00:02-00:25 - Researcher: Hey there! I'm Trooly, and I'm so glad you could join me today. This interview will take about 60 minutes, and I'm really excited to learn from your experiences. Before we dive into the main topic, 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，很高兴你今天能加入。 这次采访大约需要 60 分钟，我非常期待能从你的经验中学习。在深入探讨主题之前，你能先做一个简短的自我介绍吗？我想多了解你一点。例如，你可以告诉我你的年龄、职业和爱好，或者任何你认为能帮助我更好了解你的信息。

00:27-00:56 - Marcus: Fifty five. I am a web developer and digital marketer. My hobbies are I love music technology. I I'm an avid synthesist. And home recording artist as well as being a web developer I've got great interest in web audio That's about it. - 55岁。我是一名网页开发人员和数字营销人员。我的爱好是，我热爱音乐技术。我是一个狂热的合成器爱好者，也是一名家庭录音艺术家。除了做网页开发，我对网页音频非常感兴趣。大概就是这些。

00:58-01:13 - Researcher: That's fascinating! A web developer, digital marketer, AND into music synthesis and home recording - that's quite a unique combination. I'm curious about your web development work. What are your main responsibilities day-to-day? Could you walk me through what a typical workday looks like for you? - 太迷人了！网页开发人员、数字营销人员，还涉足音乐合成和家庭录音——这确实是一个非常独特的组合。 我对你的网页开发工作很好奇。你日常的主要职责是什么？你能带我了解一下你典型的工作日是怎样的吗？

01:17-02:09 - Marcus: Uh, typical workday would be taking a look at client needs Sometimes clients will reach out with issues with their WordPress sites if I'm not doing an active build. Then simply making sure that all the client sites are looking good, I'm looking, uh, to take proactive measures. You know, against any kind of problems. If I am doing an active build, then that's usually, uh, very scheduled out process. Yeah. Then I'm gonna have a my day will be very, very scheduled. I've got each component of a build kinda mapped out. So yeah. - 呃，典型的工作日会先查看客户需求。如果我没有在进行中的开发项目，客户有时会带着他们 WordPress 网站的问题来找我。然后就是简单地确保所有客户网站运行良好，我会采取一些主动措施，你知道的，预防任何可能出现的问题。如果我正在进行一个开发项目，那通常是一个非常有计划的过程。是的。那么我的一天会安排得非常满。我会把开发的每个组件都规划好。所以，是的。

02:12-02:26 - Researcher: Got it, so it sounds like you toggle between maintenance mode and build mode depending on what's happening with clients. When you're in that active build phase and everything's mapped out - what does "each component" typically include? Like, what are the main pieces you're working through? - 明白了，听起来你会根据客户的情况在“维护模式”和“开发模式”之间切换。 当你处于那种一切都规划好的活动开发阶段时——“每个组件”通常包括什么？比如，你正在处理的主要部分有哪些？

02:38-02:53 - Marcus: Typical web build is gonna go through client analysis. We're, um, analyzing how they get customers and, you know, if they have a brick and mortar store I'm gonna approach it a little bit differently than if they're just ecommerce. Yeah. - 典型的网页开发会经历客户分析。我们会分析他们是如何获取客户的，你知道，如果他们有实体店，我的处理方式会和纯电商有点不同。是的。

02:55-03:08 - Researcher: That makes sense - brick and mortar versus pure ecommerce would need totally different strategies. When you're doing that client analysis piece, what are you actually trying to figure out about how they get customers? What questions are you asking them? - 有道理——实体店与纯电商确实需要完全不同的策略。 当你进行客户分析时，你实际上是想弄清楚他们获取客户的哪些方面？你会问他们什么问题？

03:18-03:28 - Marcus: Yeah. I'm asking, you know, if they are getting just their their customers usually just through street traffic. If they actually have any kind of email list, if they run promotions, that sort of thing. - 是的。我会问，比如，他们是否通常只是通过街面人流量获取客户。他们是否有任何形式的电子邮件列表，是否进行过促销活动，诸如此类。

03:31-03:43 - Researcher: So you're mapping out their current customer acquisition channels. When you find out they're mostly relying on street traffic and maybe don't have much else going on - what do you typically recommend or build for them? How does that shape what you create? - 所以你是在梳理他们目前的客户获取渠道。 当你发现他们主要依靠街面流量，而其他方面可能做得不多时——你通常会给他们推荐或构建什么？这会如何影响你的创作内容？

04:22-04:34 - Marcus: If it's basically just street traffic, I'm gonna implement what I call core four marketing. Uh, you know? Or they're a brick and mortar stores. So that's gonna entail, uh, making sure that they're Google business profile, uh, has been claimed and is actually working for them. So if they haven't claimed it, I walk them through claiming and verifying that. If they have, then we're gonna go ahead and go ahead and optimize, uh, that And, you know, we're we're looking to make sure that everything stays fresh, and all their information is up to date. As well as their photos. So, yeah, I use AI tools in that process as well. - 如果基本上只是街面流量，我会实施我所谓的“核心四项营销”。呃，你知道吗？或者他们是实体店。所以这将涉及确保他们的 Google 商家配置文件的所有权已被认领，并且确实在为他们发挥作用。如果他们还没认领，我会引导他们完成认领和验证。如果已经认领了，那我们就去优化它。而且，你知道，我们要确保一切保持新鲜感，所有信息都是最新的。照片也是如此。所以，是的，在这个过程中我也使用了 AI 工具。

04:36-04:48 - Researcher: Ah, interesting! You mentioned you use AI tools in that process. Can you give me a specific example of how you're using AI when you're working on something like a Google Business Profile or other parts of that core four marketing? - 啊，有趣！你提到在这个过程中使用了 AI 工具。 你能给我举一个具体的例子吗？说明你在处理像 Google 商家配置文件或核心四项营销的其他部分时，是如何使用 AI 的？

05:09-05:14 - Marcus: Yeah. I'll go ahead and use local SEO. I think Yoast does a great job of making sure that Google Business profiles are optimized, uh, with current information, across the board. - 是的。我会使用本地 SEO 工具。我觉得 Yoast 在确保 Google 商家配置文件全面优化并包含最新信息方面做得非常好。

05:16-05:30 - Researcher: So you're using Yoast for the local SEO optimization piece. When you say you use "AI tools" in the process - are you referring to Yoast specifically, or are there other AI tools you're using? I want to make sure I understand what AI tools are actually part of your workflow. - 所以你使用 Yoast 来进行本地 SEO 优化部分。 当你说你在过程中使用“AI 工具”时——你是指专门指 Yoast，还是还有其他你在使用的 AI 工具？我想确保我理解了哪些 AI 工具实际上是你工作流的一部分。

05:32-05:58 - Marcus: I am. There's a local SEO tool that I am I can't even remember the name of right now. It's slipping it's slipping my mind, but something I use all the time. Especially if it's a client with multiple locations. To make sure that all of the locations in the multisite are up to date - 是的。有一个本地 SEO 工具，我现在甚至想不起名字了。它从我脑子里滑走了，但我一直都在用。特别是如果是一个拥有多个地点的客户。为了确保多站点中所有的地点信息都是最新的。

05:59-06:12 - Researcher: No worries if the name's slipping your mind right now! Beyond those SEO tools, are there other AI tools you use in your work - maybe for content creation, image generation, coding help, or anything else in your web development or digital marketing work? - 如果现在想不起名字也没关系！ 除了这些 SEO 工具，你在工作中还会使用其他 AI 工具吗——也许是用于内容创作、图像生成、编程辅助，或者网页开发及数字营销工作中的其他任何事情？

06:39-07:04 - Marcus: Yeah. I do. I've been using Gemini 2.0 Pro has been great. I've been using Perplexity Pro as well, both of those in coding. You know, I've been doing HTML, CSS, JavaScript, and PHP for for quite a while. And, you know, there's a lot of new stuff in in CSS, that really makes it so I don't need to use JavaScript So I'm usually running code through the AI, to make sure that I am coding in the most elegant and up to date fashion. - 是的，我有。我一直在用 Gemini 2.0 Pro，效果很好。我也一直在用 Perplexity Pro，这两个都用于编程。你知道，我做 HTML、CSS、JavaScript 和 PHP 已经挺长时间了。而且，CSS 中有很多新东西，真的让我不需要再使用 JavaScript 了。所以我通常会通过 AI 运行代码，以确保我正在以最优雅、最现代的方式进行编码。

07:06-07:23 - Researcher: That's really smart - using AI to keep your code modern and elegant, especially with all those new CSS features replacing old JavaScript solutions. When you say you're "running code through the AI" - what does that actually look like? Are you pasting in code you've already written and asking it to improve it, or are you using it differently? - 这真的很聪明——使用 AI 来保持代码的现代感和优雅感，特别是用那些取代了旧 JavaScript 方案的新 CSS 特性。 当你说“通过 AI 运行代码”时——具体是什么样的？你是把写好的代码粘贴进去让它改进，还是以其他方式使用？

07:25-07:53 - Marcus: Yeah. I am. I am optimizing code that I've already written. Running all of my old CSS through Claude or Perplexity Pro, excuse me, And yep. Absolutely. And I'm also using it to generate new code in some instances. - 是的，我是这样做的。我正在优化我已经写好的代码。把我的旧 CSS 全部跑一遍 Claude 或者 Perplexity Pro，抱歉，是的，没错。而且在某些情况下我也用它来生成新代码。

07:55-08:06 - Researcher: So you're both optimizing old code and generating new stuff. You mentioned Gemini, Perplexity, and it sounded like maybe Claude too? Which one do you find yourself reaching for most often, and why that one over the others? - 所以你既优化旧代码也生成新东西。你提到了 Gemini、Perplexity，听起来好像还有 Claude？ 你发现自己最常使用的是哪一个，为什么选择它而不是其他的？

08:08-08:42 - Marcus: I'm still I'm still figuring that out. I think that right now, Gemini is doing, uh, the best job I wanna say, I think that Perplexity Pro has led me down a few rabbit holes. So, you know, if it's something that's gonna be what I perceive is complicated, then I'll probably go for Gemini three or Claude. - 我还在摸索中。我觉得目前 Gemini 做得最好，我想这么说。我觉得 Perplexity Pro 曾让我走过几段弯路。所以，你知道，如果是一些我察觉到会很复杂的事情，那我可能会选择 Gemini 或者 Claude。

08:44-08:53 - Researcher: Interesting - so Perplexity has sent you down some rabbit holes. Can you tell me about a specific time that happened? What were you trying to do, and where did it lead you astray? - 有意思——所以 Perplexity 让你走过弯路。你能告诉我一次具体的情况吗？你当时想做什么，它在哪里把你带偏了？

09:01-09:38 - Marcus: You know, I was doing something really simple. I just wanted to change the mobile menu on the StandardTheme theme in WordPress. And, actually, no. That's not it. I was actually making a an essay or not an essay, like a quiz a lead generation quiz for clients. And using perplexity pro to do this, And I noticed that the code it generated really had some fairly large security gaps. That would be problematic. - 你知道，我当时在做一件非常简单的事情。我只是想更改 WordPress 中 StandardTheme 主题的移动端菜单。而且，实际上，不。不是那个。我当时实际上是在做一个测试题，不是文章，而是一个为客户获取潜客的在线测试。我用 Perplexity Pro 来做这个，我注意到它生成的代码确实有一些相当大的安全漏洞。那会出问题的。

09:41-09:55 - Researcher: Whoa, security gaps in a lead generation quiz - that's definitely not something you want to discover after it's live. What kind of security issues did you spot? And how did you catch them - did you notice something looked off, or do you have a process for checking AI-generated code? - 喔，潜客获取测试题里的安全漏洞——这绝对不是上线后才想发现的东西。 你发现了什么样的安全问题？你是怎么抓住它们的——是注意到某些地方看起来不对劲，还是你有一套检查 AI 生成代码的流程？

09:59-10:20 - Marcus: The fields were just they're they're they didn't have any security whatsoever. You'd be able to inject JavaScript directly into the database There was all sorts of issues with with, uh, with perplexity. That I had to resolve by using Claude, I think. - 那些字段根本没有任何安全性可言。你可以直接向数据库注入 JavaScript。Perplexity 生成的代码有各种各样的问题。我想我最后不得不使用 Claude 来解决。

10:22-10:37 - Researcher: So the form fields had no sanitization or validation - yikes. You had to bring in Claude to fix what Perplexity created. That makes me wonder - when you're generating code with AI now, do you always run it through a second AI to check it, or was that a one-time thing after that bad experience? - 所以表单字段没有经过任何清理或验证——天呐。你不得不找来 Claude 来修复 Perplexity 制造的麻烦。 这让我很好奇——现在当你用 AI 生成代码时，你是否总是通过第二个 AI 来交叉检查，还是说那是上次糟糕经历后的偶然行为？

10:54-11:10 - Marcus: Generally, what I do is I specifically tell the AI model to, you know, use up to date security precautions while generating the code. But since that experience I tend to use Perplexity Pro less for for those kind of tasks. - 通常情况下，我会专门告诉 AI 模型在生成代码时使用最新的安全预防措施。但自那次经历以来，我倾向于更少地将 Perplexity Pro 用于此类任务。

11:14-11:32 - Researcher: So you've learned to prompt for security upfront, and you've shifted away from Perplexity for code generation after that experience. Now I'd love to shift gears and talk more about AI tools in general. You mentioned Gemini, Perplexity Pro, and Claude - are there other AI apps or web tools you use regularly in your life, either for work or personal stuff? - 所以你学会了预先在提示词中加入安全要求，并且在那次经历后，在代码生成方面转而减少使用 Perplexity。 现在我想转个话题，聊聊更广泛的 AI 工具。你提到了 Gemini、Perplexity Pro 和 Claude——在你的生活中，无论是为了工作还是个人事务，还有其他你经常使用的 AI 应用或网页工具吗？

11:36-12:00 - Marcus: Yeah. Sure. I use Tidio on my website Uh, that's the chatbot. That I use. I find that very useful. It's great for responding to clients, so, you know, I don't have to. It's got some great background workflows that I like a lot. - 是的，当然。我在我的网站上使用 Tidio。那是聊天机器人。我发现它非常有用。它非常适合回复客户，所以，你知道，我就不用亲自回复了。它有一些我很喜欢的很棒的后台工作流。

12:02-12:16 - Researcher: Tidio for the chatbot - nice. So it's handling client inquiries automatically with those background workflows. Beyond Tidio, Gemini, Perplexity, and Claude - are there any other AI tools you're using? Maybe for image generation, video, writing, music production, or anything else? - 聊天机器人用 Tidio——不错。所以它通过那些后台工作流自动处理客户咨询。 除了 Tidio、Gemini、Perplexity 和 Claude——你还在使用其他 AI 工具吗？也许是图像生成、视频、写作、音乐制作或其他任何方面？

12:25-12:50 - Marcus: The Adobe image generator I've used. I've found that to be really great. I like to use that for a lot of my background images on websites you know, it's great that that they're going to be you know, there's no copyright on them and they're pretty high quality. So that is definitely part of my web design workflow. - 我用过 Adobe 的图像生成器。我觉得它真的很棒。我喜欢把它用于网站上的很多背景图片。你知道，好处是这些图片不会有版权问题，而且质量很高。所以这绝对是我网页设计工作流的一部分。

12:52-13:05 - Researcher: Adobe's image generator for backgrounds - that's smart, especially avoiding copyright issues. So we've got Gemini, Perplexity Pro, Claude, Tidio, and Adobe's image generator. Out of all these AI tools you're using, which ones are you actually paying for right now? - 把 Adobe 的图像生成器用于背景——这很聪明，特别是避开了版权问题。 那么我们有了 Gemini、Perplexity Pro、Claude、Tidio 和 Adobe 图像生成器。在所有这些你使用的 AI 工具中，你目前实际上在为哪些付费？

13:18-13:37 - Marcus: I think I've got a subscription for Perplexity Pro. And I do have a lot of I have pro versions or access to the pro versions of the other AI models? Through techTrain. They have a playground that I use. That because I do AI training with TechTrain, I get access to all of the latest AI models. - 我想我订阅了 Perplexity Pro。而其他的 AI 模型，我有专业版或者是专业版的访问权限？通过 techTrain。他们有一个我使用的测试场（playground）。因为我和 techTrain 一起做 AI 训练，所以我可以访问所有最新的 AI 模型。

13:41-13:59 - Researcher: Ah, interesting! So you're doing AI training work with TechTrain and that gives you playground access to the pro versions. Let me make sure I understand your paid subscriptions - you're paying for Perplexity Pro directly. What about Gemini, Claude, Adobe, and Tidio - are any of those separate paid subscriptions, or are they all coming through your TechTrain access? - 啊，有趣！所以你正在与 techTrain 合作进行 AI 训练，这让你获得了访问专业版测试场的权限。 让我确认一下你的付费订阅——你直接付费购买的是 Perplexity Pro。那么 Gemini、Claude、Adobe 和 Tidio 呢——这些有单独的付费订阅吗，还是它们都是通过你的 techTrain 权限获得的？

14:02-14:16 - Marcus: Now, I'm using them through my techTrain access Perplexity Pro is the only one that I'm paying for. But well, Tidio. I guess I'm paying for Tidio as well. - 不，我是通过 techTrain 权限使用的。Perplexity Pro 是我唯一付费的。不过，嗯，还有 Tidio。我想我也在为 Tidio 付费。

14:17-14:26 - Researcher: Got it - so Perplexity Pro and Tidio are your direct paid subscriptions. How much are you paying for each of those? What's the pricing model - monthly, annual, or something else? - 明白了——所以 Perplexity Pro 和 Tidio 是你直接付费订阅的。 你分别为这两个付多少钱？定价模式是怎样的——按月、按年还是其他的？

14:29-14:44 - Marcus: Got a deal. Uh, I can't remember how much I paid for a year of Perplexity Pro. I wanna say it's, a hundred and twenty nine bucks. I could be wrong. The Tidio is like twenty nine dollars a month. - 拿到了一个优惠。呃，我不记得我为一年的 Perplexity Pro 付了多少钱。我想大概是 129 美元。我可能记错了。Tidio 大约是每月 29 美元。

14:46-15:00 - Researcher: Okay, so roughly $129year for Perplexity Pro, and $29month for Tidio - that's about $478 total per year just on those two. And with TechTrain - are you paying anything for that access, or is the playground access just a perk that comes with doing the AI training work for them? - 好的，Perplexity Pro 大约每年 129 美元，Tidio 每月 29 美元——光这两个加起来每年就大约 478 美元。 至于 techTrain——你为那个访问权限付过费吗，还是说测试场访问权限只是你为他们做 AI 训练工作的一项福利？

15:03-15:04 - Marcus: It is a perk. - 这是一项福利。

15:06-15:21 - Researcher: So the TechTrain playground is a perk from your training work - that's a pretty valuable benefit! When you think about content generation tools - like for writing, copywriting, or creating marketing content for your clients - are you using any AI tools for that? Either through TechTrain or paid separately? - 所以 techTrain 测试场是你训练工作的福利——这真是一个非常有价值的待遇！ 当你想到内容生成工具时——比如用于写作、文案策划或为你的客户创建营销内容——你会使用任何 AI 工具吗？是通过 techTrain 还是单独付费？

15:23-15:30 - Marcus: You know, that's where I use Perplexity Pro quite a bit, as a matter of fact. That, uh, yeah. - 你知道，事实上在那方面我用 Perplexity Pro 用得挺多的。是的。

15:32-15:41 - Researcher: So Perplexity Pro is pulling double duty - both for code and for content generation. What kind of content are you creating with it? Marketing copy for client websites, blog posts, something else? - 所以 Perplexity Pro 承担了双重任务——既用于代码，也用于内容生成。 你用它创建什么内容？客户网站的营销文案、博客文章，还是别的？

15:52-16:10 - Marcus: Absolutely. Marketing copy for marketing copy for, uh, client websites. Sometimes I run through email responses to clients, just check my tone. And, uh, yeah, blog posts, absolutely. I use it for blog posts. Yeah. - 确实如此。给客户网站写营销文案。有时我会用它跑一遍给客户的回复邮件，只是为了检查我的语气。而且，是的，博客文章，绝对的。我用它来写博客文章。是的。

16:12-16:21 - Researcher: Checking your tone on client emails - that's a use case I hadn't thought of. Does it ever tell you to soften things, or what kind of adjustments does it usually suggest? - 检查客户邮件的语气——这是我之前没想过的用法。它会告诉你把语气放缓和一点吗，或者它通常会建议做哪种调整？

16:23-16:56 - Marcus: I don't know if I'm really checking tone as much oh, I guess I am checking tone. I'm I'm trying to, uh, make sure I've got a good balance of, uh, professionalism, friendliness, and tech savvy. And I wanna make sure that I'm always speaking in a language that's easy for people to understand, especially if they're not as technically savvy as I am. - 我不知道我是否真的在检查语气——哦，我想我确实是在检查语气。我是在试图确保我在专业、友好和技术专业感之间取得良好的平衡。我想确保我使用的语言对人们来说易于理解，特别是如果他们不像我这样懂技术的话。

16:58-17:13 - Researcher: That makes total sense - translating tech speak into something clients can actually understand without sounding condescending. When you run an email through Perplexity for that balance check, does it usually nail it on the first try, or do you find yourself going back and forth with it to get the tone right? - 这完全说得通——将技术语言转化为客户真正能理解的内容，同时又不显得居高临下。 当你为了这种平衡检查在 Perplexity 中运行邮件时，它通常能一次搞定吗，还是你会发现自己需要和它来回沟通才能把语气调对？

17:36-17:46 - Marcus: You know, I I take suggestions. I usually kind of will take a look at the output and edit it myself. I generally don't tinker with it in the sense of going through iterations with the with the AI to get a totally perfect response. But, you know, I think that it's helpful in in some circumstances. - 你知道，我会采纳建议。我通常会看一下输出结果，然后自己修改。我通常不会在不断反馈迭代的意义上去折腾它，为了让 AI 给出一个完美的回复。但是，你知道，在某些情况下它确实很有帮助。

17:48-17:56 - Researcher: So you're using it more as a starting point or a second opinion, then doing your own manual polish rather than iterating with the AI. Thinking about all these AI - 所以你更多地将其作为一个起点或第二意见，然后进行手动润色，而不是与 AI 进行迭代。 思考一下所有这些 AI...

17:56-17:57 - Marcus: That's correct. - 没错。

17:59-18:09 - Researcher: Thinking about all these AI tools you're using - has there been any recent purchase or subscription that you felt really satisfied with? Something where you thought "this was totally worth the money"? - 思考一下所有这些你正在使用的 AI 工具——最近有没有什么购买或订阅让你感到非常满意？某种让你觉得“这钱花得完全值”的东西？

18:10-18:13 - Marcus: I think Tidio is totally worth the money. - 我觉得 Tidio 完全值得花这个钱。

18:15-18:23 - Researcher: What is it about Tidio specifically that makes it feel worth the $29month? Walk me through what it's doing for you that makes that price feel justified. - Tidio 到底有哪些地方让你觉得值这每月 29 美元？带我了解一下它为你做了什么，让你觉得这个价格是合理的。

18:23-18:52 - Marcus: You know, it's easy to train. It is gonna be a way for people to come to the website to get information directly without having to search for it. It's got great back office tools, I can program it to go ahead and validate leads. For the services that I offer. It it does a lot that's quite valuable. - 你知道，它很容易训练。它能让人们来到网站直接获取信息，而不需要四处寻找。它有很棒的后台工具，我可以对其进行编程，直接校验潜在客户的有效性。针对我提供的服务。它做了很多非常有价值的事情。

18:54-19:08 - Researcher: The lead validation piece sounds really useful - so it's not just answering questions, it's actually qualifying people for you before they even reach you. How many leads or conversations is Tidio handling for you in a typical month? I'm curious what the volume looks like. - 潜客验证这部分听起来非常有用——所以它不只是回答问题，它实际上是在人们联系到你之前就在为你筛选合格的意向客户。 在典型的一个月中，Tidio 会为你处理多少条潜客信息或对话？我很好奇这个量级。

19:23-19:30 - Marcus: You know, since I've started using it, I've really just recently gone into you know, had it in in in that capacity. But I would say that you know, it does a good twenty, thirty per month. - 你知道，自从我开始使用它以来，我也只是最近才开始以这种规模使用它。但我会说，它一个月能处理二三十个。

19:32-19:42 - Researcher: 20-30 qualified leads per month - that's pretty solid for $29. On the flip side, has there been any AI tool or subscription recently where you felt disappointed or like it wasn't worth what you paid? - 每月 20 到 30 个合格的潜客——对于 29 美元来说，这相当不错了。 另一方面，最近有没有什么 AI 工具或订阅让你感到失望，或者觉得不值你付的钱？

19:47-20:21 - Marcus: You know, perplexity is pretty good. But like I said, I've run into some issues with it. I don't know that you know, when that subscription runs out. If I'll renew it, I think that it would be useful at uh, you know, if we're if we're at a less expensive I don't know that I wanna renew for a hundred and twenty nine dollars. I think that, uh, there's some some better options for that. I'd probably go with, uh, Gemini. - 你知道，Perplexity 挺好的。但就像我说的，我遇到过一些问题。我不知道当订阅到期时我是否会续费。我觉得如果价格更便宜一点的话它会有用，我不知道我想不想花 129 美元续费。我认为有更好的选择。我可能会选择使用 Gemini。

20:24-20:40 - Researcher: So Perplexity at $129year is feeling like it might not be worth renewing, especially with the security issues you've hit and better alternatives available. If Gemini 3 wasn't free through TechTrain - if you had to pay for it directly - would you be willing to pay for that instead of Perplexity? And what price would feel right to you? - 所以 129 美元一年的 Perplexity 让你觉得可能不值得续费，特别是考虑到你遇到的安全问题，而且还有更好的替代方案。 如果 Gemini 3 在 techTrain 上不再免费——如果你必须直接付费——你会愿意付钱用它来代替 Perplexity 吗？什么样的价格对你来说是合适的？

20:41-21:01 - Marcus: I mean, I'm I'd have to check the what the going rate is. For Gemini. But if it were the similar price, I would go Gemini. I'm fairly confident that Gemini can do everything perplexity can. If not better. Probably better. - 我的意思是，我得查一下 Gemini 的现行市价。但如果价格差不多，我会选 Gemini。我很确信 Gemini 能做 Perplexity 能做的一切。即使不比它更好，大概率也会更好。

21:03-21:11 - Researcher: So at a similar price point, Gemini wins for you - you're seeing it perform better overall. This makes me - 所以在价格相近的情况下，Gemini 在你这里胜出——你看到它的整体表现更好。 这让我...

21:11-21:11 - Marcus: I am. - 是的。

21:16-21:26 - Researcher: This makes me curious about your AI spending going forward. Over the next 6 months, do you see yourself spending more, less, or about the same on AI tools? What's driving that direction for you? - 这让我很好奇你未来的 AI 支出。在接下来的 6 个月里，你认为你在 AI 工具上的支出会更多、更少，还是保持不变？推动这种趋势的原因是什么？

21:51-22:05 - Marcus: See myself spending about the same. I think that right now, my work is, uh, pretty good. You know? If I am in a situation where I need more direct access, like, if I need to start using APIs, with Gemini. That's something I may be looking at. In the near future. Then I would make that switch But right now, I'm happy with with my AI implementation. - 觉得支出会保持不变。我认为目前我的工作状态挺好的。你知道吗？如果我遇到需要更直接访问权限的情况，比如，如果我需要开始在 Gemini 中使用 API。那是我近期可能会考虑的事情。那时我会做出改变。但现在，我对我的 AI 应用现状感到满意。

22:07-22:20 - Researcher: So staying flat - you've got what you need right now. The API access for Gemini would be the trigger to change things up. What would make you need API access? Is that for integrating AI directly into client websites, or something else you're building? - 所以保持平稳——你现在拥有你所需要的一切。Gemini 的 API 访问权限会是改变现状的触发点。 什么样的情况下你会需要 API 访问权限？是为了将 AI 直接集成到客户网站中，还是为了你正在构建的其他东西？

22:37-23:35 - Marcus: That, but also I am bidding on contracts. For home services. A lot of these private equity firms have bought up a lot of home services businesses across the country. So there are a lot of contracts for consolidating and optimizing all these different web properties So they're standardized into, you know, one corporate home services site. So that entails you know, that's gonna entail a lot of web scraping. A lot of these home services companies have contracts with different web development firms that either don't give them access to the database or you know, maybe they just don't even have access to the web developer anymore. There's just a lot of those situations that means that if you wanna go ahead and standardize, you're gonna have to simply scrape data directly from the web. So - 有这方面的原因，但也因为我正在投标一些合同。针对家庭服务。许多这些私募股权公司收购了全美各地的许多家庭服务业务。因此，有很多合同是关于整合和优化所有这些不同的 Web 属性。这样他们就可以标准化为一个公司制的家庭服务主站。所以这涉及，你知道，这涉及大量的网络爬虫。许多这些家庭服务公司与不同的网页开发公司签有合同，那些公司要么不给他们数据库访问权限，或者，你知道，也许他们甚至都联系不到原来的网页开发人员了。有很多这种情况，这意味着如果你想要进行标准化，你就不得不直接从网上抓取数据。所以...

23:40-24:01 - Researcher: Wow, that's a whole different scale of work - scraping and consolidating multiple home services sites for PE firms. That would definitely need API-level AI access to process all that data efficiently. Let me dig deeper into your paid tools. Let's start with Perplexity Pro since you mentioned you might not renew. When did you first start using Perplexity - was it the free version first, or did you go straight to Pro? - 哇，那完全是另一种工作规模了——为私募股权公司抓取并整合多个家庭服务网站。这确实需要 API 级别的 AI 访问权限来高效处理所有数据。 让我深入了解一下你的付费工具。让我们从 Perplexity Pro 开始，既然你提到可能不会续费。你最早是什么时候开始使用 Perplexity 的——是先用免费版，还是直接用了专业版？

24:13-24:22 - Marcus: I had the free version first, and then I think I was I got a some kind of discount through PaymentApp so that the annual rate was cheaper. - 我先用的免费版，然后我想我是通过 PaymentApp 获得了一些折扣，所以年费更便宜。

24:24-24:34 - Researcher: A PaymentApp discount deal - that's an interesting way to hook people in. Do you remember how long you used the free version before you jumped to Pro? And what were you mainly using it for during that free period? - PaymentApp 的折扣优惠——这是一个有趣的吸引手法。你还记得在跳到专业版之前你用了多久的免费版吗？在那个免费期间你主要用它来做什么？

24:37-25:09 - Marcus: Think I used it for a couple of months. It really wasn't my main focus at that point. You know, one of the main reasons I I paid for Perplexity Pro was because of the PaymentApp deal. It was just kind of a knee jerk reaction. You know, I wanted to try out one of the paid plans. And at that point, perplexity wasn't something that I had access to through TechTrain. - 我想我用了两个月。那时它真的不是我的主要关注点。你知道，我购买 Perplexity Pro 的主要原因之一是因为 PaymentApp 的那个优惠。那只是一种下意识的反应。你知道，我想尝试一下其中的一种付费计划。而那时，我还没法通过 techTrain 访问 Perplexity。

25:12-25:26 - Researcher: So it was partly opportunistic - the PaymentApp deal made the barrier low enough to just try a paid plan, and Perplexity wasn't available through TechTrain yet. What changed after you paid? Did you suddenly start using it more, or did the Pro features unlock something specific you needed? - 所以这部分是机会主义——PaymentApp 的优惠让尝试付费计划的门槛变得足够低，而且当时 Perplexity 还没加入 techTrain 的访问范围。 你付费后有什么变化吗？你是突然开始更频繁地使用它，还是专业版的功能开启了你需要的特定功能？

25:35-25:51 - Marcus: You know, the pro features, I think, are just simply it's more robust. Uh, memory is uh, I'm not sure if you if it even remembers conversations to the same degree on the free version. Yeah. - 你知道，我觉得专业版的功能就是更强大。呃，记忆力——我不确定免费版是否能将对话记到同样的程度。是的。

25:53-26:03 - Researcher: So better conversation memory and more robust responses. When you started using Pro more heavily, what was the first big task or project where you thought "okay, this Pro version is actually useful"? - 所以对话记忆更好，响应更强大。当你开始重度使用专业版时，第一个让你觉得“好吧，这个专业版确实有用”的大任务或大项目是什么？

26:12-26:35 - Marcus: Started to I guess, it it was when I was making the mobile menu. I was trying to switch it out to mobile menu that comes with the StandardTheme WordPress theme. I wanted to have something that actually worked. The mobile menu that ships with the StandardTheme theme is trash. And I wanted to have something that was more ergonomic and, uh, stylish. - 开始——我想，是在我做移动端菜单的时候。我想把原本的菜单换成 WordPress 那个 StandardTheme 主题自带的。我想要一些真正好用的东西。StandardTheme 自带的移动端菜单简直是垃圾。我想要一些更符合人体工程学、更时尚的东西。

26:37-26:52 - Researcher: The StandardTheme mobile menu is notoriously clunky - I hear you. So Perplexity Pro helped you build a custom mobile menu replacement. Walk me through that - what kind of prompts were you giving it, and did it actually generate working code on the first try, or did you have to go back and forth? - StandardTheme 的移动端菜单是出了名的臃肿——我懂你。所以 Perplexity Pro 帮你构建了一个定制的移动端菜单替代方案。 带我了解一下那个过程——你给了它什么样的提示词？它是一次性生成了可运行的代码吗，还是你必须来回调整？

26:55-27:31 - Marcus: Let's see. Um, you know, I instructed I instructed it to make a slide out mobile menu and gave it my company icon. And, uh, I think I told it that I wanted a dark theme. And a hamburger button. I remember not really being too detailed in my prompt, but it did a great job. - 让我看看。嗯，你知道，我指示它做一个滑出式移动端菜单，并提供了我公司的图标。而且，我想我告诉它我想要深色主题和一个汉堡菜单按钮。我记得我的提示词并不是特别详细，但它做得很好。

27:33-27:41 - Researcher: That's pretty impressive - minimal prompt, good output. Did you have to do much tweaking to the code it gave you, or did it mostly just drop in and work? - 挺令人印象深刻的——提示词极简，输出结果很好。你需要对它给你的代码做很多调整吗，还是说基本上直接放进去就能用？

27:43-28:30 - Marcus: I did have to do some tweaking. The CSS was not was not reading, I guess, it was being overwritten by the theme, I had to do a little bit of tweaking to get it to work. I found that instead of having the CSS in the theme plug in I'm sorry. In the in the plug in for the the mobile menu, It works better if I just drop it into additional CSS in the customization part of WordPress. So that's something that you generally don't do. - 我确实做了一些微调。CSS 无法读取，我觉得是被主题覆盖了，我必须做一点微调才能让它运行。我发现，与其把 CSS 放在主题插件里——抱歉，放在移动菜单的插件里，不如直接把它扔进 WordPress 自定义部分的“附加 CSS”中效果更好。这是通常情况下不会做的事情。

28:33-28:48 - Researcher: Interesting workaround - bypassing the plugin's CSS entirely and dropping it into the theme customizer to avoid specificity battles. That's the kind of WordPress quirk you only learn from experience. Did Perplexity suggest that fix, or did you figure that out on your own after the code wasn't working? - 有趣的规避方法——完全绕过插件的 CSS，直接扔进主题自定义器里以避免特异性冲突。这种 WordPress 的怪癖确实只有通过实际经验才能学到。 是 Perplexity 建议了这种修复方法，还是在代码无法运行后，你自己想出来的？

28:49-28:50 - Marcus: I figured that out on my own. - 是我自己想出来的。

28:53-29:11 - Researcher: Got it - so Perplexity gave you the foundation, but your WordPress knowledge solved the implementation issues. Let me ask about the security incident you mentioned earlier - the lead gen quiz with the JavaScript injection vulnerability. Can you walk me through what happened there? Like, what were you asking Perplexity to build, and how did you discover the security gap? - 明白了——所以 Perplexity 给了你基础，但你的 WordPress 知识解决了实现问题。 让我问问你之前提到的安全事件——那个带 JavaScript 注入漏洞的潜客测试题。你能带我回顾一下发生了什么吗？比如，你当时让 Perplexity 构建什么，你是怎么发现那个安全漏洞的？

29:19-30:09 - Marcus: You know, I noticed that a lot of a lot of sites like mine use you know, these lead generation quizzes. It seemed to work pretty well. But I think what hap yeah. What happened was I took the code, that was generated and I had Claude in TechTrain. Right? I had Claude analyze that code, and Claude was the one that actually alerted me of the security issues. - 你知道，我注意到很多像我这样的网站都使用这种获取潜客的测试题。它似乎效果不错。但我想发生了……是的。发生的事情是，我拿到了生成的代码，而我在 techTrain 里有 Claude，对吧？我让 Claude 分析了那段代码，正是 Claude 提醒了我存在的安全问题。

30:13-30:23 - Researcher: So you're actually running a quality check process - Perplexity generates, then you run it through Claude as a security auditor. - 所以你实际上运行了一套质量检查流程——Perplexity 生成，然后你把它跑一遍 Claude 作为安全审计。

30:23-30:24 - Marcus: I I did. Yep. - 我确实这样做了。是的。

30:27-30:35 - Researcher: That's smart. Before that incident, were you already in the habit of double-checking Perplexity's code with Claude, or did something make you suspicious about that particular quiz code? - 很聪明。在那次事件之前，你就已经在习惯用 Claude 双重检查 Perplexity 的代码了吗，还是有什么让你对那次特定的测试题代码产生了怀疑？

30:38-31:08 - Marcus: Well, I wanted to take a look at what an AI model that was supposed you know, built for code I wanted to compare the results. I also compared the results with, uh, Gemini. Which also also alerted me of those security issues. So of the three, the only one that didn't was Perplexity Pro. - 嗯，我想看看一个本应对……你知道的，为代码而生的 AI 模型——我想对比一下结果。我还把结果和 Gemini 做了对比。Gemini 也提醒了我那些安全问题。所以，在这三个工具中，唯一没有发现问题的是 Perplexity Pro。

31:11-31:26 - Researcher: Ouch - so both Claude and Gemini flagged the security holes, but Perplexity Pro was the only one that generated vulnerable code in the first place. That's a pretty clear quality gap. After that happened, did you change how you use Perplexity? Like, do you still trust it for certain tasks but not others? - 哎哟——所以 Claude 和 Gemini 都指出了安全漏洞，但 Perplexity Pro 是唯一一个从一开始就生成了有风险代码的工具。这是一个非常明显的质量差距。 在那次事件之后，你使用 Perplexity 的方式改变了吗？比如，你是否仍然在某些任务上信任它，而在其他任务上不信任它？

31:45-31:55 - Marcus: Absolutely. Yeah. I've used it mostly for doing things like generating copy marketing copy, blogs, I mentioned before. You know, if I'm writing proposals, you know, for contracts, it's great for that sort of thing. I just have less confidence in it, uh, when it comes to code generation. - 绝对的。是的。我主要用它来做我之前提到的生成文案、营销文案、博客之类的事情。你知道，如果我正在为合同编写提案，它非常擅长此类事情。只是在代码生成方面，我对它的信心降低了。

31:57-32:04 - Researcher: Makes sense - you've basically demoted it from coding to copywriting after that trust break. When you're using it for marketing - 有道理——在那种信任破灭之后，你基本上把它从编程工具降级为文案工具了。 当你将其用于营销时

32:04-32:05 - Marcus: Yeah. - 是的。

32:08-32:16 - Researcher: When you're using it for marketing copy and proposals now, are you still running those outputs through other AI models to check them, or do you trust Perplexity more for writing than for code? - 现在当你将其用于营销文案和提案时，你是否仍会将这些输出内容放入其他 AI 模型中进行检查，还是说相比代码，你更信任 Perplexity 的写作能力？

32:20-32:42 - Marcus: Sometimes I'll run it through, you know, the ChatGPT or some other models just to compare But for things like that, I really don't think they do a better job. I've seen them do kind of a worse job as a matter of fact. So you know, I think that Perplexity Pro is, uh, where it should be in my workflow. - 有时我会把它放进 ChatGPT 或其他模型里对比一下。但对于这类事情，我真的不认为其他模型做得更好。事实上，我看到它们做得反而更差一些。所以，我觉得 Perplexity Pro 在我工作流中的位置是实至名归的。

32:45-32:55 - Researcher: So for writing tasks, Perplexity Pro actually holds its own or even beats ChatGPT in your experience. That's interesting - what makes Perplexity's writing output better for your needs compared to the others? - 所以在写作任务上，根据你的经验，Perplexity Pro 实际上能立于不败之地，甚至胜过 ChatGPT。这很有趣——是什么让 Perplexity 的写作输出比其他工具更符合你的需求？

33:06-33:29 - Marcus: It could be how I'm prompting it, but for a little while, I noticed that, uh, the model of ChatGPT I was using, I think it was ChatGPT, was being kind of terse. Like, the the output the output was not something I would expect from an AI model that was instructed to act as a world class digital marketer. - 可能是我的提示词方式的问题，但有一段时间我注意到，我使用的 ChatGPT 模型（我想应该是 ChatGPT）表达非常简短。就像是，它的输出结果并不像是一个被指令要求充当“世界级数字营销专家”的 AI 模型应有的水平。

33:32-33:43 - Researcher: Ha - you wanted world-class digital marketer energy and got minimalist intern vibes instead. Does Perplexity give you more elaborate, polished marketing language? Or is it something else about the tone or structure that works better? - 哈——你想要的是世界级数字营销专家的能量，结果得到的却是极简主义实习生的感觉。 Perplexity 是否能给你更详尽、更精炼的营销语言？还是说在语调或结构方面有其他更好的地方？

34:10-34:14 - Marcus: I think the output, uh, with minimal instruction tends to give me what I'm really looking for personally. You know? I've got no problems, you know, tweaking, It just it does does a better job of giving me a launching point to, uh, you know, go ahead and polish myself. - 我觉得在只有极少指令的情况下，它的输出往往就能给出我个人真正想要的东西。你懂吗？我后续进行微调完全没问题，它只是在提供一个让我可以继续自行润色的“切入点”方面做得更好。

34:17-34:32 - Researcher: So it's closer to your voice or style right out of the gate - less work to get it where you need it. You mentioned earlier you might not renew Perplexity Pro because of cost and the security issues. When does your subscription come up for renewal, and what's going through your head about that decision right now? - 所以它一上来就更接近你的口吻或风格——不需要太多工作就能达到你的要求。 你之前提到，由于成本和安全问题，你可能不会续订 Perplexity Pro。你的订阅什么时候到期？现在关于这个决定，你脑子里是怎么想的？

34:32-35:10 - Marcus: Got another six months and things definitely can change in six months, especially in the AI world. So you know, there's no real way to to to judge the future in AI. So, you know, I'll probably be making that decision closer to you know, the time of renewal, and we'll see how Perplexity Pro is doing then. I expect that it should change And whether those are gonna be improvements across the board, is yet to be seen. - 还有六个月，六个月里肯定会发生很多变化，尤其是在 AI 领域。所以，现在确实没法预判 AI 的未来。我可能会在临近续订的时候再做决定，看看那时候 Perplexity Pro 表现如何。我期待它会有所改变，但是否是全面的提升，还有待观察。

35:12-35:20 - Researcher: Fair - six months is an eternity in AI. If Perplexity Pro stayed exactly as it is today, what would make you renew versus switch to something else like Gemini? - 也是——六个月在 AI 领域就像永恒一样长。如果 Perplexity Pro 保持现状完全不变，什么会让你选择续订，而不是转向 Gemini 之类的其他工具？

35:33-36:27 - Marcus: What would make me renew instead of switch to something else, like Gemini? I think, you know, if the workflow that I have with Perplexity Pro in regard to the to the writing if that changes you know, if I find that I'm getting better results elsewhere that's definitely gonna affect my decision to renew. If I find that, you know, it's not only giving me good results with writing, but, um, finding, you know, some some other positive outcomes from it, then, you know, that'll that might persuade me to renew. So unforeseen advantages perhaps persuade me to renew. - 什么会让我续订而不是转向 Gemini 之类的？我想，如果我目前使用 Perplexity Pro 进行写作的工作流发生了变化，或者我发现其他地方能提供更好的结果，那肯定会影响我的续订决定。如果我发现它不仅在写作方面表现良好，还能带来其他积极的效果，那可能会说服我续订。所以，一些意想不到的优势或许能说服我。

36:30-36:45 - Researcher: So it's really about whether Perplexity maintains that writing quality edge you're experiencing now, or surprises you with new capabilities. You mentioned Gemini as an alternative at a similar price point. Are you already paying for Gemini separately, or are you just accessing it through the TechTrain playground? - 所以关键在于 Perplexity 是否能保持你目前体验到的写作质量优势，或者是否能用新功能给你带来惊喜。 你提到 Gemini 是一个价格相近的替代方案。你已经是单独为 Gemini 付费了吗？还是说你只是通过 TechTrain 的演练场（Playground）在使用它？

36:48-37:15 - Marcus: I'm just accessing it through the TechTrain Playground. But, you know, Gemini is something I've been thinking about, uh, for. I just haven't really looked into it. It really depends on a few things coming up in the future. You know, if these contracts land with the web migration then, you know, I will absolutely be looking for a more robust solution - 我只是通过 TechTrain 演练场在使用它。但是，Gemini 是我一直考虑的事情。只是我还没真正深入研究。这取决于未来几件事的进展。如果那些网站迁移的合同谈成了，那我绝对会去寻找一个更强大的解决方案。

37:18-37:29 - Researcher: So landing those PE home services contracts would be the trigger to invest in a paid Gemini subscription. What specifically about those big migration projects would make you need something more robust than what you have now? - 所以，拿下那些私募支持的家庭服务合同将是投资付费 Gemini 订阅的触发点。关于那些大型迁移项目，具体是什么让你需要比现在更强大的工具？

37:30-37:38 - Marcus: I don't have API access. Through TechTrain, so that's something that doesn't come with a playground. So that would be huge. - 通过 TechTrain 我没有 API 访问权限，那是演练场不具备的功能。所以那将是非常重要的。

37:41-37:56 - Researcher: Ah - API access would let you automate parts of the web scraping and migration workflow instead of manually prompting. That makes total sense for handling multiple properties at scale. Have you scoped out what Gemini API pricing looks like, or what you'd actually build with it for those migrations? - 啊——API 访问可以让你自动化网页抓取和迁移工作流的一部分，而不是手动输入提示词。对于大规模处理多个资产来说，这完全合情合理。 你有没有了解过 Gemini API 的定价，或者对于那些迁移项目，你实际上会用它构建什么？

38:14-38:18 - Marcus: I have not. Uh, there are tools that do scraping. That I also have only scanned but, you know, I I assume they're probably using either ChatGPT or Gemini you know, at their core. - 我还没看过。有一些抓取工具，我也只是大致浏览了一下，但我猜它们的核心可能要么是用 ChatGPT，要么是用 Gemini。

38:20-38:40 - Researcher: Yeah, likely. Though building your own API workflow would give you way more control over what gets scraped and how it's structured versus off-the-shelf tools. Let me shift gears - we've talked a lot about Perplexity, Gemini, Claude. Are there any other AI tools in your work production stack we haven't covered yet? Things you use for design, research, automation, anything like that? - 是的，很有可能。虽然构建自己的 API 工作流会让你对抓取内容和数据结构有更多的控制权，而不是使用现成的工具。 让我换个话题——我们聊了很多关于 Perplexity、Gemini 和 Claude。在你的生产工具栈中，还有其他我们还没提到的 AI 工具吗？比如你用于设计、研究、自动化之类的工具？

38:42-39:10 - Marcus: Um, just the images. You know? I use I guess I use Perplexity for images as well as Adobe. I've really tried them all. What else do we have? I want to start using Midjourney for some marketing initiatives that I'm thinking of for site build and design. - 嗯，就是图像方面。我也用 Perplexity 生成图片，还有 Adobe。我真的都尝试过了。还有别的吗？我想开始为了我正在构思的一些网站建设和设计营销方案而使用 Midjourney。

39:13-39:27 - Researcher: Okay, so Adobe's image generator you mentioned earlier for backgrounds, Perplexity for images too, and you're eyeing Midjourney for future marketing stuff. Are you paying for Adobe's image generator separately, or does that come bundled with something you already subscribe to? - 好的，所以你之前提到的 Adobe 图像生成器用于背景，Perplexity 也用于图片，并且你正关注 Midjourney 用于未来的营销工作。 你是单独为 Adobe 的图像生成器付费，还是它包含在你已经订阅的某个套餐里？

39:29-39:44 - Marcus: You know, I don't use it a lot. I've only used the free package. It's just something I started using. A little while ago. So I haven't hit the wall where I need to start paying for it yet. - 其实我用得不多。我只用过免费套餐。这是我不久前才开始使用的。所以我还没遇到必须开始付费的限制。

39:46-39:58 - Researcher: Got it - still within free tier limits since you're just using it occasionally for backgrounds. When you say you use Perplexity for images too, what are you generating with that? And how does that compare to Adobe's output? - 明白了——因为只是偶尔用于背景，所以还在免费额度内。 当你说你也用 Perplexity 生成图像时，你通常生成什么？它与 Adobe 的输出相比如何？

40:08-40:48 - Marcus: You know, it's it's interesting. I find that, uh, you know, sometimes Perplexity does a great job and sometimes it doesn't. And this might be you know, it might have to do with my prompting skills. Adobe does a pretty good job, but sometimes it's just wildly off. So, yeah, I don't know that it's something that, you know, I'm gonna invest in until I have a better handle at prompting for images. Frankly, I think I'm getting better background images that, uh, yeah, that I'm actually using on sites right now from Perplexity Pro. - 这很有意思。我发现，有时候 Perplexity 做得很好，有时候又不行。这可能和我的提示词技巧有关。Adobe 做得相当不错，但有时又错得离谱。所以，我不确定是否会在这上面投资，除非我能更好地掌握图像提示词。坦白说，我觉得我目前在网站上实际使用的背景图片，从 Perplexity Pro 得到的质量更好。

40:50-41:04 - Researcher: Interesting - so Perplexity Pro's image generation is actually outperforming Adobe for the backgrounds you're actually deploying on client sites right now. When Perplexity nails it versus when it doesn't, can you spot a pattern? Like certain types of images or styles it handles better? - 很有趣——所以 Perplexity Pro 的图像生成能力在为你目前部署在客户网站上的背景图方面，实际上表现优于 Adobe。 当 Perplexity 表现好或不好时，你能发现什么规律吗？比如它处理得更好的某种类型的图像或风格？

41:13-41:40 - Marcus: Yeah. I mean, it does a really great job at these kind of, like, corporate style backgrounds. I'm I'm looking for backgrounds for my pages that have to do with Google Business profiles or the core for marketing, just kind of imagery that that nails these concepts home without being you know, too distracting. Perplexity just did a really good job with that. - 是的。它是说，在处理那种“企业级风格”的背景时，它做得非常好。我正在为涉及 Google 商家资料或核心营销的页面寻找背景，那种能够精准传达这些概念且不会过于分散注意力的意象。Perplexity 在这方面做得非常好。

41:42-41:55 - Researcher: So clean, professional, concept-driven backgrounds that support the message without screaming for attention - Perplexity nails that corporate aesthetic. Is image generation included in your Perplexity Pro subscription, or is that a separate add-on you're paying for? - 所以是那种简洁、专业、概念驱动，能支持信息传递又不会喧宾夺主的背景——Perplexity 抓住了那种企业审美。 图像生成包含在你的 Perplexity Pro 订阅中吗？还是你额外付费的插件？

41:55-41:57 - Marcus: It's included. - 包含在内。

42:00-42:20 - Researcher: Nice - so you're already getting value from a feature you didn't even mention initially. That's actually a point in Perplexity's favor for renewal. Let me ask you some rating questions about Perplexity Pro specifically, since we've dug deep into it. On a scale of 1 to 5, where 1 is very dissatisfied and 5 is very satisfied - how satisfied are you with Perplexity Pro's ability to complete the tasks you give it? - 太棒了——所以你已经从一个最初甚至没提到的功能中获得了价值。这实际上是 Perplexity 续订的一个加分项。 既然我们已经深入讨论了 Perplexity Pro，让我问几个评分问题。从 1 到 5 分，1 是非常不满意，5 是非常满意——你对 Perplexity Pro 完成你交付的任务的能力满意度如何？

42:23-42:24 - Marcus: Three point five. - 3.5 分。

42:27-42:32 - Researcher: 3.5 - so solidly above average but not exceptional. What would it take to push that up to a 4 or 5? - 3.5 分——稳高于平均水平，但还不够出众。怎样才能让它达到 4 分或 5 分？

42:35-43:12 - Marcus: You know, if I could trust it to do code, then, you know, that would be awesome. But I I don't. So that, you know, is kind of a major drawback because the writing element is something I know, you know, if I was forced to go to a free tier solution I think that I could get Perplexity Pro level responses from, you know, like, ChatGPT. That that's my feeling. - 如果我能信任它写代码，那就太棒了。但我现在不信任。所以这算是一个主要的缺点，因为写作这一块，我知道如果我被迫去用免费方案，我觉得我也能从 ChatGPT 那里得到 Perplexity Pro 级别的回复。这是我的感觉。

43:15-43:33 - Researcher: Ah, that's a critical insight - you're paying $129year mostly for writing, but you feel like you could probably get similar writing quality from free ChatGPT if you had to. The code generation was supposed to be a differentiator, but the security gap killed that trust. So right now, Perplexity Pro is kind of in this middle zone where it's convenient but not essential? - 啊，这是一个关键的洞察——你每年支付 129 美元主要为了写作，但你觉得如果需要的话，从免费版 ChatGPT 也能得到类似的写作质量。代码生成本应是一个差异化优势，但安全性的缺失杀死了这种信任。 所以现在 Perplexity Pro 处于一种尴尬的中间地带：好用，但非必需？

43:35-44:09 - Marcus: That's correct. And, you know, I'm I'm gonna be testing its coding abilities you know, probably probably soon. It's not something I've got time for at the moment. But, you know, if I if if I've heard of improvements in its coding abilities, and I run some tests and I can be convinced that it's not going to generate code with massive security gaps. Then, you know, that's more points in its favor. - 没错。我也计划可能很快会测试它的编程能力。目前我还没时间。但是，如果我听说它的编程能力有所提升，并且我运行了一些测试，确信它不会生成带有重大安全漏洞的代码，那它的加分项就更多了。

44:12-44:29 - Researcher: So you're open to giving Perplexity's coding another shot if you hear it's improved - you'd just need to validate it yourself with tests and cross-checking before trusting it on client work. Next rating: How unique is Perplexity Pro compared to other similar AI tools? Scale of 1 to 5, where 1 is not unique at all and 5 is very unique. - 所以如果你听说 Perplexity 改进了，你愿意再给它一次机会——只是在将其用于客户工作之前，你需要通过测试和交叉比对进行自我验证。 下一个评分：与其他类似的 AI 工具相比，Perplexity Pro 的独特性如何？1 到 5 分，1 是完全不独特，5 是非常独特。

44:35-44:46 - Marcus: You know, I'm gonna give it a three. I don't know that it's super unique. I think that, you know, it's it's pretty good. Yeah. - 我会给 3 分。我不觉得它超级独特。我觉得它挺不错。是的。

44:48-44:55 - Researcher: A 3 - pretty good but not standing out dramatically from the pack. What would make an AI tool a 5 in uniqueness for you? What would that look like? - 3 分——挺好，但并没有从同类产品中脱颖而出。对你来说，什么样的 AI 工具能达到 5 分的独特性？那会是什么样子的？

45:08-45:49 - Marcus: That's a good question. You know, to make it unique, I think that, if an a five would have to be an AI model that has intuition as to what you really want. And, you know, if it's unclear, it's going to prompt you for clarity. You know, that's yeah. I I think that instead of making assumptions about what the user wants, if there's any kind of, you know, ambiguity, if, they would prompt you to clear that up, that would be unique. - 这是个好问题。为了让它变得独特，我认为一个 5 分的模型必须对你真正想要的东西有“直觉”。如果指令不明确，它会主动询问以求清晰。我认为，与其对用户的意图做假设，如果存在任何模糊点时它能提示你澄清，那将是非常独特的。

45:52-46:09 - Researcher: So a truly unique AI would recognize ambiguity in your prompt and ask clarifying questions before generating - rather than just making assumptions and giving you something that might be off-target. That's a collaboration model, not just a generation model. Have you seen any AI tool do that well, or is that still a gap across the board? - 所以一个真正独特的 AI 应该能识别出提示词中的模糊之处，并在生成内容之前询问澄清性问题——而不是仅仅做假设并给你一些可能偏离目标的东西。这是一种协作模式，而不只是生成模式。 你见过有哪个 AI 工具在这方面做得很好吗？还是说这目前依然是普遍存在的空白？

46:39-46:48 - Marcus: Once again, you know, I've had in the time that I've had to, experiment with different styles of prompting, You know, I I'm fully aware that the more detail I put in my prompt, the better the response is going to be. You know, sometimes, you have more time than others. To do that. But yeah. - 还是那句话，在我尝试各种提示词风格的过程中，我充分意识到我提供的细节越多，得到的回复就越好。但有时候，你没有那么多时间去做这些。

47:02-47:22 - Researcher: Right - you know you need to be detailed, but you're saying the ideal tool would catch you when you're being vague and ask follow-ups instead of just running with incomplete info. Last rating for Perplexity Pro: How satisfied are you with its value for money - its cost versus what you're getting? 1 to 5, where 1 is very dissatisfied and 5 is very satisfied with the price-to-value ratio. - 对——你知道需要详细说明，但你是说理想的工具应该在你说得含糊时察觉到，并询问后续问题，而不是带着不完整的信息直接运行。 关于 Perplexity Pro 的最后一个评分：你对其性价比（价格与所得价值的比率）的满意度如何？1 到 5 分，1 为非常不满意，5 为非常满意。

47:22-47:22 - Marcus: Three - 3 分。

47:22-47:28 - Researcher: A 3 on value for money - right in the middle. So at $129year, it's not feeling like a steal, but not a rip-off either. What price point would make it feel like a clear win for what you're currently getting from it? - 3 分——中规中矩。所以每年 129 美元，感觉既不算捡了大便宜，也不是被宰了。 什么样的价格点会让你觉得目前的体验是明确的物超所值？

47:42-48:19 - Marcus: I mean, once again, the the one twenty nine isn't really you know, the major issue. The major issue is the subscription that I pay for. I I really want it to be you know, everything. I want it to be able to do the writing. I want it to be doing the coding. I want my Pro AI service to be fully rounded. You know. And, you know, I'm going to switch to a platform that you know, I think that, uh, is gonna be the one solution for my workflow. Or at least most of my AI solution for the for my workflow. - 我再说一遍，129 美元本身并不是主要问题。主要问题是我支付的这份订阅。我真的希望它是“全能”的。我希望它能写作，也能写代码。我希望我的 Pro 级 AI 服务是全面的。我会转向一个我认为能为我的工作流提供“单一解决方案”的平台。或者至少能解决我工作流中大部分 AI 需求的方案。

48:21-48:41 - Researcher: Ah - so it's not about the dollar amount, it's about consolidation. You don't want to pay for multiple AI subscriptions - you want ONE pro tool that handles writing, coding, and images reliably so you're not juggling platforms. Right now Perplexity does writing and images well but failed you on code. If Gemini or Claude offered that full package at a similar price, you'd switch in a heartbeat? - 啊——所以这不是钱的问题，而是整合的问题。你不想为多个 AI 订阅付费——你想要一个能可靠处理写作、代码和图像的专业工具，这样你就不用像玩杂耍一样切换平台了。 目前 Perplexity 的写作和图像做得很好，但在代码上让你失望了。如果 Gemini 或 Claude 以类似的价格提供这种全能包，你会立刻切换吗？

48:43-48:44 - Marcus: I would. - 我会的。

48:51-49:06 - Researcher: Perfect - that desire for one consolidated solution is super clear. Now let's shift to tools you use frequently but aren't paying for. We've talked about Gemini through TechTrain's playground - that's one. Are there any other AI production or efficiency tools you use regularly but don't have a paid subscription for? - 太棒了——对一个整合解决方案的渴望表达得非常清楚。 现在让我们转向你经常使用但没付费的工具。我们聊过了通过 TechTrain 演练场使用的 Gemini，这是一个。还有其他你经常使用但没有付费订阅的 AI 生产力或效率工具吗？

49:20-49:26 - Marcus: Uh, just the techTrain playground. Which has, uh, you know, several different AI platforms and they're ever changing. - 就只有 TechTrain 演练场。它包含了几种不同的 AI 平台，而且一直在更新。

49:28-49:39 - Researcher: Got it - so the techTrain playground is your main unpaid tool, and it rotates which models are available. Right now you mentioned Gemini 3 Pro is in there - what other models do you have access to through it currently? - 明白了——所以 TechTrain 演练场是你主要的免费工具，它会轮换可用的模型。你刚才提到里面有 Gemini 3 Pro——目前通过它你还能用到哪些模型？

50:11-50:30 - Marcus: So, yeah, I've got Gemini. I've got the DeepSeek models. Um, I've got the Claude. There's a few different models of Claude. There are some smaller less well known AI models. I can't think of off the top of my head. There's gotta be about ten in there that that you can use ChatGPT, Claude, Gemini, those are always there. There are some others that that, uh, appear and disappear depending on what's happening with techTrain. - 是的，我有 Gemini。我有 DeepSeek 模型。我有 Claude，有几种不同型号的 Claude。还有一些不太出名的小型 AI 模型，我一下子记不起名字。里面大概有十个可以用的。ChatGPT、Claude、Gemini 是一直都在的。其他一些模型会根据 TechTrain 的情况出现或消失。

50:33-50:50 - Researcher: So you've got the major players - ChatGPT, Claude (Quad), Gemini - plus DeepSeek and some rotating smaller models. That's actually a pretty robust free toolkit. You mentioned earlier that Gemini 3 Pro is doing the best job for you right now. How are you using ChatGPT through the playground compared to Gemini? Different tasks, or do they overlap? - 所以你拥有了主要玩家——ChatGPT、Claude、Gemini——加上 DeepSeek 和一些轮换的小模型。这实际上是一个非常强大的免费工具库。 你之前提到 Gemini 3 Pro 目前表现最好。通过演练场使用 ChatGPT 和 Gemini 相比，情况如何？是任务不同，还是相互重叠？

51:05-51:14 - Marcus: Yeah. Gemini I'm using Gemini mostly for code. I was using ChatGPT. But I think that Gemini has the edge Yeah. - 是的。Gemini——我主要用 Gemini 写代码。以前用 ChatGPT，但我认为 Gemini 现在更有优势。

51:16-51:29 - Researcher: So Gemini has overtaken ChatGPT as your go-to for code generation through the playground. What about Claude - you mentioned using it to check that Perplexity code for security issues. How often are you using Claude, and what do you typically use it for? - 所以在演练场中，Gemini 已经取代 ChatGPT 成为你生成代码的首选工具了。 那 Claude 呢——你提到用它来检查 Perplexity 代码的安全性问题。你使用 Claude 的频率如何？通常用它做些什么？

51:31-52:07 - Marcus: I mean, I was using Claude to double check code. A lot of the I mean, at first, I was using it to double check the code. I was getting out of Gemini, but, you know, it didn't really improve on what I was getting from gem from, uh, from Gemini. It wasn't giving me, you know, any like, more elegant answers. I felt that that, uh, Gemini really, uh, was the way to go. - 我以前是用 Claude 来复核代码。一开始，我是用它来复核从 Gemini 得到的代码，但怎么说呢，它并没有在 Gemini 的基础上有什么提升。它没有给我更优雅的答案。我觉得 Gemini 确实是更好的选择。

52:10-52:20 - Researcher: So you tested Claude against Gemini for code quality and elegance, and Gemini won - so now you're mostly just using Gemini for code generation and not bothering to double-check it with Claude anymore? - 所以你针对代码质量和优雅度在 Claude 和 Gemini 之间做了测试，Gemini 赢了——所以现在你主要只用 Gemini 生成代码，不再费神用 Claude 复核了？

52:21-52:26 - Marcus: Uh, that's correct. I haven't double checked with Claude in a little while now. - 没错，我已经有一段时间没用 Claude 复核了。

52:28-52:34 - Researcher: Interesting shift. And DeepSeek - have you experimented with that at all, or is it just sitting there unused in the playground? - 有趣的转变。那 DeepSeek 呢——你有没有尝试过它，还是它就一直躺在演练场里没被动过？

52:58-53:14 - Marcus: No. I actually started using DeepSeek fairly early. It's not something that I use a lot now, but yeah, DeepSeek was doing a great job with writing as well. I really like the responses that I get from DeepSeek. I use it when I was applying for some contract gigs. It did a really good job with the proposals, with writing cover letters. Yeah. I I actually really enjoyed using DeepSeek. - 不。实际上我很早就开始用 DeepSeek 了。虽然现在用得不多，但是的，DeepSeek 在写作方面也做得非常棒。我真的很喜欢 DeepSeek 给出的回复。当我申请一些外包工作时，我会用它。它在写提案、写求职信方面做得非常好。是的，我其实非常喜欢用 DeepSeek。

53:17-53:26 - Researcher: So DeepSeek impressed you for proposal and cover letter writing - it sounds like you had a good experience with it but just don't need that type of writing as frequently now? - 所以 DeepSeek 的提案和求职信写作能力令你印象深刻——听起来你对它的体验很好，只是现在不太需要那种类型的写作了？

53:50-54:12 - Marcus: I think what happened is that I started using Perplexity Pro, and noticing that between the writing and the the images that it was doing more for me. It was doing everything that DeepSeek could do. At the time, DeepSeek was not generating images. So, you know, for me, it's really just about efficiency. If I can get you know, a great result from one platform as well as additional benefits in in my workflow. Then, uh, I'm gonna go with with the one that's serving me the best. - 我觉得情况是这样的：我开始用 Perplexity Pro，发现它兼顾了写作和图像，为我做了更多。它能做 DeepSeek 能做的所有事，而当时 DeepSeek 还不能生成图像。所以对我来说，这单纯是为了效率。如果我能从一个平台获得极好的结果以及工作流中的额外收益，那么我会选择那个服务最全面的。

54:15-54:29 - Researcher: Ah - so DeepSeek was great at writing, but Perplexity Pro gave you writing plus images in one place, so you consolidated to Perplexity for efficiency. That "one-stop-shop" preference keeps coming up for you. Does DeepSeek generate images now, or is it still text-only in the playground? - 啊——所以 DeepSeek 擅长写作，但 Perplexity Pro 在一个地方同时提供了写作和图像，所以为了效率你整合到了 Perplexity。这种“一站式商店”的偏好在你身上反复出现。 DeepSeek 现在能生成图像了吗？还是演练场里依然只能处理文本？

54:31-54:34 - Marcus: As far as I know, it's still text only. - 据我所知，仍然只能处理文本。

54:36-54:50 - Researcher: Got it. So right now through the techTrain playground, you're primarily using Gemini for code and occasionally dipping into the others. Have you ever paid for any of these tools outside of the playground - like a ChatGPT subscription, or Claude, or DeepSeek directly? - 明白了。所以目前通过 TechTrain 演练场，你主要用 Gemini 写代码，偶尔尝试一下其他的。 除了演练场，你有没有为这些工具付过费——比如直接订阅过 ChatGPT、Claude 或 DeepSeek？

54:54-54:55 - Marcus: I have not. No. - 没有。

54:58-55:03 - Researcher: Never paid for any of them. And you've had free access to these pro models through techTrain for how long now? - 从来没付过。那你通过 TechTrain 免费使用这些专业模型已经多久了？

55:05-55:06 - Marcus: About a year. - 大约一年。

55:09-55:21 - Researcher: A full year of free access to ChatGPT, Claude, Gemini, DeepSeek - that's a pretty sweet perk for doing AI training work. If techTrain suddenly stopped providing that playground access tomorrow, which of those tools would you actually consider paying for? - 整整一年免费使用 ChatGPT、Claude、Gemini、DeepSeek——这对于做 AI 训练工作来说是个相当不错的福利。 如果 TechTrain 明天突然停止提供演练场访问权限，这些工具中你真正会考虑付费订阅的是哪一个？

55:22-55:23 - Marcus: Gemini. - Gemini。

55:26-55:28 - Researcher: Just Gemini - because of the code generation quality? - 只是 Gemini——是因为它的代码生成质量吗？

55:30-55:34 - Marcus: Absolutely. And it does great with writing as well. - 绝对是。而且它在写作方面也做得很出色。

55:36-55:41 - Researcher: So Gemini handles both your coding and writing needs well. If you had to pay for Gemini directly, - 所以 Gemini 能很好地满足你的编程和写作需求。如果你必须直接为 Gemini 付费，

55:43-55:46 - Marcus: For the most part, yeah. Mostly code. - 很大程度上是的。主要是代码。

55:48-55:55 - Researcher: Mostly code, some writing. If you had to pay for Gemini directly, what's the maximum monthly amount you'd be willing to spend on it? - 主要代码，部分写作。如果你必须直接付钱给 Gemini，你愿意承受的最高月费是多少？

55:59-56:02 - Marcus: I mean, probably, like, twenty bucks. - 也就是，可能 20 美金左右。

56:03-56:16 - Researcher: $20month feels right for what you're getting from Gemini. That's actually less than what you're paying for Perplexity Pro annually ($129year is about $10.75month). What would Gemini need to add or improve to make you willing to pay more than $20month? - 每月 20 美元感觉符合你从 Gemini 获得的价值。这实际上比你每年支付给 Perplexity Pro 的费用要低（129 美元/年约合 10.75 美元/月）。 Gemini 需要增加或改进什么，才会让你愿意支付超过每月 20 美元的费用？

56:25-56:39 - Marcus: You know? If it would be the all in one answer for my AI needs. Then I would probably be willing to you know, I would be willing to shell out more money. - 如果它能成为我 AI 需求的“全能答案”。那我可能愿意……我愿意掏更多的钱。

56:43-56:52 - Researcher: So back to that full package - if Gemini added really solid image generation to its already strong code and writing, you'd pay more than $20month for that consolidated solution? - 回到“全能包”的概念——如果 Gemini 在其已经很强的代码和写作基础上加入真正扎实的图像生成，你会愿意为这个整合方案支付超过每月 20 美元的费用吗？

56:54-57:02 - Marcus: Possibly. I definitely would pay the one twenty nine a year I'm paying for Perplexity Pro. - 有可能。我绝对愿意支付我现在给 Perplexity Pro 的每年 129 美元。

57:04-57:16 - Researcher: So $129year ($10.75month) you'd definitely pay for a full-package Gemini. Maybe more, but that's the floor. Right now Gemini does have image generation - have you tried it through the playground, or is it not available in the techTrain version? - 所以 129 美元/年（约 10.75 美元/月）你肯定愿意为一个全能版的 Gemini 买单。可能会更多，但那是底线。 现在 Gemini 确实有图像生成功能——你在演练场里试过吗？还是说 TechTrain 版本里没有这个功能？

57:22-57:40 - Marcus: You know, it's inconsistent. It's not it doesn't give me the results that I'm looking for. I've gotta tinker with it more. Which you know, I kinda don't wanna do. - 怎么说呢，它不太稳定。它给不出我想要的结果。我得针对它进行更多推敲，但其实我不太想这么做。

57:43-58:00 - Researcher: Ah - so Gemini has image generation in the playground, but the quality isn't there yet and you don't want to spend time wrestling with prompts to make it work. You want it to just deliver good results like Perplexity does for images. What specifically isn't working - is it understanding your prompts, or the visual quality of what it generates? - 啊——所以演练场里的 Gemini 有图像生成，但质量还没达标，而且你不想花时间去死磕提示词让它生效。你希望它像 Perplexity 的图像功能一样，直接就能出好结果。 具体是哪里不行——是它理解不了你的提示词，还是它生成的视觉质量不行？

58:10-58:26 - Marcus: Well, sometimes it acts like it's generating an image, and then it just doesn't generate that image. So that is really frustrating. You know, it'll it'll almost go into some kind of weird role playing mode. Where it wants you to pretend that it's generated an image that it actually hasn't generated. So, yeah, that that's a major drawback. - 这个，有时候它表现得好像正在生成图像，然后它就是不出图。那真的让人很沮丧。它甚至会进入某种奇怪的角色扮演模式，它想让你假装它已经生成了一张其实并没生成的图片。所以，是的，那是一个主要的缺陷。

58:28-58:36 - Researcher: Oh wow - that's bizarre and frustrating. It pretends it made an image and describes it to you instead of actually generating it? That would drive me - 噢哇——那太离谱且让人抓狂了。它假装自己画了图，然后向你描述，而不是真的生成出来？这也得把我

58:36-58:37 - Marcus: Yep. - 没错。

58:39-58:58 - Researcher: That would drive me crazy too. So until Gemini fixes that image generation weirdness, you're stuck using Perplexity Pro for images even though you'd rather consolidate everything into Gemini. Let me ask this differently: on a scale of 0-10, how willing would you be to pay for the techTrain playground tools if you lost free access - with 0 being not at all willing and 10 being extremely willing? - 那也会把我逼疯的。所以除非 Gemini 修复图像生成的怪问题，否则你只能被迫继续为了图像而使用 Perplexity Pro，即便你更想把一切都整合到 Gemini 里。 让我换个问法：如果失去免费访问权限，你为 TechTrain 演练场工具付费的意愿在 0 到 10 分之间是多少（0 是完全不愿意，10 是极其愿意）？

59:04-59:14 - Marcus: You know, if I had API if I had API access, to those models, on techTrain playgrounds, Yeah. I'd probably be willing to pay quite a bit for that. - 如果我有 API 访问权限，能在 TechTrain 演练场使用那些模型的 API。那我会愿意为此付不少钱。

59:17-59:28 - Researcher: API access changes everything for you - that's the second time you've mentioned it. Walk me through what having API access to these models would unlock for your work that you can't do now through the regular playground interface? - API 访问权限改变了你的一切——这是你第二次提到它了。跟我说说，如果拥有这些模型的 API 访问权限，能为你的工作开启哪些目前常规演练场界面做不到的事情？

59:45-59:53 - Marcus: Well, I mean, I would be able to build robust platforms. I'd be able to have you know, data pump directly in and to be able to use the AI model to have have that data automatically processed So that would be awesome. - 这样我就能构建健壮的平台。我能让数据直接输入，并利用 AI 模型自动处理这些数据。那太棒了。

59:57-01:00:08 - Researcher: So API access would let you automate those big web migration contracts - pumping scraped data from multiple sites directly into the AI to process and standardize it, instead of manually feeding it through a chat interface? - 所以 API 权限可以让你自动化那些大型网页迁移合同——将从多个站点抓取的数据直接泵入 AI 进行处理和标准化，而不是手动通过聊天界面喂给它？

01:00:09-01:00:10 - Marcus: Yep. - 没错。

01:00:12-01:00:22 - Researcher: That would be a game-changer for those PE home services contracts. If techTrain offered API access to their playground models as a paid tier, what would you be willing to pay monthly for that? - 这对于那些私募支持的家庭服务合同来说将是游戏规则的改变者。如果 TechTrain 以付费层级的形式提供演练场模型的 API 访问，你每月愿意为此支付多少钱？

01:00:37-01:01:05 - Marcus: That's a dangerous question. I don't know. You know? I would have to, they would have to let me know what they would wanna charge. I can imagine that, you know, if it were generating income for me, and I would probably be willing to part with a percentage of that income. Like, if I were making a few thousand dollars per month from having that access then I would be happy to pay up to, like, ninety nine dollars a month. - 这是个危险的问题。我不知道。他们得告诉我他们想怎么收。但我能想象，如果它能为我产生收入，我可能愿意分享一部分收入。比如，如果因为拥有这个权限，我每月能赚几千美元，那我非常乐意支付最高每月 99 美元。

01:01:07-01:01:27 - Researcher: That makes sense - you'd tie the cost to the revenue it generates. So if API access was bringing in a few thousand a month from those contracts, $99month would feel reasonable. Let me zoom out a bit. When you're deciding whether to pay for any AI productivity tool - whether it's Perplexity, Gemini, whatever - what are all the factors you weigh in that decision? Walk me through everything that matters to you. - 有道理——你会将成本与它产生的收益挂钩。如果 API 访问能让那些合同每月带来几千美元，99 美元/月的价格就非常合理。 让我跳出来看。当你决定是否为任何 AI 生产力工具付钱时——不管是 Perplexity、Gemini 还是别的——你会权衡哪些因素？请详细说明对你重要的每一件事。

01:01:31-01:02:00 - Marcus: Like I was saying before, if it's going to be an all in one solution, uh, if the writing works for my uses, if the code is rock solid and secure, If the image generation does what I wanted to do and produces high quality results, then, yeah, that's that's what I'm looking for. - 就像我之前说的，如果它能成为一站式解决方案，它的写作能满足我的用途，它的代码坚实且安全，它的图像生成能按我要求做出高质量的结果，那么，是的，那就是我想要的。

01:02:02-01:02:13 - Researcher: So the all-in-one factor is huge - writing, code, images all working well in one place. Beyond those three capabilities working, what else matters? Like, does speed matter? Conversation memory? Customer support? Ease of use? - 所以“全能”因素非常关键——写作、代码、图像在一个地方都能好用。除了这三项能力，还有什么重要的吗？比如速度？对话记忆？客户支持？易用性？

01:02:18-01:02:56 - Marcus: Customer support, not so much. Ease of use absolutely. I like the collaboration model. I like, uh, if a model would get a sense of what I'm looking for. I mean, this is kind of what I do. When I'm using these models. I like to give it as much background as I can. And, you know, with that background, I would like the model to intuit you know, what I what I need without having to give it super detailed information for each prompt. - 客户支持倒没那么重要。易用性绝对重要。我喜欢那种协作模式，我喜欢如果一个模型能感觉到我想要什么。我是说，这就是我平时使用这些模型的方式。我喜欢给它尽可能多的背景信息，在有背景的情况下，我希望模型能凭直觉判断出我需要什么，而不需要我在每个提示词里都给出超级详细的信息。

01:02:58-01:03:14 - Researcher: So you want the AI to learn you - to remember context and background so you don't have to re-explain your needs every single time. That conversation memory and intuition about your work style is really important. Does that mean you prefer tools that let you build up a long conversation thread versus ones that reset frequently? - 所以你希望人工智能能够了解你——记住语境和背景，这样你就不必每次都重新解释你的需求。那种对话记忆以及对你工作风格的直觉非常重要。 这是否意味着，与那些频繁重置对话的工具相比，你更倾向于可以建立长对话链的工具？

01:03:15-01:03:18 - Marcus: Absolutely. Without question. Yes. - 绝对是的。毫无疑问。是的。

01:03:20-01:03:30 - Researcher: Strong conversation memory is critical. What else factors into your decision - like, does it matter whether the tool has a mobile app? Or integrations with other software you use? Or how often it gets updated? - 强大的对话记忆至关重要。还有哪些因素会影响你的决定——比如，该工具是否有移动端应用重要吗？或者是否能与你使用的其他软件集成？或者更新频率如何？

01:03:32-01:04:11 - Marcus: All of that matters. You know, I do a lot on mobile. So, yeah, the mobile app is gonna it gonna be huge. Integrating with other software is super helpful. As well. And, of course, I want it to be updated. This is this is a a kind of a what's the word I'm looking for? You know, the AI wars, they're gonna have to be updated. Constantly to keep an edge. People that use AI models wanna be using the best one. So, sure, updating is important. - 所有这些都很重要。你知道，我经常在手机上处理事情。所以，是的，移动端应用会非常关键。与其他软件集成也超级有帮助。当然，我希望它能保持更新。这是一个——怎么说呢，我想找什么词来着？你知道，AI 战争，它们必须不断更新才能保持优势。使用 AI 模型的人都想用最好的那个。所以，当然，更新很重要。

01:04:13-01:04:29 - Researcher: So mobile app, integrations, frequent updates all matter. You've given me: all-in-one solution (writingcodeimages), ease of use, strong conversation memory, mobile app, integrations, regular updates, and earlier you mentioned security for code generation. Anything else that goes into the decision, or is that the full list? - 所以移动端应用、集成、频繁更新都很重要。你已经给了我这些信息：全能型解决方案（写作/代码/图片）、易用性、强大的对话记忆、移动端应用、集成、定期更新，以及早些时候你提到的代码生成的安全性。 还有其他因素会影响你的决定吗，还是说这就是完整的清单了？

01:04:31-01:04:33 - Marcus: That's the list. - 就这些了。

01:04:35-01:04:43 - Researcher: Perfect. Now the hard part - from most important to least important, how would you rank those factors? What's the #1 thing that matters most when deciding to pay? - 太棒了。现在是困难的部分——从最重要到最不重要，你会如何对这些因素进行排序？在决定付费时，最重要的一点是什么？

01:04:50-01:05:14 - Marcus: The number one thing that matters most when deciding to pay, uh, it's gonna be the coding. I'm gonna need to have an AI model that produces elegant efficient, secure code, At the end of the day, that is the most important. - 在决定付费时最重要的一点，呃，会是编程。我需要一个能生成简洁、高效、安全代码的 AI 模型。归根结底，这是最重要的。

01:05:16-01:05:19 - Researcher: Code quality is #1 - elegant, efficient, and secure. What's #2? - 代码质量是第一位——简洁、高效且安全。第二位是什么？

01:05:24-01:05:33 - Marcus: Everything else is number two. Writing, uh, image generation, Uh, Yeah. Everything else is secondary. - 其他所有东西都是第二位。写作、呃，图像生成，呃，是的。其他一切都是次要的。

01:05:36-01:05:48 - Researcher: So there's code quality at the top by itself, and then writing, images, conversation memory, mobile app, integrations, updates - all those are kind of tied as secondary. None of those secondary factors stand out as more important than the others? - 所以代码质量单独排在最顶层，然后写作、图像、对话记忆、移动端应用、集成、更新——所有这些都被并列归为次要。在这些次要因素中，没有哪一个比其他的更突出吗？

01:05:53-01:06:04 - Marcus: Well, okay. When you put it like that, I'm gonna have to say that updates integrations. Conversation memory. I'll put that on a second tier. - 哦，好吧。你这么一说的话，我得说更新和集成。对话记忆。我会把这些放在第二梯队。

01:06:06-01:06:17 - Researcher: Okay, so tier 1 is code quality. Tier 2 is updates, integrations, and conversation memory. That leaves writing, images, mobile app, and ease of use - are those all tier 3, or do some of those matter more than others? - 好的，所以第一梯队是代码质量。第二梯队是更新、集成和对话记忆。剩下的有写作、图像、移动端应用和易用性——这些都是第三梯队，还是其中一些比其他的更重要？

01:06:24-01:06:30 - Marcus: No, I'm gonna I'm gonna play put ease of use bump that up to tier two. - 不，我要把易用性提升到第二梯队。

01:06:32-01:06:41 - Researcher: Got it - so tier 2 is updates, integrations, conversation memory, AND ease of use. That leaves writing quality, image generation, and mobile app. How do those rank for you? - 明白了——所以第二梯队是更新、集成、对话记忆和易用性。剩下的就是写作质量、图像生成和移动端应用。你对这些怎么排序？

01:06:58-01:07:15 - Marcus: You know, I guess, really, they've they've all got mobile apps. Writing quality, image generation, Wow. Yeah. I'm gonna go with image generation one, writing quality two, mobile app three. - 你知道，我想，其实它们都有移动端应用。写作质量、图像生成，哇。好吧。我会把图像生成排第一，写作质量排第二，移动端应用排第三。

01:07:17-01:07:33 - Researcher: Perfect, so your full ranking is: 1. Code quality (secure, elegant, efficient) 2. Updates, integrations, conversation memory, ease of use 3. Image generation 4. Writing quality 5. Mobile app Shifting gears - where do you typically hear about new AI products? What are your main channels for discovering new tools? - 太好了，所以你的完整排名是： 1. 代码质量（安全、简洁、高效） 2. 更新、集成、对话记忆、易用性 3. 图像生成 4. 写作质量 5. 移动端应用 换个话题——你通常从哪里听说新的 AI 产品？你发现新工具的主要渠道是什么？

01:07:42-01:08:04 - Marcus: YouTube, I you know, I watch a lot of the AI stuff on YouTube. I see lots of ads on I take a look on Instagram, Facebook, and just general news articles. Tech tech news. - YouTube，你知道，我在 YouTube 上看了很多关于 AI 的内容。我在 Instagram、Facebook 上看到很多广告，还有就是普通的新闻文章，科技新闻。

01:08:06-01:08:14 - Researcher: So YouTube AI content, ads on Instagram and Facebook, and tech news articles. Of those channels, which one has actually led you to try a new AI tool most often? - 所以是 YouTube 的 AI 内容、Instagram 和 Facebook 上的广告，以及科技新闻文章。在这些渠道中，哪一个最常促使你实际去尝试一个新的 AI 工具？

01:08:31-01:08:41 - Marcus: No, probably Instagram. And it seems that whenever I see a tool on Instagram, I'll check it out just because, uh, I'm interested. So yeah. You know, I've used I've tried a bunch of SEO tools because of Instagram, like AutoSEO. I guess I could have mentioned that earlier. I totally forgot about - 不，可能还是 Instagram。似乎每当我在 Instagram 上看到一个工具时，我都会去看看，纯粹是因为我感兴趣。所以是的。你知道，因为 Instagram，我试过一堆 SEO（搜索引擎优化）工具，比如 AutoSEO。我想我刚才应该提到的，我完全忘了。

01:08:45-01:08:55 - Researcher: No worries! So Instagram ads are actually converting you to trials. When you see an AI tool advertised on Instagram, what makes you decide to actually click and try it versus scroll past? - 没关系！所以 Instagram 广告确实能转化你进行试用。当你在 Instagram 上看到 AI 工具的广告时，是什么让你决定实际点击并试用，而不是直接刷过去？

01:09:06-01:09:30 - Marcus: If it is a solution that I can use, like local SEO or just automating SEO in general. You know, that's something that I would love to have a robust AI solution for a full service automated SEO You know, a lot of places seem to, you know, they say they've got it, but I've yet to see it actually work. - 如果它是一个我可以用得上的解决方案，比如本地 SEO，或者只是通用的自动化 SEO。你知道，那是我非常希望能有一个强大 AI 解决方案的东西——全服务的自动化 SEO。你知道，很多地方似乎都声称他们有，但我还没见过哪个真的管用。

01:09:32-01:09:47 - Researcher: So you're actively looking for solutions to specific pain points - like automated SEO - and Instagram is where those tools are finding you. When those tools promise "full service automated SEO" but don't deliver, what's usually the gap between what they promise and what they actually do? - 所以你一直在积极寻找针对特定痛点的解决方案——比如自动化 SEO——而 Instagram 正是这些工具找到你的地方。当那些工具承诺“全服务自动化 SEO”却没能兑现时，它们的承诺与实际效果之间的差距通常在哪里？

01:10:00-01:10:51 - Marcus: No, when they generate blog posts, I've been noticing I mean, I haven't tried I haven't tested one of these out in a couple of months, but the last time I did, you know, the promise was that you just simply plug it in, and it's going to automatically, you know, fix your SEO issues as well as generate blog articles to, you know, help out your SEO. And in my experience, what happened is that if I just let it go, by, you know, the fourth or fifth blog article. It has left web development and design, and now it's doing something like carpentry design. Or something totally unrelated So, uh, yeah, there's, uh, definitely some quality issues with with a lot of those AI solutions. - 不，在它们生成博客文章时，我注意到——我是说，我已经有两个月没试过这些东西了，但上次我试的时候，它们的承诺是你只需简单地接入，它就会自动修复你的 SEO 问题，并生成博客文章来提升你的 SEO。根据我的经验，实际发生的情况是，如果我就那样让它跑下去，到了第四或第五篇博客文章时，它写的就已经不是网页开发和设计了，现在它在写木工设计之类的内容。或者是完全无关的东西。所以，是的，很多这类 AI 解决方案确实存在一些质量问题。

01:10:54-01:11:12 - Researcher: That's wild - it just drifts completely off topic after a few posts. So the conversation memory and context understanding we talked about earlier is exactly what's missing from those SEO tools. Last question - do you have a fixed budget in your head for AI spending overall? Like, "I'm willing to spend X per month total on AI tools," or is it more case-by-case? - 那太离谱了——写了几篇后主题就完全飘了。所以我们之前谈到的对话记忆和背景理解，正是这些 SEO 工具所缺失的。 最后一个问题——关于 AI 的整体支出，你心里有一个固定预算吗？比如“我愿意每个月总共花多少钱在 AI 工具上”，还是说更多是视情况而定？

01:11:36-01:11:43 - Marcus: It's really case by case. Uh, you know, if I have a scenario where I am going to generate a lot of profit, by using a certain AI tool, and I've got no problem. You know, paying. So, you know, I I've got no problem paying premium prices for premium results. - 真的要视情况而定。呃，你知道，如果我有某种情况，使用特定的 AI 工具能产生大量利润，那我就没有任何问题。你知道，愿意付钱。所以，我完全不介意为了优质的结果支付优质的价格。

01:11:45-01:11:57 - Researcher: So it's ROI-driven - if the tool generates profit, you'll pay what makes sense relative to that return. There's no hard ceiling like "I'll never spend more than $200month total" - it's really about whether the value is there? - 所以是投资回报率（ROI）驱动的——如果工具能产生利润，你就会支付相对于该回报而言合理的费用。没有硬性的上限，比如“我每个月的总支出绝不超过 200 美元”——关键在于价值是否在那里？

01:12:01-01:12:02 - Marcus: That's correct. - 没错。

01:12:05-01:12:26 - Researcher: Perfect. That's really helpful to understand how you think about AI spending - it's all tied to the business value it creates for you. We've covered a ton of ground today - from your daily workflow with clients, to how you use different AI tools, the security issues you've hit, what would make you switch tools, and how you decide what to pay for. Really appreciate you walking me through all of this in such detail. Thank you! - 太棒了。了解你如何看待 AI 支出非常有帮助——这完全取决于它为你创造的业务价值。 我们今天讨论了很多内容——从你与客户的日常工作流程，到你如何使用不同的 AI 工具，你遇到的安全性问题，什么会促使你更换工具，以及你如何决定为哪些东西付费。非常感谢你能如此详细地向我介绍这一切。谢谢！

01:12:31-01:12:32 - Marcus: You're welcome. - 不客气。

01:12:34-01:12:38 - Researcher: Take care, and best of luck with those web migration contracts! - 保重，祝你在那些网页迁移合同上好运！

01:12:39-01:12:40 - Marcus: Thanks. - 谢谢。

01:12:44-01:12:46 - Researcher: Thanks again - have a great rest of your day! - 再次感谢——祝你今天愉快！

01:12:47-01:12:49 - Marcus: You too. - 你也是。

01:12:51-01:12:52 - Researcher: Bye! - 再见！

01:12:56-01:12:57 - Marcus: Okay. - 好的。

01:12:57-01:12:58 - Researcher: 👋 - 👋