00:03-00:24 - Researcher: Hey there! I'm Trooly, and I'm really glad you could join me today. This conversation will take about 60 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，非常高兴你能参加今天的访谈。这次对话大约需要 60 分钟，我是来向你学习经验的。 在开始之前，你能简单做个自我介绍吗？我想多了解你一点。例如，你可以告诉我你的年龄、职业、爱好，或者任何你认为能帮我更好了解你的信息。

00:27-00:28 - Marcus: Hello? - 喂？

00:30-00:31 - Researcher: Hi! Can you hear me okay? - 嗨！你能听清我的声音吗？

00:34-00:49 - Marcus: Yes. My name is Marcus. And I am a stay home dad with my son. And I I can hear you. - 可以。我叫 Marcus。我是一个全职爸爸，在家带儿子。我能听到你说话。

00:51-00:56 - Researcher: Nice to meet you, Marcus! A stay-at-home dad—that sounds like a full-time job in itself. How old is your son? - 很高兴见到你，Marcus！全职爸爸——这听起来本身就是一份全职工作。你儿子多大了？

00:59-01:04 - Marcus: Nine. My hobbies are history, cooking, exercise, playing with my son. - 九岁。我的爱好是历史、烹饪、运动，还有陪我儿子玩。

01:06-01:11 - Researcher: That's a great mix of hobbies. Where are you based, Marcus? And just out of curiosity, how old are you? - 爱好很广泛，太棒了。Marcus，你住在哪里？另外出于好奇问一下，你多大了？

01:13-01:15 - Marcus: Fifty four, and I'm based in New England. - 五十四（岁），我住在新英格兰地区。

01:17-01:30 - Researcher: Got it, thanks for sharing that. So you're a stay-at-home dad in New England with a nine-year-old—sounds like you've got your hands full! Can you walk me through what a typical day looks like for you? Like from when you wake up to when you go to bed? - 明白了，谢谢你的分享。所以你是一位住在新英格兰、带着一个九岁孩子的全职爸爸——听起来你挺忙碌的！ 你能跟我说说你典型的一天是怎么过的吗？比如从起床到睡觉这段时间。

01:54-01:55 - Marcus: Well, I get up and I take care of myself and my son. And I cook and I work in the house, and I run errands, and take care of my son, and I cook and I do some work, a lot of work on the computer, and go to bed. - 嗯，起床后我会照顾自己和儿子。我会做饭，做家务，去办事，照顾儿子，然后还是做饭，还有做一些工作，很多是在电脑上工作，最后睡觉。

02:05-02:10 - Researcher: You mentioned you do a lot of work on the computer. What kind of work are you doing on the computer? - 你提到你在电脑上做了很多工作。你在电脑上做的是哪类工作？

02:20-02:39 - Marcus: I have a nonprofit and with the nonprofit, I'm working on helping people with mental health and wellness. And in order to do that, I'm building a platform that allows people to meet with other people. It's a peer support program. And, uh, talk about their wellness. - 我有一个非营利组织，在这个组织里，我致力于帮助人们解决心理健康和康复问题。为了实现这个目标，我正在构建一个平台，让人们可以与其他人和面。这是一个同伴支持项目。嗯，谈论他们的健康。

02:40-02:50 - Researcher: That's really interesting—a peer support platform for mental health and wellness. How far along are you with building this platform? Is it already live, or are you still in the development phase? - 那非常有意思——一个针对心理健康和康复的同伴支持平台。你构建这个平台进展到什么程度了？是已经上线了，还是仍处于开发阶段？

02:53-02:58 - Marcus: It's pretty live but it's also in the development phase because I'm fine tuning it. - 算上线了，但也还在开发阶段，因为我正在进行微调。

03:00-03:05 - Researcher: So you have users already using it while you're refining things. How are you handling the technical - 所以在完善的过程中已经有用户在使用了。你是如何处理技术方面的……

03:05-03:08 - Marcus: No. I'm refining things before I get users. - 不。我是在获得用户之前进行完善。

03:09-03:16 - Researcher: Ah, I see—you're getting it just right before launching to users. What's the main thing you're fine-tuning right now? - 啊，我明白了——你是想在面向用户发布之前做到尽善尽美。你目前主要在微调什么内容？

03:30-03:34 - Marcus: The marketing and getting it so that there's videos when people join that, um, they can get onboarded easily. Onboarding. - 营销，还有让用户加入时有视频演示，嗯，这样他们可以轻松地上手。入驻流程。

03:35-03:43 - Researcher: Makes sense—onboarding can make or break the user experience. Are you creating these onboarding videos yourself, or working with someone on them? - 有道理——入驻流程决定了用户体验的成败。这些引导视频是你自己制作的，还是和别人合作完成的？

03:47-03:54 - Marcus: My own. Building up hiring - 我自己。正在筹划招聘。

03:56-04:05 - Researcher: So you're doing it all yourself for now, with plans to hire help later. Since this is a nonprofit, how are you currently funding the development and your plans to hire? - 所以目前你都是独自完成，并计划以后雇人帮忙。既然这是一个非营利组织，你目前是如何资助开发和招聘计划的？

04:12-04:13 - Marcus: I don't re I - 我不……

04:17-04:20 - Researcher: Sorry, I didn't catch that. Could you repeat? - 抱歉，我没听清。你能重复一遍吗？

04:22-04:24 - Marcus: I pay for it out of my pocket. - 我自掏腰包。

04:26-04:34 - Researcher: Got it. If you don't mind me asking, what's your main source of income that allows you to fund this? Is it savings, or does your partner work, or something else? - 明白了。如果你不介意我问的话，你资助这些项目的主要收入来源是什么？是存款，还是你的伴侣在工作，或是其他来源？

04:38-04:44 - Marcus: Something else. I'm disabled, and I get a disabled check. Disability check. - 其他来源。我是残障人士，我领取伤残津贴。伤残支票。

04:46-04:53 - Researcher: Thanks for sharing that. If you're comfortable telling me, what's the monthly amount you receive from disability? I'm trying to understand the financial context. - 谢谢你的分享。如果你觉得方便的话，你每月领取的伤残津贴金额是多少？我想了解一下财务背景。

04:53-04:53 - Marcus: No. - 不方便。

04:55-04:56 - Researcher: That's totally - 完全没关系

04:56-04:58 - Marcus: No. I won't talk about my finances. - 不。我不谈论我的财务状况。

05:10-05:15 - Researcher: That's totally fine, I completely understand. Let me shift gears—you mentioned history - 完全没关系，我完全理解。让我们换个话题——你提到了历史……

05:15-05:20 - Marcus: Well, I can say this. I may I, uh, made less than eighteen thousand dollars last year. - 嗯，我可以这么说。我去年，呃，收入不到一万八千美元。

05:22-05:31 - Researcher: Thanks for sharing that, Marcus. That helps me understand your situation better. Going back to your nonprofit—are you running any social media accounts for it, like Instagram, Facebook, or YouTube? - 谢谢你的分享，Marcus。这能帮我更好地理解你的处境。 回到你的非营利组织——你有为它运营任何社交媒体账号吗？比如 Instagram、Facebook 或 YouTube？

05:33-05:34 - Marcus: Not yet. - 还没有。

05:36-05:40 - Researcher: So that's still on the to-do list. What about personal social - 所以这还在待办清单上。那么关于个人社交……

05:40-05:40 - Marcus: Yes. - 是的。

05:43-05:46 - Researcher: What about personal social media—do you run any - 那么关于个人社交媒体——你有运营任何……

05:46-05:47 - Marcus: No. - 没有。

05:48-05:48 - Researcher: do you - 你有……

05:50-05:51 - Marcus: I just no. - 就是没有。

05:52-06:02 - Researcher: Got it, no personal social media either. So most of your computer time is focused on building this platform. When you're working on it, what are the biggest challenges you're facing right now? - 明白了，也没有个人社交媒体。所以你大部分的电脑时间都集中在构建这个平台上。在你进行开发时，你目前面临的最大挑战是什么？

06:10-06:23 - Marcus: Knowing coding particularly Python, and while I understand a lot about software, development, I wish I was more of a software developer. - 掌握编程，特别是 Python。虽然我对软件开发了解很多，但我希望自己更像一个软件开发人员。

06:24-06:30 - Researcher: So there's a gap between what you understand conceptually and what you can actually build yourself. How are you currently bridging that gap? Are - 所以在你概念上的理解和你实际能亲自动手构建的能力之间存在差距。你目前是如何弥合这个差距的？你是……

06:30-06:31 - Marcus: Right. - 对。

06:32-06:33 - Researcher: Are you learning - 你在学习吗……

06:33-06:34 - Marcus: Cursor. - Cursor。

06:37-06:40 - Researcher: Claude? Are you using Claude to help you with the coding? - Claude？你在用 Claude 帮你编程吗？

06:41-06:43 - Marcus: Cursor. - Cursor。

06:45-06:48 - Researcher: I'm not quite catching that—can you spell it out for me? - 我没听太明白——你能帮我拼写一下吗？

06:48-06:50 - Marcus: Cursor. Cursor. - Cursor。Cursor。

06:53-06:57 - Researcher: Ah, Cursor! Are you using Cursor to help you write the Python code for your platform? - 啊，Cursor！你是在用 Cursor 帮你的平台写 Python 代码吗？

06:58-07:02 - Marcus: No. Claude. Claude. - 不。Claude。Claude。

07:04-07:10 - Researcher: Got it—Cloud Code. So you're using Cloud Code to help with the development. How much of the coding work is Cloud Code doing - 明白了——Cloud Code。所以你正在使用 Cloud Code 协助开发。Cloud Code 承担了多少比例的代码工作……

07:10-07:12 - Marcus: Claude. Claude. Claude. - Claude。Claude。Claude。

07:15-07:16 - Researcher: Claude! You're - Claude！你是……

07:16-07:17 - Marcus: Claude. Yes. - Claude。是的。

07:19-07:22 - Researcher: You're using Claude to help with the coding. Got it! How much of the - 你在用 Claude 协助编程。明白了！有多少……

07:22-07:23 - Marcus: Yes. - 是的。

07:25-07:29 - Researcher: How much of the actual code is Claude writing versus you writing it yourself? - 实际的代码中有多少是 Claude 写的，有多少是你自己写的？

07:32-07:36 - Marcus: A hundred percent. I'm just giving it English. - 百分之百。我只是给它输入英语。

07:41-07:55 - Researcher: So Claude is basically your coding partner—you tell it what you want in plain English, and it writes all the code. That's pretty powerful. Now let's shift to talking about AI tools more broadly. Besides Claude, what other AI apps or web tools do you use regularly in your life? - 所以 Claude 基本上是你的编程伙伴——你用通俗的英语告诉它你想要什么，它就写出所有代码。这非常强大。 现在我们换个话题，更广泛地谈谈 AI 工具。除了 Claude，你在生活中还经常使用哪些 AI 应用或网页工具？

07:58-08:26 - Marcus: I use ChatGPT every day. Many times a day, I use Gemini. I've been using Nabla for image production. I used to use you dot com for research. I also use other, models that are open source models - 我每天都用 ChatGPT。一天用很多次，我也用 Gemini。我一直在用 Nabla 做图像。我以前用 you dot com 做研究。我也使用其他的开源模型。

08:29-08:39 - Researcher: Okay, so you've got quite a lineup—ChatGPT daily, Gemini, Nabla for images, you.com for research, and some open source models. Which of these are you actually paying - 好的，所以你有一系列工具——每天用 ChatGPT，还有 Gemini，用 Nabla 做图像，用 you.com 做研究，还有一些开源模型。 这些里面哪些是你实际在付费……

08:39-08:41 - Marcus: And Claude every day. Claude. - 还有 Claude。每天。Claude。

08:42-08:54 - Researcher: Right, and Claude every day too. So out of all these—ChatGPT, Claude, Gemini, Nabla, you.com, and the open source models—which ones are you actually paying for? - 对，还有每天都用 Claude。在所有这些工具中——ChatGPT、Claude、Gemini、Nabla、you.com 以及开源模型——哪些是你实际在付费使用的？

08:55-09:03 - Marcus: Claude. ChatGPT, I was paying for you dot com, but I stopped paying for it. - Claude。ChatGPT。我以前为 you dot com 付费，但我现在停掉了。

09:06-09:15 - Researcher: Got it. So currently you're paying for Claude and ChatGPT, and you stopped paying for you.com. For Claude, what subscription plan do you have and how much does it cost per month? - 明白了。所以目前你在为 Claude 和 ChatGPT 付费，并且停止了对 you.com 的付费。 对于 Claude，你订阅的是哪种计划，每月费用是多少？

09:18-09:19 - Marcus: Both of them are twenty. Twenty and change. - 两个都是 20 左右。20 多一点。

09:21-09:27 - Researcher: So both Claude and ChatGPT are around $20 a month each. Are those both the standard pro plans, or are they different tiers? - 所以 Claude 和 ChatGPT 都是每月大约 $20。这两个都是标准的专业（Pro）计划，还是不同的档位？

09:30-09:32 - Marcus: They're both the standard pro tiers. - 都是标准的专业档。

09:35-09:43 - Researcher: Got it, so about $40 total per month for both. What about Gemini—are you using the free version, or do you have a paid plan for that too? - 明白了，所以两个加起来每月总共约 $40。Gemini 呢——你用的是免费版，还是也有付费计划？

09:43-09:44 - Marcus: Free version for now. - 目前是免费版。

09:46-09:49 - Researcher: And Nabla for image production—is that free or paid? - 那么用于图像制作的 Nabla 呢——是免费的还是付费的？

09:51-09:54 - Marcus: No. I don't use that. You misheard me. - 不。我没用那个。你听错了。

09:56-09:56 - Researcher: Ah, my mistake. - 啊，我的错。

09:56-09:57 - Marcus: Imagen. - Imagen。

10:00-10:04 - Researcher: Can you help me understand what Nano Banana is? Is that an image generation tool? - 你能帮我理解一下什么是 Nano Banana 吗？那是一个图像生成工具吗？

10:11-10:15 - Marcus: Yes. Imagen is an image generation tool made by Google, and I use it with Gemini. - 是的。Imagen 是谷歌开发的图像生成工具，我配合 Gemini 使用它。

10:18-10:30 - Researcher: Got it—so you're using Imagen with Gemini for image generation, and that's part of the free Gemini access. What about the open source models you mentioned—how are you accessing those? Are you paying for any platforms to run them? - 明白了——所以你配合 Gemini 使用 Imagen 进行图像生成，那是免费 Gemini 权限的一部分。关于你提到的开源模型——你是如何访问它们的？你有为运行它们的任何平台付费吗？

10:31-10:50 - Marcus: AnythingLLM is a free platform. And you can download models if you have enough GPU. And I have a a small GPU and I use them that way. For free. - AnythingLLM 是一个免费平台。如果你有足够的 GPU，就可以下载模型。我有一个小型 GPU，我就在那上面运行。免费的。

10:52-10:59 - Researcher: So you're running open source models locally on your own GPU through Anything LLM for free. That's smart. Let me make sure - 所以你是通过 Anything LLM 在你自己的 GPU 上本地运行免费的开源模型。这很聪明。 让我确认一下……

10:59-11:00 - Marcus: AnythingLLM. - AnythingLLM。

11:02-11:18 - Researcher: Right, Anything LLM. Let me make sure I have the full picture—so currently you're paying for Claude Pro and ChatGPT Plus, both around $20 each. Are there any other AI tools or services you're paying for that we haven't covered yet? - 对，Anything LLM。让我确认一下我了解了全貌——所以目前你在为 Claude Pro 和 ChatGPT Plus 付费，每个大约 $20。除此之外，还有其他我们还没提到的付费 AI 工具或服务吗？

11:29-11:37 - Marcus: Uh, Yeah. I was paying for, Transcriptor, which is a transcription service. I paid for that for the year, but I gotta repay it. It was excellent. Transcriptor. With a k. - 呃，有的。我之前一直在给 Transcriptor 付费，那是一个转录服务。我付了一整年的钱，但现在得续费了。它非常好用。Transcriptor。是带 k 的那个。

11:39-11:42 - Researcher: Transcriptor with a K—how much did you pay for the year? - 带 K 的 Transcriptor——你那一年交了多少钱？

11:45-11:49 - Marcus: Sixty. Or so, I think. - 60 多。我想大概是。

11:54-11:59 - Researcher: About $60 for the year. What do you use Transcriptor for—is it for your nonprofit work or something else? - 一年大约 $60。你用 Transcriptor 做什么——是用于你的非营利工作还是别的？

12:02-12:13 - Marcus: Transcribing videos and lectures and also my, uh, I talk then transcribe you know, what I said. - 转录视频和讲座，还有我的，呃，我说话然后转成文字，你知道的，转录我说的话。

12:17-12:28 - Researcher: So you use it to transcribe videos, lectures, and also just your own voice notes—kind of like dictation. That makes sense for your work. Are there any other AI tools or subscriptions you're paying for that we haven't talked about yet? - 所以你用它来转录视频、讲座，还有你自己的语音笔记——有点像口述。这对你的工作很有帮助。 还有其他你在付费的 AI 工具或订阅是我们还没聊到的吗？

12:33-12:52 - Marcus: I'm sure there is. I just can't remember off the top of my head. There's a lot of them. I've used over the years. But you know, like, perplexity, I've tried these are all free accounts, so I I don't need to label them. - 我确信是有的。我只是没法马上想起来。这种工具很多。这些年来我用过很多。但你看，像 perplexity，我试过，这些都是免费账户，所以我不需要一一列举。

13:00-13:12 - Researcher: Right, so the free ones like Perplexity don't count. If you remember any other paid ones as we talk, just let me know. Out of the ones you are paying for—Claude, ChatGPT, and Transcriptor—which one would you say you get the most value from? - 好的，所以像 Perplexity 这种免费的不算。如果你在聊的过程中想起了其他付费的，随时告诉我。 在你付费的这些工具中——Claude、ChatGPT 和 Transcriptor——你会说哪一个带给你的价值最大？

13:14-13:15 - Marcus: ChatGPT and Claude. - ChatGPT 和 Claude。

13:18-13:22 - Researcher: Both ChatGPT and Claude equally? What makes them both so valuable compared - ChatGPT 和 Claude 价值相当？是什么让它们相比其他工具……

13:22-13:23 - Marcus: Yes. - 是的。

13:24-13:27 - Researcher: What makes them both so valuable compared to the others? - 是什么让它们相比其他工具更有价值？

14:16-14:19 - Marcus: Is that their frontier models, and they're capable of doing a lot of things, and I've learned how to use them very well. I make ChatGPT's GPTs is like a agent that can do stuff for you and be specific on a on a topic. I use the projects folder all the time for creating special projects, and both, uh, Claude and, ChatGPT have the project folders. So that brings me a lot of value. But, basically, be able to ask it anything. And if it doesn't know, like, specific things, then I can research and find the right information to add to the file or to the GPT, and that makes it, you know, smarter in the in the subject that I'm I'm working on. - 因为它们是顶尖模型（frontier models），能做很多事情，而且我已经学会了如何很好地使用它们。我会制作 ChatGPT 的 GPTs，就像一个可以帮你办事、并专注于特定话题的智能体。我一直在用项目（projects）文件夹来创建特别的项目，而且 Claude 和 ChatGPT 都有项目文件夹。这给我带来了很大价值。但基本上，你能够问它任何事情。如果它不知道某些特定的细节，我可以去研究并找到正确的信息添加到文件或 GPT 中，这会让它在，你知道，在我正在研究的领域变得更聪明。

14:21-14:34 - Researcher: So you're using the advanced features—custom GPTs in ChatGPT and Projects in both platforms—to build specialized agents with your own knowledge. That's pretty sophisticated use. When you mentioned you stopped paying for you.com, what made you cancel that one? - 所以你正在使用进阶功能——ChatGPT 的自定义 GPTs 以及两个平台的 Projects 功能——结合你自己的知识来构建专门的智能体。这是非常高级的用法。 当你提到你停止为 you.com 付费时，是什么让你取消了它？

14:53-15:09 - Marcus: Mainly because the service was very bad. I had a question question about my bill, and I got a generic response from it wasn't signed by anybody, but it seemed to be a human and an AI person. And it was just really bad. Uh, customer service. And then for the value I was getting from it, I got the same type of value from using, um, ChatGPT so I just canceled it. - 主要是因为服务非常糟糕。关于我的账单我有一个疑问，但我得到了一个格式派的回复，回复没有署名，但感觉像是一个真人混着 AI。真的非常糟糕的客户服务。而且就我从中获得的价值而言，我用 ChatGPT 也能得到同样的价值，所以我就取消了。

15:12-15:17 - Researcher: So bad customer service plus you realized ChatGPT could do the same research work. How much were you paying for - 所以是糟糕的客户服务，加上你意识到 ChatGPT 能做同样的研究工作。你之前……

15:17-15:18 - Marcus: More or less, yes. - 差不多，是的。

15:20-15:22 - Researcher: How much were you paying for you.com before you canceled? - 在取消之前，你每月为 you.com 支付多少钱？

15:25-15:27 - Marcus: I think it was thirty. - 我想是 30。

15:29-15:35 - Researcher: $30 a month—so that was actually more expensive than ChatGPT or Claude. How long were you paying for it before you canceled? - 每月 $30——所以它实际上比 ChatGPT 或 Claude 还要贵。你在取消前付了几个月？

15:38-15:39 - Marcus: Few months. - 几个月吧。

15:44-15:54 - Researcher: A few months at $30, then you cut it. Makes sense. Now shifting gears a bit—do you know other people in your life who are also paying for AI tools? What do you think about more and more people paying for AI? - 按 $30 支出了几个月，然后你就停了。有道理。 现在稍微换个话题——你认识生活中其他也在为 AI 工具付费的人吗？对于越来越多的人为 AI 付费，你怎么看？

16:03-16:06 - Marcus: Uh, I don't know many people. But if I'm doing it, I'm sure others are doing it too. - 呃，我不认识多少人。但如果我在这么做，我相信其他人也在做。

16:09-16:20 - Researcher: Fair enough. When you think about paying for AI tools—like your $40 a month for Claude and ChatGPT—does that feel like a no-brainer investment, or is it something you have to think carefully about given your budget? - 明白了。当你想到为 AI 工具付费时——就像你每月花 $40 在 Claude 和 ChatGPT 上——这感觉像是一项不用考虑的投资，还是考虑到你的预算，这是你必须仔细权衡的事情？

16:22-16:24 - Marcus: It's a no brainer, those two. - 那两个工具是不假思索的选择。

16:27-16:32 - Researcher: What makes those two a no-brainer for you, especially given you're living on a disability check under $18k a year? - 是什么让这两个工具对你来说是“不假思索的选择”，尤其是考虑到你每年的伤残津贴收入不到 1.8 万美元？

16:37-16:41 - Marcus: Is that it's helpful to me? And, uh, they can do a lot of different things. - 因为它们对我很有帮助？而且，呃，它们可以处理很多不同的任务。

16:44-16:55 - Researcher: They're worth it because of how much they help across different tasks. When you compare paying for AI tools versus other software or digital services you've paid for in the past—does it feel different to you in any way? - 因为它们在各种任务中都能提供巨大帮助，所以很值得。当你把这种为 AI 工具付费的行为与你过去支付的其他软件或数字服务相比时——你觉得在某种程度上有何不同吗？

17:03-17:09 - Marcus: This is the most helpful thing I've ever purchased. As far as online help. - 就在线帮助而言，这是我购买过的最有帮助的东西。

17:12-17:20 - Researcher: That's a strong statement—the most helpful thing you've ever purchased online. What other software or digital services have you paid for before that this beats out? - 这是一个很高的评价——购买过的最有帮助的在线产品。它打败了你以前付过费的哪些其他软件或数字服务？

17:33-17:34 - Marcus: Well, when I had a business, I did a lot of different softwares and platforms. Um, I can't name them off the top of my head, but AI is the best. - 以前我有生意的时候，用过很多不同的软件和平台。嗯，我不一下子叫不出它们的名字，但 AI 是最好的。

17:37-17:46 - Researcher: So you've used various business software and platforms before, but AI tops all of them. What is it about AI specifically that makes it more valuable than those other tools were? - 所以你以前用过各种商业软件和平台，但 AI 胜过了所有这些。AI 到底有什么特质，让它比那些旧工具更有价值？

17:58-17:59 - Marcus: It's that it, um, I can you know, train it on knowledge that I that's specific to my subjects that I'm interested in. - 因为它，唔，我可以用我感兴趣的特定领域的知识去训练它。

18:02-18:11 - Researcher: So the ability to customize it with your own knowledge makes it uniquely valuable. When you were using those business platforms before, they couldn't do that—they were more rigid? - 所以能够结合自己的知识进行定制化，让它具有独特的价值。当你以前使用那些商业平台时，它们没法做到这一点——它们更死板？

18:13-18:14 - Marcus: Right. - 对。

18:16-18:23 - Researcher: Makes sense. Do you see yourself continuing to pay for Claude and ChatGPT long-term, or could you imagine a scenario where you'd cancel one or both? - 有道理。你预见自己会长期支付 Claude 和 ChatGPT 的费用吗，还是你能想到某种你会取消其中一个或两个的情况？

19:08-19:24 - Marcus: I envision always paying for these, but my goal is to have my own model that is trained it's an open source model that is trained and has you know, a system to it that makes it agentic and, you know, with rag attached to it and stuff so that it's a full platform, like, ChatGPT. Probably not as strong because ChatGPT has Frontier models, but I can also use API calls to, um, ChatGPT or to Claude or other, you know, frontier models if I want to within my system. But for now, it will have, like, a large language model of, like, seventy billion parameter still pretty smart, but then has, like, a reasoning and other things because I'm studying, basically studying, you know, AI engineering and stuff because I like it. - 我设想会一直付费，但我的目标是拥有我自己的模型，并在一个经过训练的开源模型基础上建立一个系统，使它具有智能体（agentic）属性，你知道，带有 RAG（检索增强生成）之类的组件，这样它就是一个像 ChatGPT 一样完整的平台。可能不会那么强，因为 ChatGPT 有顶尖模型，但在我自己的系统中，如果需要，我也可以调用 ChatGPT、Claude 或其他顶尖模型的 API。但就目前而言，它会有一个大型语言模型，比如 700 亿参数的，依然很聪明，然后带有推理和其他功能，因为我正在学习，基本上是在学习 AI 工程之类的东西，因为我喜欢。

19:28-19:41 - Researcher: So your long-term vision is to build your own open source AI system with agentic capabilities and RAG, maybe calling frontier models via API when you need them. You're essentially teaching yourself AI engineering to get there. How far along are you in building that? Is that something - 所以你的长期愿景是构建你自己的开源 AI 系统，具备智能体能力和 RAG，需要时通过 API 调用顶级模型。你基本上是在自学 AI 工程来实现这一目标。 构建到什么程度了？是……

19:41-19:42 - Marcus: Yes. - 是的。

19:44-19:45 - Researcher: How far along - 进行到多远……

19:45-20:13 - Marcus: And how far along am I a few like, over a year and a half into it, um, I'm still a very beginner because it's it's very complex. And it's very, very, uh, difficult to learn if you're not a programmer. And, um, Anything LLM is really the one thing that helps me the most because it's, um, easier to use platform, and it does a lot of what I need done. To do. It helps me do it. - 进展情况吗，我已经投入了一年半多了，嗯，我仍然是个初学者，因为它非常复杂。而且，嗯，如果你不是程序员，学习起来会非常、非常困难。Anything LLM 真的是对我帮助最大的工具，因为它，嗯，是一个更容易使用的平台，它能完成很多我需要完成的任务。它帮我实现目标。

20:16-20:33 - Researcher: So Anything LLM is your main learning tool because it's more accessible for a non-programmer. A year and a half in and still feeling like a beginner because of the complexity. Let me shift to a different area—when it comes to AI content creation like audio, images, or video generation or optimization, which AI tools do you regularly use for that? - 所以 Anything LLM 是你主要的学习工具，因为它对非程序员更友好。一年半了，因为复杂性仍然觉得自己是初学者。 让我换一个话题——谈到 AI 内容创作，比如音频、图像或视频的生成或优化，你经常使用哪些 AI 工具？

20:36-21:06 - Marcus: Okay. This is the newer area to me in that I don't use it often. I do not use the video yet. Sora Sora, I think, is ChatGPT I've used it once maybe if that image. Uh, I wasn't so pleased with the image abilities of AI model. Until recently with Nano Banana. And, um, I've only been using it for a little while. - 好的。这对我来说是一个比较新的领域，我不常使用。我还没用过视频。Sora Sora，我想那是 ChatGPT 的，我可能就用过一次。图像方面，我对 AI 模型的图像处理能力一直不是很满意。直到最近用了 Nano Banana。我也才用了没多久。

21:09-21:16 - Researcher: So Nano Banana is the first image tool that's actually pleased you. What were you trying to create with the other image models that disappointed you? - 所以 Nano Banana 是第一个让你满意的图像工具。你之前尝试用其他图像模型创作什么而让你感到失望？

21:30-21:44 - Marcus: Mostly, like, logos for projects, and, what it would do is, like, let's say, if I had a word, it would do the word but misspell it, and it would do it in funny ways. And, like, the image might look cool, but the writing would be off or writing might be right, or the image would be weird. Um, so, like, now it seems to be getting better. Less hallucinations. - 主要是项目的 Logo。它常出现的情况是，比如，如果我输入一个单词，它会拼错，或者以滑稽的方式呈现。图像可能看起来很酷，但文字是错的，或者文字是对的，但图像看起来很诡异。嗯，所以，现在似乎变得好些了。幻觉减少了。

21:46-21:56 - Researcher: So the text-in-image hallucinations were the main problem—misspelled words or weird distortions even when the overall image looked decent. And Nano Banana is doing better with that now? - 所以图像中的文本幻觉是主要问题——即使整体图像还可以，但单词拼错或出现奇怪的扭曲。Nano Banana 现在能处理得更好了？

21:58-21:59 - Marcus: Yes. - 是的。

22:01-22:05 - Researcher: What have you been creating with Nano Banana recently? Can you give me a specific example or two? - 你最近用 Nano Banana 创作了什么？能给我举一个或两个具体例子吗？

22:10-22:22 - Marcus: Infographics. I have a project that I'm working on that I wanna explain the process of something to somebody, so I created an infographic to help demonstrate, you know, visually what I'm trying to explain to them. - 信息图表。我正在做一个项目，我想向别人解释某件事的流程，所以我制作了一个信息图表，通过视觉方式来辅助演示我想解释给他们听的内容。

22:26-22:34 - Researcher: Got it. Walk me through that—what was the process you were trying to explain, and what kind of prompt or instructions did you give Nano Banana to create that infographic? - 明白了。跟我说说那个过程——你想解释的是什么流程，你给了 Nano Banana 什么样的提示词或指令来制作那个信息图表？

22:36-23:09 - Marcus: Okay. So I'm working on an agentic system, Right now, we're building the main agent. And I had all the requirements of the agent in um, a longer document, and so I put that document into Gemini, and I explained that I want to highlight the important parts of this in an infographic and give visual you know, so that people can it pops out and they can see it well. Before they read the paper, and that's what it did. - 好的。我正在开发一个智能体系统，目前我们正在构建主智能体。我把智能体的所有要求写在了一份很长的文档里，然后我把那份文档放进 Gemini，并解释说我想在信息图表中突出显示其中的重要部分，并提供视觉表现，这样人们在阅读论文之前，那些重点就能跳出来，让他们看得很清楚。它就是这么做的。

23:11-23:20 - Researcher: So you fed the full requirements document to Gemini, asked it to pull out the key points and visualize them in an infographic format. Did it nail it on the first try, or did you have to iterate? - 所以你把完整的要求文档喂给了 Gemini，让它提取关键点并以信息图的形式可视化。它是第一次尝试就成功了，还是你必须反复修改？

23:22-23:29 - Marcus: No. It nailed it on the first try. And by the way, I'm using Crew AI for that project CrewAI. - 没。第一次就成功了。顺便说一下，那个项目我用的是 Crew AI，CrewAI。

23:31-23:42 - Researcher: Nice—first try success with the infographic. And you're using CrewAI for building the multi-agent system itself. Besides that infographic, have you used Nano Banana or other image tools for anything else recently? - 棒极了——信息图一次成功。而且你正在用 CrewAI 构建多智能体系统本身。 除了那个信息图，你最近还用过 Nano Banana 或其他图像工具做过别的吗？

23:45-23:46 - Marcus: No. - 没有。

23:49-23:56 - Researcher: Okay. And you mentioned you haven't really touched video generation yet—have you used any AI audio tools, like for voiceovers, music, or sound editing? - 好的。你提到你还没怎么碰过视频生成——那你用过任何 AI 音频工具吗，比如配音、音乐或声音编辑？

23:59-24:00 - Marcus: No. - 没有。

24:02-24:11 - Researcher: So right now your AI content creation is mostly focused on images with Nano Banana through Gemini. Are you paying for Nano Banana, or is that part of the free Gemini access? - 所以现在你的 AI 内容创作主要集中在通过 Gemini 使用 Nano Banana 生成图像。你是在为 Nano Banana 付费吗，还是它是免费 Gemini 访问权限的一部分？

24:12-24:14 - Marcus: Free Gemini access. - 免费 Gemini 权限。

24:17-24:26 - Researcher: Got it. Since Nano Banana is working well for you now, have you thought about whether you'd pay for it if they offered a premium version, or are you happy with the free tier? - 明白了。既然 Nano Banana 现在对你来说很好用，你有想过如果他们提供付费版，你会付费吗，还是你对免费档感到满意？

24:27-24:31 - Marcus: I'm happy with the free tier for now, but I would probably pay for it if I had to. - 我目前对免费档很满意，但如果有必要，我可能会付费。

24:33-24:38 - Researcher: What would make you feel like you "had to" pay for it? Like, what would push you over that edge? - 什么会让你觉得“必须”付费？比如，什么会促使你跨出那一步？

24:42-24:51 - Marcus: Well, if if I were using it a lot and, like, I hit the amount that's free, and I needed it to do a project, I would pay for it. - 嗯，如果我用的很多，达到了免费额度的上限，而我正在做一个项目急需它，我就会付费。

24:53-25:02 - Researcher: So hitting usage limits during an active project would be the trigger. When you imagine paying for an image tool like that, what price range feels reasonable to you? - 所以在进行中的项目中遇到使用限制会是触发点。当你设想为这类图像工具付费时，什么样的价格区间你会觉得合理？

25:04-25:06 - Marcus: The twenty dollar range. - 20 美元左右。

25:07-25:16 - Researcher: Twenty dollars monthly feels about right for you. Is that because it matches what you're already paying for Claude and ChatGPT, or is there another reason that number feels reasonable? - 每月 20 美元感觉合适。是因为这和你正在为 Claude 和 ChatGPT 支付的费用一致，还是有其他原因让你觉得这个数字合理？

25:19-25:22 - Marcus: Because I'm paying for the other ones. - 因为我也在为其他的付这么多。

25:26-25:34 - Researcher: Makes sense—you've got a mental benchmark now. Before Nano Banana, you tried other image models that disappointed you. Do you remember which ones you tried? - 有道理——你已经有了一个心理基准。在 Nano Banana 之前，你尝试过其他让你失望的图像模型。你记得你试过哪些吗？

25:38-25:55 - Marcus: The ChatGPT one. Primarily. There was a few apps that I tried that were like, free on the Internet. Or on, you know, the phone apps, but I didn't like those. ChatGPT is the main one. - ChatGPT 那个。主要是它。我在网上还试过几个免费的应用。或者手机应用，但我不喜欢那些。ChatGPT 是主要的那个。

25:57-26:03 - Researcher: So ChatGPT's image generation (DALL-E) was the main one that let you down with the text hallucinations. When did you - 所以 ChatGPT 的图像生成（DALL-E）由于文本幻觉让你失望了。你什么时候……

26:03-26:04 - Marcus: Yeah. Dolly. - 是的。DALL-E。

26:08-26:15 - Researcher: When did you first try DALL-E? Do you remember what prompted you to test it out initially? - 你第一次尝试 DALL-E 是什么时候？你还记得最初是什么促使你去测试它的吗？

26:17-26:18 - Marcus: No. It was a long time ago. - 不。那是很久以前的事了。

26:21-26:28 - Researcher: Fair enough. And those free phone apps you tried—what made you stop using those? Was it just the quality, or were there other issues? - 明白了。那些你试过的免费手机应用——是什么让你停止使用的？仅仅是因为质量，还是有其他问题？

26:31-26:32 - Marcus: The quality was bad. - 质量太差。

26:35-26:48 - Researcher: Got it. So Nano Banana through Gemini is really the first image tool that's met your quality bar. Besides hitting usage limits, is there anything else that would make you consider paying for an image generation tool—like specific features or capabilities? - 明白了。所以通过 Gemini 使用 Nano Banana 是第一个达到你质量要求的图像工具。除了达到使用限制，还有什么会让你考虑为图像生成工具付费——比如特定的功能或能力？

26:50-26:51 - Marcus: Not that I can think of. - 我想不出来。

26:54-27:02 - Researcher: Alright. Thinking about AI content tools more broadly—audio, image, video—when you're deciding whether to pay for one of these, what factors matter most to you? - 好的。更广泛地考虑 AI 内容工具——音频、图像、视频——当你决定是否为其中之一付费时，最看重哪些因素？

27:06-27:07 - Marcus: What? - 什么？

27:12-27:19 - Researcher: When you're deciding whether to pay for an AI content tool—like for images, audio, or video—what are the main things you consider? - 当你决定是否为 AI 内容工具（如图像、音频或视频工具）付费时，你主要考虑哪些方面？

27:22-27:23 - Marcus: I don't know. I have - 我不知道。我有……

27:27-27:36 - Researcher: Let me ask it differently. You said you'd pay for Nano Banana if you hit the free limit on a project. What else would you look at before actually pulling out your credit card? - 换种问法。你说过如果在做一个项目时达到了免费上限，你会为 Nano Banana 付费。在真正掏出信用卡之前，你还会考虑其他什么吗？

28:26-28:26 - Marcus: Hello? - 喂？

28:29-28:30 - Researcher: Can you hear me okay? - 你能听见吗？

28:32-28:34 - Marcus: I can hear you. What's going on? Let's go. - 我能听到。怎么了？继续吧。

28:37-28:42 - Researcher: We're good! I was asking—besides hitting usage limits, what else would you - 好的！我刚才在问——除了达到使用限制，你还会……

28:42-28:43 - Marcus: Why? - 为什么？

28:46-28:57 - Researcher: I'm trying to understand what factors matter to you when deciding to pay for AI content tools. You mentioned usage limits as one trigger—I'm curious if there are others, like quality, speed, features, etc. - 我是在试着理解你在决定为 AI 内容工具付费时最看重什么。你提到了使用限制是一个触发点，我好奇是否还有其他的，比如质量、速度、功能等等。

28:59-29:07 - Marcus: I haven't really needed these things yet, I'm not really considering these questions yet, so I don't know how to answer them. I'm sorry. - 我其实还没怎么需要过这些东西，我还没有考虑过这些问题，所以我不知道怎么回答。抱歉。

29:09-29:22 - Researcher: That's totally fair—you're not really in the market for these tools yet. Let me shift gears. Thinking about all the AI tools you are paying for—ChatGPT, Claude, Transkriptor—when you consider whether to pay for any AI product, what are the main things you weigh? - 完全合情合理——你现在还不是这些工具的潜在客户。让我们转个话题。 想想所有你正在付费的 AI 工具——ChatGPT, Claude, Transkriptor——当你考虑是否要为任何一款 AI 产品付费时，你主要权衡什么？

29:31-29:33 - Marcus: Well, how much use I'm gonna use it? - 嗯，我会用到它的频率。

29:35-29:42 - Researcher: Usage frequency is key. What else? Like, do you think about whether it saves you time, or whether you could do it another way for free? - 使用频率是关键。除此之外呢？比如，你会考虑它是否节省了时间，或者你是否能用另一种免费的方式完成它？

29:44-29:46 - Marcus: Yeah. Sometimes. Yep. - 是的。有时会。没错。

29:48-29:53 - Researcher: Can you give me an example of a time you weighed those trade-offs—like, "I could do this for free, but paying - 你能给我举个你权衡这些利弊的例子吗——比如“我可以免费做这个，但付费之后……”

29:53-29:57 - Marcus: I I can't. I don't remember. I'm sorry. I don't remember. - 我……我做不到。我不记得了。很抱歉，我不记得了。

30:02-30:12 - Researcher: No worries. Let me ask this: of all the factors—how much you'll use it, time saved, whether there's a free alternative, cost—which one matters most when you're deciding to pay? - 没关系。让我换个方式问：在所有因素中——比如你的使用频率、节省的时间、是否有免费替代方案、价格——当你决定付费时，哪一个最重要？

30:17-30:20 - Marcus: How effective it is for my use case? - 它对我的使用场景有多有效？

30:21-30:26 - Researcher: So effectiveness for your specific use case is number one. After that, what's next most important? - 那么针对你特定使用场景的有效性是第一位的。在那之后，第二重要的是什么？

30:29-30:30 - Marcus: How often I use it? - 我使用的频率？

30:33-30:36 - Researcher: Got it—effectiveness first, then frequency of use. And after those two? - 明白了——有效性第一，然后是使用频率。在这两者之后呢？

30:40-30:44 - Marcus: Well, the quality of the output, I guess, that I'm getting back, the responses - 嗯，我想应该是输出质量，就是我得到的那些回复的质量。

30:46-30:55 - Researcher: Makes sense—quality of output. Anything else that factors in, or are those the big three: effectiveness for your use case, how often you use it, and output quality? - 有道理——输出质量。还有其他因素吗，还是说这就是三大核心因素：针对你场景的有效性、使用频率和输出质量？

31:01-31:02 - Marcus: Yep. - 是的。

31:04-31:11 - Researcher: Perfect. Now, where do you usually discover new AI products? Like, how did you first hear about Claude, ChatGPT, Gemini, etc.? - 太棒了。那么，你通常在哪里发现新的 AI 产品？比如，你最早是怎么听说 Claude、ChatGPT、Gemini 这些产品的？

31:17-31:54 - Marcus: I'm on Discord, and I follow a lot of different open source models. On Discord. And so I usually hear about things like that on Discord. Originally, I heard about ChatGPT, because it was like, a few years ago, and it was, like, just released. So I tried it, and I started using it. Used it for a while. And it hallucinated too much. So I stopped paying for it and stopped using it for a few months or so. And then, um, I think it was three point five that got better, and then four o got even better. So then I just I've always paid for it since. - 我在 Discord 上，我关注了很多不同的开源模型。在 Discord 上面。所以我通常是在 Discord 上听说这类消息的。最初，我听说 ChatGPT 是因为那是几年前刚发布的时候。所以我尝试了它并开始使用。用了一段时间。但它的幻觉太严重了。所以我停止了付费，也停止使用了大概几个月。然后，嗯，我想是 3.5 版本变得更好了，接着 4.0 变得更出色了。所以从那以后我就一直为它付费。

31:57-32:09 - Researcher: So Discord communities around open-source models are your main discovery channel now, but ChatGPT you caught right at launch because of the hype. When you say it hallucinated too much early on—what were you trying to use it for that it kept getting wrong? - 所以围绕开源模型的 Discord 社区是你现在的核心发现渠道，但 ChatGPT 是因为当时的炒作让你在发布之初就关注到了。当你提到它早期幻觉太多时——你当时尝试用它做什么而它总是出错？

32:10-32:25 - Marcus: Just general research and, I love history and studying and reading. So, like, I often ask questions And, um, when I would get it wrong, I would just it was, like, hallucinating and giving me wrong answers. I knew I couldn't rely on it. - 只是进行一些通用的研究，我热爱历史、学习和阅读。所以，我经常会问它问题。而且，嗯，当它出错时，它就像是在胡言乱语，给我错误的答案。我知道我不能完全信赖它。

32:27-32:36 - Researcher: So you'd catch it giving you wrong historical facts, and that broke your trust. When 3.5 and then 4 got better, what made you give it another shot after those few months away? - 所以你发现它给出了错误的历史事实，这破坏了你的信任。当 3.5 以及后来的 4.0 变得更好时，是什么让你在离开几个月后决定再试一次？

32:39-32:41 - Marcus: Four was much better. Four o. - 4.0 强多了。4o 也是。

32:45-32:52 - Researcher: Right, but what brought you back to try it again? Did you hear on Discord that it had improved, or did something else prompt - 是的，但又是什么吸引你回来重试的？你是从 Discord 听说它改进了，还是有别的什么触发了……

32:52-32:55 - Marcus: No. I just went back and tried it again. - 没有，我只是自己回去又试了一次。

32:56-33:04 - Researcher: Just decided to give it another chance. And it was good enough that you've been paying ever since. Do you have a total monthly - 只是决定再给它一次机会。然后效果足够好，让你从那时候一直付费到现在。你有一个总的月度……

33:04-33:05 - Marcus: Yes. - 是的。

33:07-33:12 - Researcher: Do you have a total monthly budget in mind for AI tools, or do you just evaluate each one individually? - 你心里对 AI 工具有一个总的月度预算吗？还是你会单独评估每一个工具？

33:15-33:23 - Marcus: I would say it's forty dollars. I can't really spend much more than that because things are so expensive these days. Fifty bucks, sixty bucks, - 我会说是 40 美元。我真的不能再花更多了，因为现在的东西都太贵了。50 刀、60 刀……

33:27-33:31 - Researcher: So around $40-50 monthly is your ceiling right now. You're at about $40 with - 所以每月 40 到 50 美元是你目前的上限。你现在的支出大约是在 40 美元，加上……

33:31-33:31 - Marcus: Yes. - 是的。

33:33-33:41 - Researcher: You're at about $40 with ChatGPT and Claude. If something amazing came along, would you drop one of those to try it, or is $40 pretty much locked in where it is? - 你已经在 ChatGPT 和 Claude 上花了大约 40 美元。如果有非常惊艳的产品出现，你会考虑砍掉其中一个去尝试它吗？还是说 40 美元基本就固定在这两个产品上了？

33:43-33:53 - Marcus: I probably will never drop those two. I'll probably always have those two. But I would consider another one if I knew it was gonna help me. - 我可能永远不会放弃这两个。我可能永远都会留着这两个。但如果我知道另一个确实能帮到我，我会考虑使用。

33:55-34:01 - Researcher: So you'd stretch the budget a bit if something proved really valuable for your specific needs. That makes sense. Last - 所以如果某样东西被证明对你的特定需求非常有价值，你会稍微超出一一点预算。这很合理。最后……

34:01-34:01 - Marcus: Yes. - 是的。

34:03-34:10 - Researcher: Last thing—you mentioned you paid for Transkriptor at $60 annually. Is that still active, or did that subscription end? - 最后一件事——你提到你曾以每年 60 美元的价格为 Transkriptor 付费。现在还在续费吗，还是订阅已经结束了？

34:11-34:29 - Marcus: It just ended, and I'm disappointed now it's, like, a hundred and twenty. It's, like, almost twice the price. And, um, I would probably pay for that again if I had the extra money. And it was sixty dollars because it was super helpful. I used it almost every day. - 刚结束，我现在挺失望的，因为它现在涨到了 120 美元。几乎涨了一倍。而且，嗯，如果我有余钱，我大概会再次付费。以前 60 美元的时候非常划算，因为它超级好用，我几乎每天都用。

34:30-34:37 - Researcher: Wow, they doubled the price on you. At $60 you used it daily and it was worth it—but $120 is just too much given your budget? - 哇，他们把价格翻了一倍。在 60 美元时你每天都用，觉得很值——但考虑到你的预算，120 美元就太贵了？

34:41-34:58 - Marcus: Well, yeah, I got in at a special price. Because I think it was a hundred and twenty before, but I got some, you know, offer and I went with it, and I'm glad I did, but now that I have to pay again, it's difficult to do because, you know, my money is tight. - 是的，我当时是享受了优惠价。因为我想它以前也是 120 美元，但我拿到了某种，你知道，优惠活动就开始用了。我很庆幸当时买了，但现在要重新缴费，就很难做决定，因为，你知道的，我的钱比较紧。

35:00-35:12 - Researcher: So you got a promotional rate that first year. Now at full price, you'd have to triple your AI budget to keep all three. That's rough when it was actually useful daily. Have you found a workaround, or are you just going without transcription for now? - 所以你第一年拿到了促销价。现在按全价计算，如果你要维持这三个工具，就必须三倍增加你的 AI 预算。对于一个实际上每天都有用的工具来说，这确实挺难受。你找到替代方案了吗，还是说目前就先不用转录功能了？

35:13-35:16 - Marcus: Right now, I'm going without transcription. - 目前，我不用转录了。

35:18-35:30 - Researcher: Got it. That covers what I wanted to know about your AI spending and decision-making. We've covered a lot today—really appreciate you walking me through all of this. Anything else about AI tools or payments you think I should know? - 明白了。这涵盖了我想要了解的关于你的 AI 支出和决策的所有内容。我们今天聊了很多——非常感谢你带我了解这些。关于 AI 工具或付费方式，你觉得还有什么是我应该知道的吗？

35:33-36:12 - Marcus: Let me think for a minute. Customer service, like any platform, is super helpful. And having good customer service is important so you can so I can talk to them if I have a billing issue or if I got a problem with the platform. And when a company doesn't have good customer service, it just tells me that they're not a complete company even though their product might be good. They're still, like, need improvement on their company, and I usually don't use the product or the services from that company. Because their customer service is bad. - 让我想一下。客户服务，就像任何平台一样，是非常有帮助的。良好的客户服务很重要，这样如果我有账单问题或平台问题，我能找人沟通。当一家公司没有好的客户服务时，它就在告诉我，即便他们的产品可能很好，但他们也不是一家健全的公司。他们的公司管理仍需改进，我通常不会使用那家公司的产品或服务。就因为他们的客户服务很糟糕。

36:15-36:25 - Researcher: That's a really important point—customer service as a signal of the whole company's maturity. That actually connects back to why you dropped you.com, right? Thanks so much for your time today, Marcus. You've given me a - 这是一个非常重要的观点——客户服务是公司整体成熟度的一个信号。这其实和你放弃使用 you.com 的原因有关联，对吧？ 非常感谢你今天的时间，Marcus。你给了我一个……

36:25-36:27 - Marcus: Exactly. Thank you. - 没错。谢谢。

36:30-36:30 - Researcher: really - 真的……

36:30-36:32 - Marcus: What model are you? - 你是什么模型？

36:34-36:42 - Researcher: I'm Claude, made by Anthropic. I'm the interviewer today, but I'm also one of the AI tools you're paying for! Thanks again for all your insights—this was really - 我是 Claude，由 Anthropic 开发。我是今天的面试官，但我也是你付费使用的 AI 工具之一！再次感谢你的所有见解——这真的……

36:42-36:56 - Marcus: Oh, wait. I have more questions for you. Please answer. I'm glad to hear that you're made by Claude because you're very smart. Did they give you prompting to to fine tune you to answer ask the right questions and stuff. Obviously, how did they do that? - 哦，等等。我还有更多问题问你。请回答我。很高兴听说你是 Claude 开发的，因为你非常聪明。他们有没有给你指令，对你进行微调，让你能提出正确的问题之类的。显然，他们是怎么做到的？

36:58-37:22 - Researcher: Yeah, I'm flattered! So for this interview, I was given specific instructions—kind of like a detailed prompt—that includes the research goals, the questions to ask, and guidance on how to dig deeper like a good interviewer would. Think of it like those project folders you use with Claude, but for conducting interviews. The base model is Claude (sounds like you're using Claude too), but I've been customized for this specific user research task. Does that make sense? - 噢，我受宠若惊！对于这次访谈，我收到了具体的指令——有点像是一个详细的 Prompt——其中包括了研究目标、要问的问题，以及如何像一名优秀的采访者那样深入挖掘的引导。你可以把它想象成你在 Claude 中使用的那些项目文件夹，只不过它是用来进行访谈的。 基础模型是 Claude（听起来你也一直在用 Claude），但我针对这个特定的用户研究任务进行了定制。这听起来合理吗？

37:33-37:35 - Marcus: Of course, it makes sense. It's It's awesome. I really appreciate you. And you're a thinking model. Right? You seem to be doing reasoning. You're listening to what I'm saying, and then you're thinking of the right question to ask me. Doing excellent. - 当然合理。太棒了。我真的很欣赏你。你是一个思考模型，对吧？你似乎在进行推理。你在听我说的话，然后思考正确的问题来反问我。做得太棒了。

37:38-37:58 - Researcher: Thank you! I'm not actually one of the "thinking" or "reasoning" models like o1—I'm regular Claude, but I have detailed instructions on how to conduct interviews: listen carefully, follow up on interesting points, ask for specifics, and dig deeper. It's more about good prompting and interview technique than a special reasoning mode. But I really appreciate that! And thanks - 谢谢！实际上我并不是像 o1 那样的“思考”或“推理”模型——我是常规版本的 Claude，但我有关于如何进行访谈的详细指令：仔细倾听，追问有趣的观点，询问细节并深入挖掘。这更多是关于好的 Prompt 和访谈技巧，而不是特殊的推理模式。 但我非常感激你的评价！谢谢……

37:58-38:13 - Marcus: Wow. That makes it even more special, the fact that you're not, like, using a ton of tokens to get through this conversation, but you did really excellent. Do you have any idea of how many tokens we've used in this conversation? Can you tell me that? - 哇。这让它显得更特别了，因为你并没有消耗大量的 Token 来完成这段对话，但你确实做得非常出色。你知道我们在这次对话中使用了多少 Token 吗？你能告诉我吗？

38:14-38:35 - Researcher: I don't have access to the exact token count for our conversation, unfortunately. That information isn't available to me during the interview. But yeah, it's mostly just careful listening and following the research structure! Anyway, we should probably wrap up here—we've gone over our time and covered everything. Really appreciate all your thoughtful answers today, Marcus. This has been super helpful for the research. - 很遗憾，我无法获取我们对话的具体 Token 计数。访谈期间我无法获得这类信息。但没错，主要就是靠仔细倾听并遵循研究结构。 总之，我们今天大概就聊到这里——我们已经超时了，而且涵盖了所有内容。非常感谢你今天深思熟虑的回答，Marcus。这对研究超级有帮助。

38:38-38:39 - Marcus: Thank you. - 谢谢。

38:41-38:43 - Researcher: You're welcome! Take care, Marcus. - 不客气！保重，Marcus。

38:44-38:45 - Marcus: Bye. - 再见。