fast and correctly. That reliability is worth way more to you than Copilot's integration convenience. Last question: imagine a brand new AI product launches tomorrow that's absolutely perfect for you — does everything you need exactly how you want it. What's the absolute maximum monthly price you'd pay for that dream tool? Is it still fifty dollars, or would you go higher for perfection? - 所以是速度、准确性（95% 的把握）以及它确实管用——无论是写代码还是学校功课，它都能快速准确地交付。这种可靠性对你来说比 Copilot 的集成便利性值钱得多。 最后一个问题：想象明天推出了一款全新的 AI 产品，对你来说绝对完美——按照你想要的方式完成你需要的每一件事。你愿意为这个梦想工具支付的绝对最高月价是多少？还是 50 美元，还是会为了完美而支付更高？

36:07-36:33 - Marcus: It it'd still be fifty dollars. I mean, like, if if I was like, a full scale contracting company with, like, dozens and dozens of clients where, like, to the point where, like, I need something more, then, like, of course, I'd pay it. But, you know, like I said, I'm I'm, like, a small contractor right now. Like, I have at most, like, like, what, like, ten clients right now. I mean, really well paying clients, but, like, ten clients right now. So, yeah, fifty dollars would be my max without me possibly, like, losing revenue or anything like that. - 还是 50 美元。我的意思是，如果我是一个拥有几十个客户的大规模承包公司，到了需要更强大工具的程度，那我当然愿意多付钱。但正如我所说，我现在只是一个小承包商。我现在最多只有，比如，十个客户。虽然客户给的报酬很高，但目前也就十个。所以，是的，50 美元是我的最高限额，再多可能就会影响我的收益或产生其他影响。

36:36-37:03 - Researcher: That makes total sense — fifty dollars is your ceiling because beyond that, you'd be eating into your actual revenue as a small contractor. It's a business decision, not just a preference. So if that perfect tool cost fifty dollars, that's the same as your general max. But earlier you said you're currently paying $20 for ChatGPT and $19 for Copilot — that's $39 total. If this perfect tool replaced both and cost fifty, you'd actually be paying more than you do now. Would you still do it, or would you expect it to cost less since it's consolidating? - 完全理解——50 美元是你的上限，因为超过这个数，作为小承包商你就会损失实际利润。这是一个商业决定，而不仅仅是偏好。 因此，如果那个完美的工具价格是 50 美元，这和你一般的最高限额相同。但你之前说你现在分别为 ChatGPT 支付 20 美元，为 Copilot 支付 19 美元——总共是 39 美元。如果这个完美的工具取代了这两者，价格是 50 美元，实际上你会比现在付得更多。你还会买吗？或者你会因为它整合了功能而期望它更便宜？

37:06-37:08 - Marcus: I I would still do it. - 我还是会买。

37:10-37:32 - Researcher: Cool — so you'd actually pay more for true consolidation and quality, even though it's above your current spend. That's a strong signal. Alright Marcus, we've covered a ton — your workflow, what you value, your price limits, how you decide what to pay for. Really appreciate you walking me through all of this. Anything else about AI tools and paying for them that you think I should know, or anything you wanted to mention that didn't come up? - 酷——所以为了真正的整合和质量，你实际上愿意支付比现在更高的费用，即便它超过了你目前的支出。这是一个强烈的信号。 好的，Marcus，我们谈了很多——你的工作流程、你所看重的价值、你的价格底线，以及你如何决定为哪些东西付费。非常感谢你带我了解这些。关于 AI 工具和付费，还有没有什么你觉得我应该知道的，或者你想提到但还没谈到的？

37:34-37:36 - Marcus: Nope. Nothing at all. - 没了。完全没有。

37:37-37:45 - Researcher: Perfect. Thanks so much for your time, Marcus — this was super helpful. Good luck with the contracting company and hitting that $3000 for next semester! - 太棒了。Marcus，非常感谢你在百忙之中抽出时间——这非常有帮助。祝你的承包公司一切顺利，祝你在下学期顺利存够那 3000 美元！

37:48-37:50 - Marcus: Thank you. You too. Have a good day. - 谢谢，你也是。祝你今天过得愉快。

37:52-37:53 - Researcher: You too, take care! - 你也一样，再见！