WEBVTT 00:00:00.000 --> 00:00:05.580 We all fear it. That sudden jarring case of AI 00:00:05.580 --> 00:00:07.799 amnesia. Oh, it is the absolute worst. You switch 00:00:07.799 --> 00:00:10.240 to a new tool. And suddenly your brand new assistant 00:00:10.240 --> 00:00:13.039 has zero clue who you are. Right. It's like you 00:00:13.039 --> 00:00:15.300 never even existed. It doesn't actually have 00:00:15.300 --> 00:00:17.300 to be this way. You can transplant a digital 00:00:17.300 --> 00:00:20.399 brain in under five minutes. Welcome to our deep 00:00:20.399 --> 00:00:22.839 dive. Yeah, I am really excited about this one. 00:00:22.980 --> 00:00:25.140 I am glad you're joining us today. We are looking 00:00:25.140 --> 00:00:28.300 at a highly specific migration guide. We are 00:00:28.300 --> 00:00:31.920 dissecting the complete move from ChatGPT over 00:00:31.920 --> 00:00:34.090 to Claude. And we are going to cover it. everything 00:00:34.090 --> 00:00:36.890 you need to make this seamless. We will explore 00:00:36.890 --> 00:00:39.350 why everyone is making this jump right now. We'll 00:00:39.350 --> 00:00:41.270 show you how to properly prep your data. You 00:00:41.270 --> 00:00:44.170 will get the exact master export prompt you need. 00:00:44.289 --> 00:00:46.369 We'll also explain how Claude's memory actually 00:00:46.369 --> 00:00:48.869 works under the hood. And crucially, we will 00:00:48.869 --> 00:00:51.810 highlight the major traps you absolutely must 00:00:51.810 --> 00:00:53.890 avoid. Yeah, those traps catch a lot of people. 00:00:54.030 --> 00:00:57.270 Let's unpack this massive shift in the AI landscape. 00:00:57.570 --> 00:01:00.609 Beat. Because for a long time, ChatGPT was essentially 00:01:00.609 --> 00:01:03.740 the only real option in town. It really was. 00:01:04.120 --> 00:01:06.219 If you wanted a smart digital assistant, you 00:01:06.219 --> 00:01:08.540 went there. You built your entire workflow around 00:01:08.540 --> 00:01:10.840 it. Exactly. It was the absolute gold standard 00:01:10.840 --> 00:01:13.439 for everyone. But things have fundamentally changed 00:01:13.439 --> 00:01:16.219 recently. Very much so. Power users are finding 00:01:16.219 --> 00:01:19.040 a new favorite. They are migrating their daily 00:01:19.040 --> 00:01:21.579 operations over to Claude. Which is built by 00:01:21.579 --> 00:01:24.420 a company called Anthropic. Right. And the most 00:01:24.420 --> 00:01:27.000 common feedback I hear, it just feels more like 00:01:27.000 --> 00:01:30.180 a person. That's the primary draw right now. 00:01:30.349 --> 00:01:33.590 It feels significantly less like a robot. The 00:01:33.590 --> 00:01:36.810 Claude 3 .5 Sonnet model is particularly incredible. 00:01:36.890 --> 00:01:39.629 Oh, it's amazing. These newer models follow highly 00:01:39.629 --> 00:01:42.430 complex multi -step instructions beautifully. 00:01:42.689 --> 00:01:45.090 But more importantly, they write in such a natural 00:01:45.090 --> 00:01:47.909 tone. Yeah, a very human -like tone. Yeah. Chat 00:01:47.909 --> 00:01:51.109 GPT is still fantastic for many specific tasks. 00:01:51.290 --> 00:01:54.569 Absolutely. It is excellent for strict mathematical 00:01:54.569 --> 00:01:57.790 logic. It is wonderful for debugging highly complex 00:01:57.790 --> 00:02:01.709 code. It generates images. seamlessly with Deli. 00:02:01.890 --> 00:02:04.549 And, of course, it has that incredible advanced 00:02:04.549 --> 00:02:07.030 voice mode. Right, and it remains a total powerhouse 00:02:07.030 --> 00:02:09.650 for those things. But Claude is consistently 00:02:09.650 --> 00:02:12.270 winning the creative work category. It just sounds 00:02:12.270 --> 00:02:15.889 so much less AI -ish. It rarely uses those telltale 00:02:15.889 --> 00:02:20.110 words, right, like delve? Yes, or leverage, or 00:02:20.110 --> 00:02:23.530 testament. You almost never see those. That vocabulary 00:02:23.530 --> 00:02:26.210 difference is surprisingly huge for your daily 00:02:26.210 --> 00:02:28.430 workflow. It saves you so much editing time. 00:02:28.680 --> 00:02:31.139 But we also have to talk about the context window. 00:02:31.659 --> 00:02:34.889 Oh, yeah. That is a critical architectural difference. 00:02:35.229 --> 00:02:37.349 The context window is essentially its short -term 00:02:37.349 --> 00:02:40.169 working memory. It's how much text the AI can 00:02:40.169 --> 00:02:42.949 actively hold at once. Claude handles up to 200 00:02:42.949 --> 00:02:45.550 ,000 words easily. Which is just a massive amount 00:02:45.550 --> 00:02:47.870 of information. It completely changes how you 00:02:47.870 --> 00:02:50.210 work with long documents. You are no longer just 00:02:50.210 --> 00:02:52.389 feeding it tiny snippets. Right. You can drop 00:02:52.389 --> 00:02:55.250 in an entire code base. Or you can upload three 00:02:55.250 --> 00:02:58.090 different annual financial reports. And it synthesizes 00:02:58.090 --> 00:03:00.129 all of it perfectly without losing the plot. 00:03:00.349 --> 00:03:02.729 Then there is a transformative feature called 00:03:02.729 --> 00:03:04.930 Artifacts. Artifacts essentially gives Cloud 00:03:04.930 --> 00:03:07.449 a dedicated workspace. Right next to your chat 00:03:07.449 --> 00:03:09.770 window. Developers are absolutely loving it. 00:03:10.009 --> 00:03:12.250 If you ask it to code a Python dashboard, it 00:03:12.250 --> 00:03:15.189 doesn't just spit out text. No, it actually renders 00:03:15.189 --> 00:03:17.789 a functional interactive preview. It builds it 00:03:17.789 --> 00:03:20.150 right on the side of your screen. It is a huge 00:03:20.150 --> 00:03:23.870 win for creators. But here is the profound friction 00:03:23.870 --> 00:03:26.310 for you. The switching cost. The switching cost 00:03:26.310 --> 00:03:31.379 is very real. If you have used ChatGPT for a 00:03:31.379 --> 00:03:34.879 year, it deeply knows you. It knows your job 00:03:34.879 --> 00:03:37.479 title and your family dynamics. It knows your 00:03:37.479 --> 00:03:40.699 absolute favorite coding language. It knows exactly 00:03:40.699 --> 00:03:43.020 how you like your emails to sound. That is what 00:03:43.020 --> 00:03:45.479 we call its internal memory. It's a collection 00:03:45.479 --> 00:03:48.000 of small facts it picked up over time. The fear 00:03:48.000 --> 00:03:50.240 of losing this memory keeps people completely 00:03:50.240 --> 00:03:52.620 stuck. It really does. It was like moving into 00:03:52.620 --> 00:03:55.560 a beautiful, spacious new house. But leaving 00:03:55.560 --> 00:03:57.919 all your customized furniture behind. Exactly. 00:03:58.039 --> 00:04:00.419 You have to start decorating from scratch. That's 00:04:00.419 --> 00:04:03.479 exhausting. Luckily, Anthropic recognized this 00:04:03.479 --> 00:04:06.000 massive hurdle recently. They built a dedicated 00:04:06.000 --> 00:04:08.659 tool to help bring those memories over. The digital 00:04:08.659 --> 00:04:11.099 move is now easier than ever. You do not lose 00:04:11.099 --> 00:04:13.460 your custom instructions or your saved memories. 00:04:13.879 --> 00:04:15.780 No, you can keep all of it. So before we get 00:04:15.780 --> 00:04:17.980 into the migration itself, I have a probing question. 00:04:18.519 --> 00:04:21.240 Why should someone stick with chat GPT instead 00:04:21.240 --> 00:04:23.920 of moving? Well, you should stay if you rely 00:04:23.920 --> 00:04:27.279 heavily on voice mode or daily image generation. 00:04:27.860 --> 00:04:30.500 Stay if you rely heavily on voice mode or daily 00:04:30.500 --> 00:04:32.779 image generation. That makes perfect sense. Yeah, 00:04:32.800 --> 00:04:34.860 it's a clear dividing line. Let's move into the 00:04:34.860 --> 00:04:37.160 preparation phase. This is where the real work 00:04:37.160 --> 00:04:40.339 begins. That fear of losing your digital setup 00:04:40.339 --> 00:04:43.180 makes people hesitate. But preparing the data 00:04:43.180 --> 00:04:45.480 is surprisingly straightforward. You just have 00:04:45.480 --> 00:04:47.829 to know which switches to flip. And this is where 00:04:47.829 --> 00:04:50.129 people make their first huge mistake. If you 00:04:50.129 --> 00:04:53.370 just open a chat and ask for an export, it fails. 00:04:53.670 --> 00:04:56.189 Completely fails. You might only get 20 % of 00:04:56.189 --> 00:04:58.649 your data. You miss almost all of your nuanced 00:04:58.649 --> 00:05:00.829 information. You really have to force the system 00:05:00.829 --> 00:05:02.970 to look deeper. I have to offer a vulnerable 00:05:02.970 --> 00:05:06.459 admission here. Oh. I still wrestle with remembering 00:05:06.459 --> 00:05:09.339 to check my settings before exporting. I have 00:05:09.339 --> 00:05:12.160 definitely lost project contexts that way. Oh, 00:05:12.259 --> 00:05:13.959 man. Yeah, it happens to the absolute best of 00:05:13.959 --> 00:05:16.699 us. It is incredibly frustrating to realize your 00:05:16.699 --> 00:05:19.779 data is gone because you rushed. You have to 00:05:19.779 --> 00:05:23.040 explicitly force chat GPT to look into every 00:05:23.040 --> 00:05:25.540 corner. You need to open its entire brain before 00:05:25.540 --> 00:05:28.740 you extract anything. So instead of just clicking 00:05:28.740 --> 00:05:31.819 wildly. Be strategic. When you dive into your 00:05:31.819 --> 00:05:34.779 backend settings, look for personalization preferences. 00:05:35.339 --> 00:05:37.980 You need to ensure the AI has permission to view 00:05:37.980 --> 00:05:40.959 its own memories. You will see three main toggles 00:05:40.959 --> 00:05:43.259 in that backend menu. Make sure they're all turned 00:05:43.259 --> 00:05:45.420 on. The first one is simply called memory. This 00:05:45.420 --> 00:05:47.300 is the database of facts about you. The second 00:05:47.300 --> 00:05:50.279 toggle is personalization. That helps the AI 00:05:50.279 --> 00:05:52.839 adapt to your specific writing style. The third 00:05:52.839 --> 00:05:55.639 crucial one is chat history and training. Now, 00:05:55.860 --> 00:05:58.060 you might normally keep this turned off for privacy. 00:05:58.259 --> 00:06:00.800 Many power users do. But you need it turned on 00:06:00.800 --> 00:06:03.040 briefly right now. It allows the system to pull 00:06:03.040 --> 00:06:05.439 actively from your complete history. Next, we 00:06:05.439 --> 00:06:07.920 have to talk about the specific AI model you 00:06:07.920 --> 00:06:11.560 use. Yes. ChadGPT offers standard models like 00:06:11.560 --> 00:06:14.540 4 .0 and newer thinking models. For this export, 00:06:14.839 --> 00:06:16.910 you absolutely must use a thinking model. You 00:06:16.910 --> 00:06:20.629 should use something like O1 or O3 Mini. This 00:06:20.629 --> 00:06:23.149 is the absolute secret sauce for a successful 00:06:23.149 --> 00:06:26.110 transfer. Standard models are just too lazy for 00:06:26.110 --> 00:06:29.069 this kind of exhaustive task. If you ask a standard 00:06:29.069 --> 00:06:31.750 model for your memories, it rushes. It might 00:06:31.750 --> 00:06:34.129 give you the last five superficial things it 00:06:34.129 --> 00:06:36.550 learned. But a thinking model actually takes 00:06:36.550 --> 00:06:39.430 its time. It iterates. It thinks much longer. 00:06:39.949 --> 00:06:42.490 It searches its hidden database thoroughly. In 00:06:42.490 --> 00:06:45.189 extensive testing, a thinking model pull three 00:06:45.189 --> 00:06:48.000 times more data. It grabs everything you actually 00:06:48.000 --> 00:06:50.939 need for a complete profile. I've seen standard 00:06:50.939 --> 00:06:54.899 models just return garbage data. So why avoid 00:06:54.899 --> 00:06:58.139 standard models for this specific export? Standard 00:06:58.139 --> 00:07:01.079 models are lazy. Thinking models search the internal 00:07:01.079 --> 00:07:03.060 database much more thoroughly. Standard models 00:07:03.060 --> 00:07:05.540 are lazy. Thinking models search the internal 00:07:05.540 --> 00:07:07.800 database much more thoroughly. Got it. So now 00:07:07.800 --> 00:07:09.939 your settings are fully prepared. You have the 00:07:09.939 --> 00:07:12.699 right model selected. But you need a very specific 00:07:12.699 --> 00:07:15.019 structured prompt. You cannot just casually say, 00:07:15.139 --> 00:07:17.699 tell me what you know about me. That will just 00:07:17.699 --> 00:07:20.339 give you a messy paragraph. You need a prompt 00:07:20.339 --> 00:07:23.199 that forces a highly clean output. You want a 00:07:23.199 --> 00:07:26.100 perfect, easily transferable format. Claude actually 00:07:26.100 --> 00:07:28.920 provides a basic prompt in their documentation. 00:07:29.160 --> 00:07:31.699 But the guide we are analyzing tweaked it heavily. 00:07:32.000 --> 00:07:34.259 The tweaked version is much better for a power 00:07:34.259 --> 00:07:37.300 user's move. The real secret here is using markdown 00:07:37.300 --> 00:07:40.459 code blocks. Exactly. You tell the AI to format 00:07:40.459 --> 00:07:43.240 everything inside one single code block. This 00:07:43.240 --> 00:07:45.120 keeps the plain text from getting messy when 00:07:45.120 --> 00:07:47.790 you copy it. It strips out all the weird formatting. 00:07:47.889 --> 00:07:50.250 And it makes a convenient copy button magically 00:07:50.250 --> 00:07:53.350 appear. I love the sheer elegance of using a 00:07:53.350 --> 00:07:56.269 code block for plain text. It is such a clever, 00:07:56.470 --> 00:07:58.829 simple hack. It prevents all those rich text 00:07:58.829 --> 00:08:01.089 formatting nightmares. It really is brilliant. 00:08:01.329 --> 00:08:04.329 But the actual content of the prompt is even 00:08:04.329 --> 00:08:06.750 more important. You paste this master prompt 00:08:06.750 --> 00:08:09.689 into a new chat. Using that thinking model. You 00:08:09.689 --> 00:08:12.439 explicitly state... I am moving my workflow to 00:08:12.439 --> 00:08:15.540 another service. You command it to export every 00:08:15.540 --> 00:08:18.339 single bit of personalized data, every stored 00:08:18.339 --> 00:08:21.060 memory, every custom instruction, every piece 00:08:21.060 --> 00:08:24.480 of deep context it has about your job. You demand 00:08:24.480 --> 00:08:27.220 the data in a highly specific format. You want 00:08:27.220 --> 00:08:30.069 the category, the date. and the specific memory 00:08:30.069 --> 00:08:32.250 itself. You ask it to include your preferred 00:08:32.250 --> 00:08:35.730 name and location. Your exact job title and your 00:08:35.730 --> 00:08:37.870 overarching work goals. Your preferred coding 00:08:37.870 --> 00:08:40.570 languages and specific frameworks. You even ask 00:08:40.570 --> 00:08:43.669 for your nuanced writing style preferences. Things 00:08:43.669 --> 00:08:47.590 like always be concise or use dry humor. And 00:08:47.590 --> 00:08:50.590 any of your strict always do X or never do Y 00:08:50.590 --> 00:08:53.549 rules. And here is the absolute most critical 00:08:53.549 --> 00:08:56.870 instruction of all. You tell it, do not summarize 00:08:56.870 --> 00:08:59.309 anything. Right. Do not group these memories 00:08:59.309 --> 00:09:01.929 into general themes. I want the raw verbatim 00:09:01.929 --> 00:09:04.610 data exactly as it was saved. This preserves 00:09:04.610 --> 00:09:07.370 your entirely unique working style. If the AI 00:09:07.370 --> 00:09:10.330 summarizes your preferences, you lose that nuance. 00:09:10.490 --> 00:09:13.350 All that precious hard -earned context just vanishes. 00:09:13.549 --> 00:09:16.690 So why must we demand raw verbatim data instead 00:09:16.690 --> 00:09:19.049 of summaries? Summaries strip away the nuance 00:09:19.049 --> 00:09:21.190 -specific ways you prefer to work and write. 00:09:21.529 --> 00:09:23.610 Summaries strip away the nuance -specific ways 00:09:23.610 --> 00:09:26.340 you prefer to work and write. Perfect. That brings 00:09:26.340 --> 00:09:30.059 us to the actual import phase. Sponsor. All right. 00:09:30.059 --> 00:09:32.519 Now you have your master export completed. You 00:09:32.519 --> 00:09:35.899 have a massive clean block of text from chat 00:09:35.899 --> 00:09:38.860 GPT. It is time to head over to Claude and complete 00:09:38.860 --> 00:09:41.759 the brain transplant. Anthropic new power users 00:09:41.759 --> 00:09:44.639 were switching platforms in droves. So they built 00:09:44.639 --> 00:09:47.940 a highly specific tool just for this. It makes 00:09:47.940 --> 00:09:50.320 the platform competition very friendly for the 00:09:50.320 --> 00:09:53.139 user. You go into Claude's interface. You dive 00:09:53.139 --> 00:09:55.399 into your account settings. Under the capabilities 00:09:55.399 --> 00:09:57.620 section, looks for the memory configuration. 00:09:57.860 --> 00:09:59.820 You will see a dedicated button waiting right 00:09:59.820 --> 00:10:03.120 there. It explicitly says import memory from 00:10:03.120 --> 00:10:05.860 other AI providers. You just click start import. 00:10:06.000 --> 00:10:08.279 A clean text box will appear on your screen. 00:10:08.519 --> 00:10:11.139 This is where you paste that big perfectly formatted 00:10:11.139 --> 00:10:13.740 code block. You paste the raw text and simply 00:10:13.740 --> 00:10:15.940 click add to memory. This is where the underlying 00:10:15.940 --> 00:10:19.600 technology gets really fascinating. beat. Claude 00:10:19.600 --> 00:10:22.259 does not just save this as a dumb static note. 00:10:22.460 --> 00:10:25.259 No, it actively digests and learns the information. 00:10:25.460 --> 00:10:28.899 If your imported text says, I prefer functional 00:10:28.899 --> 00:10:31.580 programming in Python, Claude genuinely understands 00:10:31.580 --> 00:10:34.899 that concept. It knows it should default to functional 00:10:34.899 --> 00:10:37.700 Python in the future. It stores those specific 00:10:37.700 --> 00:10:41.059 details as active behavioral memories. That's 00:10:41.059 --> 00:10:43.860 exactly why the move feels so magical. The AI 00:10:43.860 --> 00:10:46.960 effectively learns your entire professional past 00:10:46.960 --> 00:10:50.840 in seconds. After you click save, you absolutely 00:10:50.840 --> 00:10:54.340 must verify the import. Start a brand new blank 00:10:54.340 --> 00:10:57.019 chat. Ask it a very simple question. Based on 00:10:57.019 --> 00:11:00.019 what you just learned, what is my job? If it 00:11:00.019 --> 00:11:03.100 answers correctly, your digital migration was 00:11:03.100 --> 00:11:05.799 a success. But you need to deeply understand 00:11:05.799 --> 00:11:08.500 how their underlying brains differ. Yeah, the 00:11:08.500 --> 00:11:10.960 memory architectures are quite different. ChatGPT 00:11:10.960 --> 00:11:13.600 saves a new memory the exact moment you say it. 00:11:13.820 --> 00:11:15.980 It is an instant database, right? Claude is a 00:11:15.980 --> 00:11:17.919 bit more architectural and thoughtful about it. 00:11:18.220 --> 00:11:20.840 Claude processes your ongoing conversations quite 00:11:20.840 --> 00:11:23.620 differently. It updates its deeper long -term 00:11:23.620 --> 00:11:26.740 memory weights every 24 hours. Usually this heavy 00:11:26.740 --> 00:11:28.659 processing happens overnight. So if you tell 00:11:28.659 --> 00:11:31.929 Claude a brand new fact today... It waits. It 00:11:31.929 --> 00:11:33.710 perfectly remembers it for your current chat 00:11:33.710 --> 00:11:35.549 session. But it might not show up in the global 00:11:35.549 --> 00:11:38.330 memory list until tomorrow. Claude is also heavily 00:11:38.330 --> 00:11:41.389 tuned as a professional work assistant. It prioritizes 00:11:41.389 --> 00:11:44.269 your professional tools and complex project goals. 00:11:44.549 --> 00:11:47.870 It might actively overlook personal trivia. Like 00:11:47.870 --> 00:11:50.470 your cat's name or your favorite movie. You can 00:11:50.470 --> 00:11:52.730 always add those personal details manually later. 00:11:52.990 --> 00:11:55.309 It focuses relentlessly on the productivity side 00:11:55.309 --> 00:11:57.769 of things. So people often get confused about 00:11:57.769 --> 00:12:00.429 the timing. Why might Claude forget your name 00:12:00.429 --> 00:12:03.750 in a brand new chat today? Claude processes and 00:12:03.750 --> 00:12:06.129 officially updates its deeper memory overnight, 00:12:06.549 --> 00:12:09.309 not instantly. Claude processes and officially 00:12:09.309 --> 00:12:12.250 updates its deeper memory overnight, not instantly. 00:12:12.610 --> 00:12:14.309 That is a great distinction. It really helps 00:12:14.309 --> 00:12:17.049 manage expectations. Let's transition to handling 00:12:17.049 --> 00:12:20.350 highly specialized workspaces. If you are a pro 00:12:20.350 --> 00:12:23.789 user on either platform, you likely use projects. 00:12:24.009 --> 00:12:26.950 Both platforms charge about $20 a month for their 00:12:26.950 --> 00:12:29.610 premium tiers. Projects are incredibly useful 00:12:29.610 --> 00:12:32.210 for compartmentalizing your life. They are completely 00:12:32.210 --> 00:12:35.009 isolated spaces for different parts of your workflow. 00:12:35.370 --> 00:12:37.629 You might have a dedicated fitness project and 00:12:37.629 --> 00:12:40.370 a separate coding project. Moving these isolated 00:12:40.370 --> 00:12:43.549 projects requires a bit more intentional effort. 00:12:43.769 --> 00:12:47.289 You cannot just execute one big bulk export for 00:12:47.289 --> 00:12:49.129 everything. You have to go into each project 00:12:49.129 --> 00:12:51.730 completely individually. It is definitely manual 00:12:51.730 --> 00:12:55.009 labor. but it is vital for maintaining context. 00:12:55.470 --> 00:12:58.409 You open your specific project over in ChatGPT. 00:12:58.490 --> 00:13:01.269 You use that exact same master export prompt 00:13:01.269 --> 00:13:04.330 we discussed. Because you are inside that specific 00:13:04.330 --> 00:13:07.529 walled garden, ChatGPT looks around locally. 00:13:07.690 --> 00:13:09.929 It pulls the unique project instructions and 00:13:09.929 --> 00:13:12.850 the context of uploaded files. You meticulously 00:13:12.850 --> 00:13:15.769 copy that highly specific data. Then you head 00:13:15.769 --> 00:13:18.370 over to Claude to manually rebuild it. In Claude, 00:13:18.409 --> 00:13:20.649 you create a brand new project. You give it the 00:13:20.649 --> 00:13:23.460 exact same name for continuity. In the new project 00:13:23.460 --> 00:13:26.139 instructions area, you paste your freshly exported 00:13:26.139 --> 00:13:28.679 data. And please, do not forget to re -upload 00:13:28.679 --> 00:13:30.440 your core knowledge files. Put your reference 00:13:30.440 --> 00:13:33.419 PDFs and your complex spreadsheets back in. Comparing 00:13:33.419 --> 00:13:36.259 the two directly, Claw's version of projects 00:13:36.259 --> 00:13:39.340 is just more powerful. This brings us right back 00:13:39.340 --> 00:13:43.039 to that incredible artifacts feature. Whoa! Imagine 00:13:43.039 --> 00:13:45.179 it building a working dashboard right inside 00:13:45.179 --> 00:13:47.960 the chat from your uploaded files. Two secs silence. 00:13:48.379 --> 00:13:50.919 It is incredibly powerful to watch that happen 00:13:50.919 --> 00:13:53.580 in real time. It fundamentally transforms how 00:13:53.580 --> 00:13:56.039 you interact with stagnant data. It takes the 00:13:56.039 --> 00:13:59.039 concept of a workspace to a completely new level. 00:13:59.240 --> 00:14:01.860 It's not just chat anymore. It's active software 00:14:01.860 --> 00:14:04.779 generation. This step usually trips up users 00:14:04.779 --> 00:14:08.000 with massive setups. Can you bulk export all 00:14:08.000 --> 00:14:10.519 your specialized workspaces at once? No. You 00:14:10.519 --> 00:14:13.100 must export and rebuild each project individually 00:14:13.100 --> 00:14:15.980 to maintain the context. No! You must export 00:14:15.980 --> 00:14:18.600 and rebuild each project individually to maintain 00:14:18.600 --> 00:14:21.289 the context. It takes time. But it is worth it. 00:14:21.470 --> 00:14:23.610 We need to clearly recap the common traps here. 00:14:23.789 --> 00:14:25.710 I want to ensure you avoid these frustrating 00:14:25.710 --> 00:14:28.370 pitfalls completely. There are three big mistakes 00:14:28.370 --> 00:14:30.970 that people constantly make. We see these exact 00:14:30.970 --> 00:14:34.190 same unforced errors over and over. The very 00:14:34.190 --> 00:14:37.029 first mistake is using the wrong AI model for 00:14:37.029 --> 00:14:39.490 the export. People often use a fast, lazy model 00:14:39.490 --> 00:14:42.090 like GPT -4 Mini. Just because it's the default. 00:14:42.429 --> 00:14:44.669 If you do that, you get a pathetically tiny list. 00:14:44.889 --> 00:14:46.769 It might only give you five highly recent memories. 00:14:47.049 --> 00:14:50.509 You absolutely must use the thinking model to 00:14:50.509 --> 00:14:53.649 extract your full history. You want the complete 00:14:53.649 --> 00:14:57.029 rich list of 50 or more distinct memories. The 00:14:57.029 --> 00:15:00.029 second major mistake is not respecting Claude's 00:15:00.029 --> 00:15:02.990 24 -hour sync cycle. People paste their massive 00:15:02.990 --> 00:15:06.210 memory blocks into Claude and start testing immediately. 00:15:06.309 --> 00:15:08.549 They start a brand new chat 10 minutes later 00:15:08.549 --> 00:15:11.629 expecting a miracle. And then they get incredibly 00:15:11.629 --> 00:15:14.110 frustrated when it fails. Claude simply says, 00:15:14.110 --> 00:15:16.029 I don't know who you are. We have to give the 00:15:16.029 --> 00:15:18.649 system a few hours. The backend architecture 00:15:18.649 --> 00:15:22.009 needs time to properly index your data. The third 00:15:22.009 --> 00:15:24.830 critical mistake is forgetting the custom instructions 00:15:24.830 --> 00:15:28.450 box. ChatGPT has a highly specific area buried 00:15:28.450 --> 00:15:31.200 in the settings. It asks. How would you like 00:15:31.200 --> 00:15:34.080 chat GPT to respond? People constantly forget 00:15:34.080 --> 00:15:37.019 to copy that specific text separately. That hidden 00:15:37.019 --> 00:15:39.320 box is arguably the most important part of the 00:15:39.320 --> 00:15:42.059 transition. Is the core persona setting? It is 00:15:42.059 --> 00:15:44.039 what makes Claude actually feel like your trusted 00:15:44.039 --> 00:15:45.960 old assistant. You have to grab every single 00:15:45.960 --> 00:15:48.700 piece of the puzzle. Absolutely. It cannot leave 00:15:48.700 --> 00:15:51.279 the most important pieces behind. Just to hammer 00:15:51.279 --> 00:15:54.500 this point home, what happens if you use a lazy 00:15:54.500 --> 00:15:57.409 model for the export? You will only get about 00:15:57.409 --> 00:15:59.909 five recent memories instead of your complete 00:15:59.909 --> 00:16:01.970 history. You will only get about five recent 00:16:01.970 --> 00:16:05.090 memories instead of your complete history. Don't 00:16:05.090 --> 00:16:07.190 do it. We have covered a tremendous amount of 00:16:07.190 --> 00:16:09.970 ground today. Let's distill the core philosophy 00:16:09.970 --> 00:16:13.029 behind all of this technical maneuvering. Beat. 00:16:13.549 --> 00:16:15.789 Ultimately, you own your preferences and your 00:16:15.789 --> 00:16:18.429 digital history. You are never locked into one 00:16:18.429 --> 00:16:21.399 specific rigid tool. That is the most empowering 00:16:21.399 --> 00:16:24.440 part of this whole technological shift. You are 00:16:24.440 --> 00:16:27.460 finally in control of your own AI persona. By 00:16:27.460 --> 00:16:30.019 pulling your data out meticulously, you take 00:16:30.019 --> 00:16:33.200 absolute control. You force the AI to work precisely 00:16:33.200 --> 00:16:35.980 for you. It is like stacking Lego blocks of data 00:16:35.980 --> 00:16:38.259 exactly how you want them. And remember this 00:16:38.259 --> 00:16:40.899 exact same logic works perfectly in reverse. 00:16:41.100 --> 00:16:42.940 If you try Claude for a month and absolutely 00:16:42.940 --> 00:16:45.200 hate it, you can move back. You can transplant 00:16:45.200 --> 00:16:47.960 Claude's newly acquired memories right back into 00:16:47.960 --> 00:16:51.000 Chad GPT. You are completely free to roam. Beat. 00:16:51.340 --> 00:16:53.340 I want to challenge you directly right now. Use 00:16:53.340 --> 00:16:55.799 the import tool today. Set up just one single 00:16:55.799 --> 00:16:59.059 project in Claude. Give it a strict, unbiased, 00:16:59.299 --> 00:17:01.700 one -week trial run. See if the coding logic 00:17:01.700 --> 00:17:03.519 and the natural writing actually feel better. 00:17:03.779 --> 00:17:06.400 I want to leave you with one final, fascinating 00:17:06.400 --> 00:17:09.099 thought. If transferring an AI's entire memory 00:17:09.099 --> 00:17:12.200 is this simple right now. what happens in a few 00:17:12.200 --> 00:17:14.640 years. Imagine when you can seamlessly merge 00:17:14.640 --> 00:17:17.259 the memories of three or four different specialized 00:17:17.259 --> 00:17:21.180 AIs into one ultimate hyper -personalized super 00:17:21.180 --> 00:17:24.019 assistant. Beat! That is a profound future to 00:17:24.019 --> 00:17:26.119 think about. Thank you for joining us on this 00:17:26.119 --> 00:17:28.799 deep dive today. Be well, Outiro Music.