WEBVTT 00:00:00.000 --> 00:00:05.240 Claude Opus 4 .7, GPT 5 .5, GPT 5 .5 Instant, 00:00:05.780 --> 00:00:09.380 Gemini 3 .5 Flash, Gemini Omni Flash, all dropping 00:00:09.380 --> 00:00:11.400 in just over a month. Looking at this timeline, 00:00:11.519 --> 00:00:14.839 it is honestly overwhelming. The AI world is 00:00:14.839 --> 00:00:17.559 moving incredibly fast. It is moving so fast, 00:00:17.640 --> 00:00:21.800 it feels paralyzing. It is beat. It is... a very 00:00:21.800 --> 00:00:23.879 real feeling. You know, people get so overwhelmed 00:00:23.879 --> 00:00:25.899 by the sheer volume of releases. Right. They 00:00:25.899 --> 00:00:28.440 just freeze up entirely. They stop trying to 00:00:28.440 --> 00:00:30.820 keep pace. Welcome to the deep dive. Our mission 00:00:30.820 --> 00:00:33.359 today is very clear. We are going to cut through 00:00:33.359 --> 00:00:36.039 the noise. We really have to. We must stop the 00:00:36.039 --> 00:00:37.979 endless tool chasing right now. We are breaking 00:00:37.979 --> 00:00:41.340 down the 2026 AI roadmap. We need to map out 00:00:41.340 --> 00:00:43.560 what actually works in the real world, not just 00:00:43.560 --> 00:00:45.780 the daily social media hype. We have to look 00:00:45.780 --> 00:00:48.100 at practical functional application. Exactly. 00:00:48.159 --> 00:00:49.820 We will map out exactly where you should start. 00:00:49.859 --> 00:00:52.140 We are explore. three distinct levels today. 00:00:52.560 --> 00:00:54.880 Basic, intermediate, and advanced. Right. Basic 00:00:54.880 --> 00:00:57.619 focuses heavily on deep prompting, intermediate 00:00:57.619 --> 00:01:00.579 covers background AI workflows, and advanced 00:01:00.579 --> 00:01:03.079 dives into agentic engineering is quite the journey. 00:01:03.380 --> 00:01:05.859 But before you build complex systems, before 00:01:05.859 --> 00:01:08.280 you build things that work while you sleep, you 00:01:08.280 --> 00:01:11.620 have to build a bulletproof foundation. And that 00:01:11.620 --> 00:01:14.140 starts with rethinking how you actually talk 00:01:14.140 --> 00:01:16.560 to these models. Most people approach the chat 00:01:16.560 --> 00:01:19.500 window completely backwards. They treat it kind 00:01:19.500 --> 00:01:22.099 of like a standard search engine. So level one 00:01:22.099 --> 00:01:25.260 is the foundation. We are moving away from basic 00:01:25.260 --> 00:01:30.099 dumb queries. We need highly structured intentional 00:01:30.099 --> 00:01:33.319 prompting. Yeah. We need to actively curate a 00:01:33.319 --> 00:01:36.799 tight tool stack and we must hedge our personal 00:01:36.799 --> 00:01:39.299 financial risk. That last point always catches 00:01:39.299 --> 00:01:41.760 people off guard. It is a critical piece of the 00:01:41.760 --> 00:01:43.959 puzzle. We will dig into the money aspect shortly, 00:01:43.959 --> 00:01:46.079 but let's look at prompting mechanics first. 00:01:46.680 --> 00:01:49.040 A weak prompt is essentially useless. Completely. 00:01:49.239 --> 00:01:51.260 You type something like, write an email to my 00:01:51.260 --> 00:01:53.420 boss asking for a day off. You will get pure 00:01:53.420 --> 00:01:55.959 generic garbage bag. It sounds like a soulless 00:01:55.959 --> 00:01:58.480 robot rode it. There is zero context for the 00:01:58.480 --> 00:02:00.780 model to work with. It is like delegating a critical 00:02:00.780 --> 00:02:03.439 task to a brand new intern. You just bark a simple 00:02:03.439 --> 00:02:05.780 command with no background information. Yeah, 00:02:05.819 --> 00:02:08.539 that never works. You cannot possibly expect 00:02:08.539 --> 00:02:11.520 a good result. You need what we call a deep prompt. 00:02:11.719 --> 00:02:14.060 You have to constrain the model's focus. You 00:02:14.060 --> 00:02:16.379 must assign them a specific role first. Right. 00:02:16.560 --> 00:02:19.199 Something like act as a seasoned professional 00:02:19.199 --> 00:02:22.280 marketing manager. That sets the behavioral guardrails 00:02:22.280 --> 00:02:25.099 right away. Then you give them exact, clear details. 00:02:25.539 --> 00:02:28.400 Email David regarding Friday, May 29th. Tell 00:02:28.400 --> 00:02:30.740 him the monthly conversion report is completely 00:02:30.740 --> 00:02:33.560 finished. And finally, you define the exact output 00:02:33.560 --> 00:02:36.699 format. Create a clear professional subject line. 00:02:37.080 --> 00:02:40.139 Write in very short punchy paragraphs. Keep the 00:02:40.139 --> 00:02:42.900 overall tone friendly but concise. The difference 00:02:42.900 --> 00:02:45.360 in the final output is night and day. It gives 00:02:45.360 --> 00:02:47.159 you something you can actually send. Exactly. 00:02:47.240 --> 00:02:49.979 But even with good structure, it is tough. I 00:02:49.979 --> 00:02:53.400 still wrestle with prompt drift myself. Keeping 00:02:53.400 --> 00:02:56.319 the AI on track during long multi -step chats 00:02:56.319 --> 00:02:58.879 is incredibly hard. It definitely wanders off 00:02:58.879 --> 00:03:01.120 over time. The context window gets cluttered 00:03:01.120 --> 00:03:03.740 with old information. That is exactly why you 00:03:03.740 --> 00:03:06.539 need a curated core stack. Stop downloading 50 00:03:06.539 --> 00:03:08.680 different AI apps on your phone. You need to 00:03:08.680 --> 00:03:11.080 pick one primary chatbot and learn it deeply. 00:03:11.539 --> 00:03:13.819 Let's break down the big three right now. Claude 00:03:13.819 --> 00:03:16.539 is fantastic for dense coding and long form writing. 00:03:17.120 --> 00:03:19.780 Its architecture holds onto large documents really 00:03:19.780 --> 00:03:22.610 well. Right. ChatGPT handles general reasoning 00:03:22.610 --> 00:03:26.050 and voice interaction perfectly. Gemini integrates 00:03:26.050 --> 00:03:28.750 deeply and seamlessly with your existing Google 00:03:28.750 --> 00:03:31.370 Workspace apps. Pick just one of those based 00:03:31.370 --> 00:03:34.370 on your daily needs. Then add a dedicated research 00:03:34.370 --> 00:03:37.569 tool to your stack. Like Perplexity. Yeah, Perplexity 00:03:37.569 --> 00:03:40.710 is amazing for finding real -time facts. It cites 00:03:40.710 --> 00:03:43.069 actual sources, which severely limits random 00:03:43.069 --> 00:03:45.870 hallucinations. Learn their keyboard shortcuts 00:03:45.870 --> 00:03:48.289 inside and out. That covers the daily tactical 00:03:48.289 --> 00:03:51.930 tools you need. But there is a much bigger macroeconomic 00:03:51.930 --> 00:03:54.810 picture here. We need to seriously look at AI 00:03:54.810 --> 00:03:57.030 financial exposure. This is huge. You have to 00:03:57.030 --> 00:03:59.229 look at your daily job and your long term money 00:03:59.229 --> 00:04:01.469 together. I want to circle back to that specifically. 00:04:01.990 --> 00:04:05.009 Why do people ignore the financial exposure aspect 00:04:05.009 --> 00:04:08.349 of AI? Because they view AI purely as a productivity 00:04:08.349 --> 00:04:10.949 software. They completely miss its macroeconomic 00:04:10.949 --> 00:04:13.310 impact on their specific industry. If your career 00:04:13.310 --> 00:04:16.230 is at risk from AI, you have to balance your 00:04:16.230 --> 00:04:19.500 personal investments. You cannot be heavily over 00:04:19.500 --> 00:04:22.199 -invested in tech stocks if tech is automating 00:04:22.199 --> 00:04:25.939 your livelihood. So treat AI as an economic shift, 00:04:26.259 --> 00:04:29.379 not just a software update. Beat. Precisely. 00:04:29.939 --> 00:04:32.100 You have to insulate your personal wealth from 00:04:32.100 --> 00:04:34.519 your industry's disruption. That brings us to 00:04:34.519 --> 00:04:37.500 level two. Intermediate workflows and personal 00:04:37.500 --> 00:04:40.560 knowledge management. Your daily foundation is 00:04:40.560 --> 00:04:43.529 solid now. Your prompts are sharp and highly 00:04:43.529 --> 00:04:46.189 constrained, but you eventually hit a hard ceiling 00:04:46.189 --> 00:04:48.250 answering single questions. You could only type 00:04:48.250 --> 00:04:50.230 so fast on a keyboard, there are only so many 00:04:50.230 --> 00:04:53.009 hours in your day. To scale your output, we have 00:04:53.009 --> 00:04:55.829 to transition. We stop chatting with AI directly 00:04:55.829 --> 00:04:58.449 in a browser. We make AI work independently in 00:04:58.449 --> 00:05:00.490 the background. This is where we move into autonomous 00:05:00.490 --> 00:05:02.990 AI agents. It completely changes how you interact 00:05:02.990 --> 00:05:05.470 with computers. Agents handle your big overarching 00:05:05.470 --> 00:05:08.009 goals. You move from single isolated questions 00:05:08.009 --> 00:05:11.040 to multi -step tasks. Instead of asking five 00:05:11.040 --> 00:05:12.959 different sequential questions, you give the 00:05:12.959 --> 00:05:16.160 system a major complex directive, like, research 00:05:16.160 --> 00:05:18.379 my top five competitors in the email newsletter 00:05:18.379 --> 00:05:21.519 space. Find their exact pricing tiers and subscription 00:05:21.519 --> 00:05:24.240 plans. Summarize the main topics they write about 00:05:24.240 --> 00:05:27.040 every week. Make a clean, formatted data table 00:05:27.040 --> 00:05:29.319 for me. The agent goes off and browses the web 00:05:29.319 --> 00:05:32.439 automatically. It iteratively searches, reads, 00:05:32.600 --> 00:05:35.180 and compiles all the pieces. It does all the 00:05:35.180 --> 00:05:37.250 tedious heavy lifting for you. But we can take 00:05:37.250 --> 00:05:39.889 this concept even further. We can create background 00:05:39.889 --> 00:05:42.790 automations and local workflows. Yes. Connecting 00:05:42.790 --> 00:05:44.769 different software apps together seamlessly. 00:05:45.290 --> 00:05:47.750 You can use visual drag and drop workflow tools 00:05:47.750 --> 00:05:51.129 like NN. That does not require you to write complex 00:05:51.129 --> 00:05:54.089 code. It is essentially drawing a map for your 00:05:54.089 --> 00:05:56.550 data. It connects your email directly to an AI 00:05:56.550 --> 00:05:58.670 reasoning model. Then it connects that model 00:05:58.670 --> 00:06:02.269 to your private notes app. Whoa. Imagine waking 00:06:02.269 --> 00:06:05.079 up. and your local agent has already read, sorted, 00:06:05.279 --> 00:06:07.740 and summarized everything. It categorizes your 00:06:07.740 --> 00:06:10.480 chaotic inbox while you are still asleep. It 00:06:10.480 --> 00:06:12.379 is incredibly powerful to experience that the 00:06:12.379 --> 00:06:14.459 first time. And for people worried about data 00:06:14.459 --> 00:06:16.360 leaks, there is a solution. Yeah, definitely. 00:06:16.579 --> 00:06:19.000 You can run private models entirely locally. 00:06:19.500 --> 00:06:22.720 Using desktop tools like OpenClaw ensures total 00:06:22.720 --> 00:06:26.050 data safety. Your data never leaves your actual 00:06:26.050 --> 00:06:29.170 physical machine. It processes everything using 00:06:29.170 --> 00:06:31.569 your laptop's own graphical processing unit. 00:06:31.589 --> 00:06:33.949 It stays completely private from the big tech 00:06:33.949 --> 00:06:36.589 companies. This directly connects to creating 00:06:36.589 --> 00:06:39.430 custom knowledge bases. You feed personal or 00:06:39.430 --> 00:06:41.569 business documents into these specialized tools. 00:06:41.810 --> 00:06:44.860 Like Google Notebook LM. Yeah. That is fantastic 00:06:44.860 --> 00:06:47.939 for handling messy information. Claude projects 00:06:47.939 --> 00:06:50.220 works incredibly well for organizing distinct 00:06:50.220 --> 00:06:53.639 client files. Custom GPTs are also a very solid 00:06:53.639 --> 00:06:56.240 option. It creates a private, deeply informed 00:06:56.240 --> 00:06:59.040 expert for you. One that won't hallucinate random 00:06:59.040 --> 00:07:01.519 internet facts to fill gaps. It is like giving 00:07:01.519 --> 00:07:03.959 the AI an open book test. You are not relying 00:07:03.959 --> 00:07:06.639 on its generic generalized training memory. You 00:07:06.639 --> 00:07:08.939 are pointing it directly to your exact verified 00:07:08.939 --> 00:07:11.199 facts. You tell it to read 20 dense marketing 00:07:11.199 --> 00:07:13.879 PDFs. You tell it to write a strategy post using 00:07:13.879 --> 00:07:16.699 only that specific data. It becomes perfectly 00:07:16.699 --> 00:07:19.319 accurate to your brand's voice. With all this 00:07:19.319 --> 00:07:22.579 automation happening, how safe is uploading private 00:07:22.579 --> 00:07:25.120 company data to these custom knowledge bases? 00:07:25.560 --> 00:07:28.399 Well, cloud tools have enterprise privacy toggles, 00:07:28.500 --> 00:07:31.920 sure, but True data paranoia is why local tools 00:07:31.920 --> 00:07:34.699 like OpenClaw are becoming essential for intermediate 00:07:34.699 --> 00:07:38.139 users. Right. You bypass the cloud entirely to 00:07:38.139 --> 00:07:40.800 prevent any theoretical leaks. For absolute privacy, 00:07:40.920 --> 00:07:43.439 run the AI locally, keep your data off the cloud. 00:07:44.000 --> 00:07:47.220 Two sick sounds. Sponsor. Welcome back to the 00:07:47.220 --> 00:07:49.540 Deep Dive. Let's move to level three on our roadmap, 00:07:50.079 --> 00:07:52.939 advanced business systems and AI coding. Yeah. 00:07:53.220 --> 00:07:55.199 Personal productivity is definitely a massive 00:07:55.199 --> 00:07:57.459 advantage. It is a total game changer for an 00:07:57.459 --> 00:07:59.779 individual contributor. But to create massive 00:07:59.779 --> 00:08:03.120 value in the broader world, to truly build a 00:08:03.120 --> 00:08:06.319 highly resilient lasting career, we have to bridge 00:08:06.319 --> 00:08:09.180 a major operational gap. Absolutely. We must 00:08:09.180 --> 00:08:12.740 move from personal workflows to messy, large 00:08:12.740 --> 00:08:15.339 scale business operations. Businesses are incredibly 00:08:15.339 --> 00:08:18.600 complicated ecosystems. They need absolute reliability 00:08:18.600 --> 00:08:21.180 to function properly without crashing. We are 00:08:21.180 --> 00:08:23.740 talking about building custom agents for business 00:08:23.740 --> 00:08:26.540 applications, building highly reliable automated 00:08:26.540 --> 00:08:30.000 back end system. Let's look at an HR onboarding 00:08:30.000 --> 00:08:32.759 agent as an example. Onboarding is usually a 00:08:32.759 --> 00:08:35.299 complete administrative mess for everyone involved. 00:08:35.720 --> 00:08:38.340 A custom agent completely automates the tedious 00:08:38.340 --> 00:08:42.039 paperwork process. It notices the exact moment 00:08:42.039 --> 00:08:45.919 a new hire joins the payroll system. It instantly 00:08:45.919 --> 00:08:49.360 sends the right localized HR tax forms. Yeah. 00:08:49.700 --> 00:08:53.200 It provides 247 interactive handbook support 00:08:53.200 --> 00:08:55.299 for their immediate questions. It automatically 00:08:55.299 --> 00:08:58.309 triggers IT setups for new company emails, it 00:08:58.309 --> 00:09:00.350 manages the whole workflow from start to finish. 00:09:00.590 --> 00:09:02.450 To build something like that you must connect 00:09:02.450 --> 00:09:05.169 systems. You need to deeply understand two key 00:09:05.169 --> 00:09:08.850 architectural tools, APIs and MCPs. Right. Let's 00:09:08.850 --> 00:09:11.269 define those clearly right now. An API is a digital 00:09:11.269 --> 00:09:13.370 waiter carrying data between different software 00:09:13.370 --> 00:09:15.809 programs. That is a perfect visual way to picture 00:09:15.809 --> 00:09:18.149 it. It takes your specific data order to the 00:09:18.149 --> 00:09:20.090 kitchen. It brings the formatted data back to 00:09:20.090 --> 00:09:22.929 your application quickly. And an MCP is a secure 00:09:22.929 --> 00:09:26.129 bridge letting AI read private local files safely. 00:09:26.330 --> 00:09:28.690 Exactly. That keeps internal company databases 00:09:28.690 --> 00:09:31.210 totally secure from outside models. It prevents 00:09:31.210 --> 00:09:33.669 massive data leaks while still allowing internal 00:09:33.669 --> 00:09:37.470 analysis. This leads us to the final major skill 00:09:37.470 --> 00:09:41.080 here. AI coding and the rise of agentic engineering. 00:09:41.580 --> 00:09:44.980 This means actually using AI to write real functional 00:09:44.980 --> 00:09:48.220 software, then using hosting tools like Vercell 00:09:48.220 --> 00:09:51.039 to launch it live on the internet. Anyone can 00:09:51.039 --> 00:09:53.399 build sophisticated web apps right now. You just 00:09:53.399 --> 00:09:56.279 describe what you want in plain English. The 00:09:56.279 --> 00:09:58.980 AI generates the interface. It is absolutely 00:09:58.980 --> 00:10:01.600 amazing. I have to push back gently on that particular 00:10:01.600 --> 00:10:04.940 narrative. The source material has a very clear, 00:10:05.139 --> 00:10:07.759 stark warning. If you don't know basic coding 00:10:07.759 --> 00:10:10.519 logic, the AI will inevitably make mistakes, 00:10:10.799 --> 00:10:13.460 and you will be completely stuck, unable to debug 00:10:13.460 --> 00:10:16.769 or fix them. That is a very fair, realistic point. 00:10:16.950 --> 00:10:19.110 You cannot be totally blind to the underlying 00:10:19.110 --> 00:10:21.590 architecture. You do need to understand the absolute 00:10:21.590 --> 00:10:24.149 basics of how data flows. With models writing 00:10:24.149 --> 00:10:26.350 entire code bases, does this mean traditional 00:10:26.350 --> 00:10:28.769 software engineers are obsolete? No, they aren't 00:10:28.769 --> 00:10:31.750 obsolete at all. But the role shifts dramatically 00:10:31.750 --> 00:10:34.429 from writing manual syntax to agentic engineering. 00:10:34.549 --> 00:10:37.289 They manage and review the architecture generated 00:10:37.289 --> 00:10:40.529 by the AI. Engineers become editors. They manage 00:10:40.529 --> 00:10:43.309 AI output instead of typing every single line. 00:10:43.610 --> 00:10:46.409 Beat? They guide the complex system instead of 00:10:46.409 --> 00:10:48.669 doing the manual labor. We have mapped out the 00:10:48.669 --> 00:10:51.629 entire destination today, from level one all 00:10:51.629 --> 00:10:54.269 the way up to level three. But knowing the destination 00:10:54.269 --> 00:10:56.909 is completely useless in practice. If you fall 00:10:56.909 --> 00:10:58.950 into the common predictable traps along the way. 00:10:59.129 --> 00:11:01.190 Let's look at the actual timeline and the top 00:11:01.190 --> 00:11:04.470 mistakes. Stage one takes just days or maybe 00:11:04.470 --> 00:11:07.620 weeks to master. Stage two takes a few solid 00:11:07.620 --> 00:11:10.720 weeks to months. And stages three and four, becoming 00:11:10.720 --> 00:11:14.259 a true AI engineer that takes months to years 00:11:14.259 --> 00:11:17.259 of nonstop dedicated learning. It is a long marathon, 00:11:17.460 --> 00:11:20.009 not a quick sprint. So let's learn to avoid the 00:11:20.009 --> 00:11:22.549 worst potential mistakes. Mistake number one 00:11:22.549 --> 00:11:25.230 is incredibly common right now, chasing every 00:11:25.230 --> 00:11:27.389 single new tool that pops up on social media. 00:11:27.610 --> 00:11:30.730 You see a flashy new tool demo online, you sign 00:11:30.730 --> 00:11:33.210 up and use it for five minutes, then you completely 00:11:33.210 --> 00:11:35.690 forget it exists. It is a huge waste of mental 00:11:35.690 --> 00:11:38.210 energy. Mistake number two is ignoring workflow 00:11:38.210 --> 00:11:41.370 thinking entirely, asking for one single generated 00:11:41.370 --> 00:11:43.649 sentence instead of automating a whole repetitive 00:11:43.649 --> 00:11:46.549 process. You need to think systematically from 00:11:46.549 --> 00:11:49.279 start to finish. How does the data logically 00:11:49.279 --> 00:11:59.419 move through your whole day? Social media makes 00:11:59.419 --> 00:12:02.580 these models look like pure magic. But AI always 00:12:02.580 --> 00:12:05.179 requires a layer of human taste. It needs human 00:12:05.179 --> 00:12:07.399 supervision and constant human correction to 00:12:07.399 --> 00:12:13.259 be valuable. People delay simply because the 00:12:13.259 --> 00:12:15.539 underlying technology keeps changing so fast. 00:12:15.759 --> 00:12:17.559 The knowledge gap just gets wider and wider if 00:12:17.559 --> 00:12:20.039 you wait. The fundamentals you learn today will 00:12:20.039 --> 00:12:22.480 apply to tomorrow's models. Waiting to learn 00:12:22.480 --> 00:12:25.539 AI until it is perfectly stable. It is like waiting 00:12:25.539 --> 00:12:27.940 to buy a smartphone until they permanently stop 00:12:27.940 --> 00:12:30.259 making new ones. You are just ensuring you get 00:12:30.259 --> 00:12:32.299 completely left behind by your peers. You have 00:12:32.299 --> 00:12:34.580 to start building your foundation today. Use 00:12:34.580 --> 00:12:36.559 whatever imperfect tool is readily available 00:12:36.559 --> 00:12:39.000 right now. What is the real danger of relying 00:12:39.000 --> 00:12:41.679 entirely on AI without any human supervision? 00:12:42.279 --> 00:12:45.379 AI lacks human taste and broad worldly context. 00:12:46.120 --> 00:12:48.399 Fully unsupervised AI will eventually make a 00:12:48.399 --> 00:12:51.240 costly, embarrassing error in a complex business 00:12:51.240 --> 00:12:54.399 setting. AI provides the raw power, but human 00:12:54.399 --> 00:12:58.120 taste remains the ultimate quality control. Beat! 00:12:58.700 --> 00:13:01.279 That is the lens you need for every single output. 00:13:01.419 --> 00:13:04.360 We have covered a massive amount of dense information 00:13:04.360 --> 00:13:07.379 today. Let's distill this entire deep dive into 00:13:07.379 --> 00:13:10.740 a single grounding takeaway. You don't need to 00:13:10.740 --> 00:13:13.879 reach level four to be highly successful in 2026. 00:13:14.259 --> 00:13:16.659 Pick your specific lane based on your actual 00:13:16.659 --> 00:13:19.840 career goals. Stick to just two or three core 00:13:19.840 --> 00:13:22.799 tools for everything. Build your personal systems 00:13:22.799 --> 00:13:25.519 very slowly and deliberately. Ignore the daily 00:13:25.519 --> 00:13:27.940 social media noise completely. It will just distract 00:13:27.940 --> 00:13:31.230 you from doing real impactful work. Focus on 00:13:31.230 --> 00:13:33.950 stacking these simple Lego blocks of data one 00:13:33.950 --> 00:13:36.730 at a time. Establish your base layer firmly before 00:13:36.730 --> 00:13:39.649 moving up the chain. Just pick one single manageable 00:13:39.649 --> 00:13:42.190 skill from today's roadmap. Refining a highly 00:13:42.190 --> 00:13:45.029 structured deep pumped. Setting up a basic local 00:13:45.029 --> 00:13:47.129 workflow on your laptop. Test it out for yourself 00:13:47.129 --> 00:13:49.190 before the day is over. Get your hands dirty 00:13:49.190 --> 00:13:51.309 in the actual software. That is how you actually 00:13:51.309 --> 00:13:53.610 retain the knowledge. As these agentic systems 00:13:53.610 --> 00:13:55.909 take over the friction of our daily work, doing 00:13:55.909 --> 00:13:58.519 all the logical heavy lifting. Yeah. What happens 00:13:58.519 --> 00:14:01.659 to the premium we place on human intuition? When 00:14:01.659 --> 00:14:04.320 execution becomes free, maybe having a unique 00:14:04.320 --> 00:14:06.620 perspective becomes the most valuable asset in 00:14:06.620 --> 00:14:06.940 the world.