WEBVTT 00:00:00.000 --> 00:00:03.500 Imagine waking up tomorrow morning. beat. The 00:00:03.500 --> 00:00:06.019 sun is just rising outside your window. And your 00:00:06.019 --> 00:00:08.000 AI didn't just answer some questions overnight. 00:00:08.339 --> 00:00:10.779 It actually ran your entire business while you 00:00:10.779 --> 00:00:13.119 slept. It negotiated contracts, right? And optimized 00:00:13.119 --> 00:00:17.059 your campaigns. Exactly. Welcome to today's deep 00:00:17.059 --> 00:00:19.620 dive. We are looking at a monumental industry 00:00:19.620 --> 00:00:23.199 shift today. We've curated some truly fascinating 00:00:23.199 --> 00:00:25.719 sources for you. The landscape is just changing 00:00:25.719 --> 00:00:27.920 incredibly fast right now. Yeah, it really is. 00:00:28.100 --> 00:00:31.109 So here's our roadmap for today. We are tracking 00:00:31.109 --> 00:00:35.490 an entirely new direction for AI. First, we explore 00:00:35.490 --> 00:00:37.990 OpenAI's major pivot at the Cannes Festival. 00:00:38.229 --> 00:00:41.369 They are becoming a full -fledged ad tech powerhouse. 00:00:41.570 --> 00:00:44.679 Oh, yeah. Massive move. Next, we cover the absolute 00:00:44.679 --> 00:00:47.780 death of prompting. We look at the rise of loop 00:00:47.780 --> 00:00:50.340 engineering instead. Right. And the ecosystem 00:00:50.340 --> 00:00:52.719 updates supporting those autonomous agents. Exactly. 00:00:53.020 --> 00:00:55.380 And finally, we unpack a breakthrough in financial 00:00:55.380 --> 00:00:58.439 data. It completely changes how AI reads complex 00:00:58.439 --> 00:01:02.259 SEC filings. But I have a vulnerable admission 00:01:02.259 --> 00:01:04.670 before we begin. Okay, let's hear it. I still 00:01:04.670 --> 00:01:07.750 wrestle with prompt drift myself. You know, you 00:01:07.750 --> 00:01:09.930 write a prompt, but you just lose the underlying 00:01:09.930 --> 00:01:12.310 logic. Well, we all do. I mean, prompting is 00:01:12.310 --> 00:01:15.209 inherently messy. It's unpredictable. It's exactly 00:01:15.209 --> 00:01:17.790 why the industry is moving past it. Yeah. So 00:01:17.790 --> 00:01:19.989 let's start right at the commercial peak today. 00:01:20.329 --> 00:01:22.750 To understand where AI is actually going next, 00:01:22.849 --> 00:01:25.010 we have to look at how it makes money. Always 00:01:25.010 --> 00:01:27.659 follow the capital. It reveals the true infrastructure. 00:01:27.980 --> 00:01:31.459 Right. So OpenAI went to the Cannes Lions Festival 00:01:31.459 --> 00:01:34.000 recently, and the narrative around them has completely 00:01:34.000 --> 00:01:36.500 shifted this year. They aren't just a quirky 00:01:36.500 --> 00:01:40.099 tech lab anymore. They are arriving as a media 00:01:40.099 --> 00:01:42.879 and advertising company. Which is huge. Credio 00:01:42.879 --> 00:01:44.719 actually invited them to the Cannes Festival. 00:01:44.840 --> 00:01:47.480 That is a major signal to traditional advertising 00:01:47.480 --> 00:01:50.159 executives. Just days before Cannes, they made 00:01:50.159 --> 00:01:52.099 a huge move. They expanded their self -serve 00:01:52.099 --> 00:01:54.579 ads manager in the UK. Yeah. They officially 00:01:54.579 --> 00:01:57.019 rolled out cost - per click bidding. Cost per 00:01:57.019 --> 00:02:00.439 click means advertisers pay only when users actually 00:02:00.439 --> 00:02:03.840 click. It is the exact model that built modern 00:02:03.840 --> 00:02:07.780 search. And now it is being applied to conversational 00:02:07.780 --> 00:02:11.360 AI interfaces. Exactly. And OpenAI's chief revenue 00:02:11.360 --> 00:02:14.419 officer headlined some core sessions there. The 00:02:14.419 --> 00:02:17.280 main topic was advertising in the age of AI. 00:02:17.870 --> 00:02:20.090 Their core strategy here is becoming incredibly 00:02:20.090 --> 00:02:22.750 obvious. I mean, they are building infrastructure 00:02:22.750 --> 00:02:27.210 to monetize their massive traffic. ChatGPT generates 00:02:27.210 --> 00:02:29.870 billions of interactions every single month. 00:02:30.150 --> 00:02:33.490 Yeah, the scale is wild. And agencies are demanding 00:02:33.490 --> 00:02:35.949 something entirely different these days. They 00:02:35.949 --> 00:02:38.689 aren't just buying standalone AI models anymore. 00:02:38.949 --> 00:02:41.909 They want end -to -end marketing operating systems. 00:02:42.189 --> 00:02:44.569 Right. They want systems that run, orchestrate, 00:02:44.590 --> 00:02:46.810 and optimize operations. They need an engine 00:02:46.810 --> 00:02:50.680 that handles... So let's unpack what an end -to 00:02:50.680 --> 00:02:52.659 -end system actually means. It doesn't just write 00:02:52.659 --> 00:02:54.860 the ad copy for you. No, not at all. It targets 00:02:54.860 --> 00:02:57.319 the user, bids for the placement, and deploys 00:02:57.319 --> 00:02:59.960 it. It monitors the analytics and adjusts the 00:02:59.960 --> 00:03:01.680 spend automatically. It essentially replaces 00:03:01.680 --> 00:03:04.819 an entire floor of media buyers. Yeah. OpenAI 00:03:04.819 --> 00:03:07.659 is using Khan to prove a major point. They want 00:03:07.659 --> 00:03:09.759 to be the commercial backbone for major brands. 00:03:09.979 --> 00:03:12.180 It's like they stopped selling the engine entirely. 00:03:12.439 --> 00:03:14.259 They started selling the whole highway system 00:03:14.259 --> 00:03:16.360 instead. That's a perfect structural analogy. 00:03:16.680 --> 00:03:19.479 They were providing the roads, the signs, and 00:03:19.479 --> 00:03:22.039 the destination. There was also some initial 00:03:22.039 --> 00:03:25.719 buzz about a film project. An AI -assisted animated 00:03:25.719 --> 00:03:28.419 film called Critters was supposed to debut. Right, 00:03:28.460 --> 00:03:30.819 but that project reportedly stalled out recently. 00:03:31.060 --> 00:03:33.659 It was tied to the shutdown of Soar's consumer 00:03:33.659 --> 00:03:36.300 tool. It really shows where their true priorities 00:03:36.300 --> 00:03:39.099 are right now. Consumer video tools are flashy 00:03:39.099 --> 00:03:42.199 but highly expensive. The real story to watch 00:03:42.199 --> 00:03:44.780 is their ad tech infrastructure. But let me ask 00:03:44.780 --> 00:03:48.060 you this. Is this just about monetizing chat 00:03:48.060 --> 00:03:51.280 GPT traffic or s***? something bigger oh it is 00:03:51.280 --> 00:03:53.060 definitely something much bigger than just chat 00:03:53.060 --> 00:03:55.879 traffic they're building the fundamental commercial 00:03:55.879 --> 00:03:58.870 backbone for brands. They want to intermediate 00:03:58.870 --> 00:04:01.349 every single transaction on the Internet. So 00:04:01.349 --> 00:04:03.430 they're becoming the operating system, not just 00:04:03.430 --> 00:04:05.750 the tool. Exactly. And that leads us directly 00:04:05.750 --> 00:04:08.530 to how that system runs. Right. If OpenAI is 00:04:08.530 --> 00:04:10.550 building this comprehensive marketing operating 00:04:10.550 --> 00:04:13.409 system, how does that system actually function 00:04:13.409 --> 00:04:15.789 in the real world? It requires AI that doesn't 00:04:15.789 --> 00:04:18.350 wait for human commands. It requires agents that 00:04:18.350 --> 00:04:20.790 can think and act continuously. And that brings 00:04:20.790 --> 00:04:23.769 us to the agentic shift. We are moving away from 00:04:23.769 --> 00:04:26.120 traditional prompt engineering completely. Thank 00:04:26.120 --> 00:04:28.860 goodness. Yeah. Now, loop engineering is the 00:04:28.860 --> 00:04:31.519 new standard in the industry. Loop engineering 00:04:31.519 --> 00:04:34.339 means AI agents that keep working autonomously 00:04:34.339 --> 00:04:37.439 until a goal is met. Forrest Cherney from Claude 00:04:37.439 --> 00:04:40.259 Code recently highlighted this shift. The team 00:04:40.259 --> 00:04:42.800 over at OpenClaw is seeing it too. Let's clearly 00:04:42.800 --> 00:04:44.839 define how this actually works. You give them 00:04:44.839 --> 00:04:47.399 a goal, and they just loop until finished. Yeah, 00:04:47.420 --> 00:04:49.699 they evaluate the problem, take an action, and 00:04:49.699 --> 00:04:52.509 check their work. Then they loop again until 00:04:52.509 --> 00:04:54.810 the objective is fully met. Like Moonshot AI, 00:04:55.089 --> 00:04:58.550 they just added goal mode to Kimi work. It is 00:04:58.550 --> 00:05:01.129 a desktop agent that tracks progress continuously. 00:05:01.490 --> 00:05:04.209 Right. You set one objective and the agent just 00:05:04.209 --> 00:05:07.069 keeps grinding. You can check its progress anytime 00:05:07.069 --> 00:05:09.689 you want. It is a fundamental shift in how we 00:05:09.689 --> 00:05:12.300 interact with software. You aren't typing back 00:05:12.300 --> 00:05:14.740 and forth anymore. You are managing a persistent 00:05:14.740 --> 00:05:16.839 digital worker. But running these persistent 00:05:16.839 --> 00:05:20.019 agents requires serious computing power. You 00:05:20.019 --> 00:05:21.959 can't just leave a laptop running constantly. 00:05:22.220 --> 00:05:24.560 No, you really can't. That is where new tools 00:05:24.560 --> 00:05:28.139 like Agent 37 come in. Agent 37 offers managed 00:05:28.139 --> 00:05:31.480 hosting for persistent AI agents. It supports 00:05:31.480 --> 00:05:34.839 popular agents like Hermes, OpenClaw, and CloudCode. 00:05:34.980 --> 00:05:36.720 So you don't need to run them on a Mac mini. 00:05:36.819 --> 00:05:38.759 They live in the cloud and work around the clock. 00:05:39.819 --> 00:05:42.759 We are also seeing tools like Atomic Male Agentic 00:05:42.759 --> 00:05:46.259 emerge. It gives autonomous AI agents their own 00:05:46.259 --> 00:05:49.160 real inbox. Right. They manage it fully without 00:05:49.160 --> 00:05:52.100 human setup or ongoing intervention. They literally 00:05:52.100 --> 00:05:54.699 read the incoming emails, graph responses, and 00:05:54.699 --> 00:05:57.100 send them off. I have to push back on that specific 00:05:57.100 --> 00:05:59.620 idea, though. Handing over autonomous control 00:05:59.620 --> 00:06:02.560 of a real business inbox. Yeah. Letting an AI 00:06:02.560 --> 00:06:05.339 reply to sensitive emails without supervision. 00:06:05.519 --> 00:06:07.779 Right. Beat. That feels like a corporate disaster 00:06:07.779 --> 00:06:10.279 waiting to happen. It is a very valid fear to 00:06:10.279 --> 00:06:12.259 have right now. But the technology is adapting 00:06:12.259 --> 00:06:15.259 quickly to address that exact risk. That is exactly 00:06:15.259 --> 00:06:17.879 why tools like Backgrind are gaining traction. 00:06:18.220 --> 00:06:20.660 Right. Backgrind runs your AI agents over any 00:06:20.660 --> 00:06:23.740 app or game. But it only pings you when it actually 00:06:23.740 --> 00:06:26.199 needs a decision. Right. And that is the crucial 00:06:26.199 --> 00:06:29.319 safety mechanism we need. It handles the mundane 00:06:29.319 --> 00:06:33.399 routing and basic formatting tasks. But it escalates 00:06:33.399 --> 00:06:36.399 the high stakes choices directly to a human. 00:06:36.680 --> 00:06:38.959 We're also seeing this autonomy inside our daily 00:06:38.959 --> 00:06:41.720 documents. Grok for Word now generates outlines 00:06:41.720 --> 00:06:44.379 right inside the panel. Oh, wow. It can restructure 00:06:44.379 --> 00:06:47.019 wording and create complex tables autonomously. 00:06:47.100 --> 00:06:49.100 It doesn't wait for you to highlight and click 00:06:49.100 --> 00:06:52.019 format. And Cloudback MCP connects to clients 00:06:52.019 --> 00:06:55.800 like Cursor or VS Code. It manages 300 backup 00:06:55.800 --> 00:06:58.240 definitions through simple chat commands. So 00:06:58.240 --> 00:07:00.600 what happens when an agent gets stuck in a loop 00:07:00.600 --> 00:07:03.040 trying to solve a problem? Well, new tools are 00:07:03.040 --> 00:07:05.199 being built specifically to handle agent friction. 00:07:05.819 --> 00:07:08.160 When they get suck, they trigger a decision ping 00:07:08.160 --> 00:07:11.079 to you. They outline the specific roadblock and 00:07:11.079 --> 00:07:13.819 ask for your human intuition. Got it. Human oversight 00:07:13.819 --> 00:07:16.899 shifts from micromanaging to just approving decisions. 00:07:17.360 --> 00:07:19.180 Exactly. You become the senior editor, not the 00:07:19.180 --> 00:07:21.819 junior writer. Right. So to make these continuous 00:07:21.819 --> 00:07:25.060 loops run smoothly, things must change. The underlying 00:07:25.060 --> 00:07:28.019 foundation models need a major structural upgrade. 00:07:28.560 --> 00:07:30.879 They definitely do. They need more speed, larger 00:07:30.879 --> 00:07:33.779 memory, and better debugging tools. You cannot 00:07:33.779 --> 00:07:37.079 run persistent loops on slow, forgetful foundation 00:07:37.079 --> 00:07:39.839 models. They will just lose track of the original 00:07:39.839 --> 00:07:42.160 goal completely. Well, rumors about Cloud Sound 00:07:42.160 --> 00:07:44.199 at 5 are really heating up now. It is currently 00:07:44.199 --> 00:07:47.100 codenamed Fennec inside the development circles. 00:07:47.620 --> 00:07:50.519 Word is it could launch as early as next week. 00:07:50.779 --> 00:07:53.879 The hardware specs leaking out are highly impressive. 00:07:54.120 --> 00:07:56.759 Yeah. It reportedly features a 1 million token 00:07:56.759 --> 00:08:00.899 context window. Tokens are tiny chunks of words 00:08:00.899 --> 00:08:03.699 AI uses to read and write. Right. And a context 00:08:03.699 --> 00:08:06.560 window is the AI's active short -term memory 00:08:06.560 --> 00:08:08.699 limit. It is supposed to have vastly stronger 00:08:08.699 --> 00:08:11.620 coding capabilities, too. A million tokens means 00:08:11.620 --> 00:08:14.220 it remembers massive amounts of data. It can 00:08:14.220 --> 00:08:16.579 hold hundreds of textbooks in its active memory. 00:08:16.819 --> 00:08:18.839 We are also seeing Grok build remote showing 00:08:18.839 --> 00:08:20.980 up online. It looks unfinished and was likely 00:08:20.980 --> 00:08:23.379 leaked by accident. Yeah, it is unclear if it 00:08:23.379 --> 00:08:26.029 lives on. folds into cursor cursor is a wildly 00:08:26.029 --> 00:08:29.370 popular ai code editor right now a grok integration 00:08:29.370 --> 00:08:32.149 would disrupt that market significantly definitely 00:08:32.149 --> 00:08:35.250 there is also big money moving in the debugging 00:08:35.250 --> 00:08:38.990 space elastic is buying an ai debugging startup 00:08:38.990 --> 00:08:42.730 called deductive ai the deal is reportedly worth 00:08:42.730 --> 00:08:46.669 up to 85 million dollars deductive ai is a startup 00:08:46.669 --> 00:08:49.950 from 2023 with solid revenue yeah they've reached 00:08:49.950 --> 00:08:52.429 about 1 million in annual recurring revenue already 00:08:52.990 --> 00:08:56.129 This will massively boost Elastic's AI observability 00:08:56.129 --> 00:08:58.850 tools. When an agent hallucinates, you need to 00:08:58.850 --> 00:09:00.990 know exactly why. You have to trace the error 00:09:00.990 --> 00:09:03.850 back to its source. It's like buying an MRI machine 00:09:03.850 --> 00:09:06.429 specifically to diagnose where the AI went wrong. 00:09:06.590 --> 00:09:09.750 You need to see the exact layer inside the black 00:09:09.750 --> 00:09:11.809 box. That is exactly what observability tools 00:09:11.809 --> 00:09:14.350 provide for developers today. They map out the 00:09:14.350 --> 00:09:17.179 AI's internal logic pathways clearly. We are 00:09:17.179 --> 00:09:19.820 also seeing fascinating Y Combinator -backed 00:09:19.820 --> 00:09:22.240 startups emerge. Palmier just unveiled an entirely 00:09:22.240 --> 00:09:25.279 new video editor tool. Oh, yeah. Cloud or Codex 00:09:25.279 --> 00:09:27.379 can directly edit videos inside the software. 00:09:27.539 --> 00:09:29.779 It works seamlessly with Sedans 2 .0 and Cling 00:09:29.779 --> 00:09:32.399 D3. And it is completely free to download right 00:09:32.399 --> 00:09:35.700 now. That strategy really drives rapid user adoption 00:09:35.700 --> 00:09:37.919 in competitive markets. There are some quick 00:09:37.919 --> 00:09:41.169 hits for your toolbox today, too. Microsoft opened 00:09:41.169 --> 00:09:46.990 a free 12 -week AI course online. It has 24 lessons 00:09:46.990 --> 00:09:50.570 covering classic AI concepts thoroughly. It dives 00:09:50.570 --> 00:09:53.190 into deep learning, neural networks, and modern 00:09:53.190 --> 00:09:55.549 architectures. It is an excellent resource for 00:09:55.549 --> 00:09:57.809 anyone trying to catch up. Stanford also released 00:09:57.809 --> 00:10:00.850 a brilliant four part prompting technique. If 00:10:00.850 --> 00:10:03.669 your chat bot gives shallow answers, try this 00:10:03.669 --> 00:10:06.470 specific framework. Yeah, it pushes any model 00:10:06.470 --> 00:10:09.730 to give deeper Ph .D. level analysis. It is perfect 00:10:09.730 --> 00:10:12.649 for complex reports, interviews or big structural 00:10:12.649 --> 00:10:15.289 decisions. Let me ask you, why do we need a one 00:10:15.289 --> 00:10:18.649 million token window just for coding? Massive 00:10:18.649 --> 00:10:21.309 context allows the AI to ingest everything at 00:10:21.309 --> 00:10:23.529 once. It can literally read an entire company's 00:10:23.529 --> 00:10:26.330 code base simultaneously. Wow. Yeah. So it understands 00:10:26.330 --> 00:10:28.230 how a change here breaks a feature over there. 00:10:28.330 --> 00:10:30.110 You need the whole map to navigate, not just 00:10:30.110 --> 00:10:32.409 a single street. Precisely. And navigating that 00:10:32.409 --> 00:10:35.309 map requires flawlessly clean data. That brings 00:10:35.309 --> 00:10:37.509 us to our final highly important topic today. 00:10:37.789 --> 00:10:40.750 Even with huge context windows and these persistent 00:10:40.750 --> 00:10:44.210 working loops, the AI is completely useless if 00:10:44.210 --> 00:10:47.629 the data is garbled. Garbage. Always guarantees 00:10:47.629 --> 00:10:51.070 garbage out. Especially in highly complex systems. 00:10:51.269 --> 00:10:53.870 Nowhere is this more critical than in high -stakes 00:10:53.870 --> 00:10:57.029 global finance. A joint research team just dropped 00:10:57.029 --> 00:10:59.970 a monumental new fix. Yeah, they did. Teams from 00:10:59.970 --> 00:11:02.250 Stanford, the University of California, and Nanjing 00:11:02.250 --> 00:11:04.529 University collaborated. They tackled one of 00:11:04.529 --> 00:11:07.080 the most frustrating problems in modern AI. Feeding 00:11:07.080 --> 00:11:11.000 raw SEC filings into LLMs is notoriously difficult. 00:11:11.460 --> 00:11:15.000 LLMs are large AI systems trained on massive 00:11:15.000 --> 00:11:17.779 text data sets. Right. If you've tried it, you 00:11:17.779 --> 00:11:20.360 already know the painful truth. Financial tables 00:11:20.360 --> 00:11:23.019 usually get completely flattened by the AI models. 00:11:23.220 --> 00:11:25.019 Rows and columns just collapse into a useless 00:11:25.019 --> 00:11:27.799 text paragraph. Exactly. And that destroys the 00:11:27.799 --> 00:11:30.159 actual underlying accounting logic entirely. 00:11:30.559 --> 00:11:33.080 A misplaced decimal or a flattened row changes 00:11:33.080 --> 00:11:36.000 millions of dollars. The research team developed 00:11:36.080 --> 00:11:38.700 a new data set and methodology entirely. It is 00:11:38.700 --> 00:11:41.379 designed specifically to make SEC filings readable 00:11:41.379 --> 00:11:44.179 for machines. It allows the AI to parse the data 00:11:44.179 --> 00:11:46.580 without losing structural meaning. The researchers 00:11:46.580 --> 00:11:49.340 propose something they officially call SCFD. 00:11:49.690 --> 00:11:53.070 It is a reconstructed version of the SEC's edGR 00:11:53.070 --> 00:11:56.250 document filings. Right. It translates the raw 00:11:56.250 --> 00:11:59.629 data into layout -faithful multi -markdown. Multi 00:11:59.629 --> 00:12:02.169 -markdown is a text format preserving complex 00:12:02.169 --> 00:12:05.009 structural formatting clearly. It acts as a perfect 00:12:05.009 --> 00:12:07.529 bridge between human documents and machine code. 00:12:07.789 --> 00:12:10.610 They successfully retain the merged headers and 00:12:10.610 --> 00:12:13.090 the critical indentation. They keep the financial 00:12:13.090 --> 00:12:16.350 signs, spans, and complex table hierarchies. 00:12:16.490 --> 00:12:24.169 And incredibly, it uses far fewer tokens. The 00:12:24.169 --> 00:12:26.809 scale of this data set is just absolutely staggering. 00:12:26.990 --> 00:12:31.690 The team dropped 152 billion token public snapshot 00:12:31.690 --> 00:12:34.389 recently. Processing the full archive yields 00:12:34.389 --> 00:12:37.870 around 550 billion tokens total. That is 550 00:12:37.870 --> 00:12:40.649 billion tokens of structurally perfect financial 00:12:40.649 --> 00:12:43.129 documents. It is an unprecedented amount of high 00:12:43.129 --> 00:12:45.289 quality training material. This data set has 00:12:45.289 --> 00:12:48.149 less than a 0 .1 % overlap with Common Crawl. 00:12:48.200 --> 00:12:50.539 Common Crawl is the standard web scrape data 00:12:50.539 --> 00:12:53.100 most models use. Right, which means it offers 00:12:53.100 --> 00:12:55.279 highly unique, specialized training material. 00:12:55.659 --> 00:12:59.340 Whoa. Imagine scaling to a billion queries on 00:12:59.340 --> 00:13:02.600 flawless financial data. Two sec silence. It 00:13:02.600 --> 00:13:05.000 completely revolutionizes algorithmic trading 00:13:05.000 --> 00:13:07.720 and financial risk analysis forever. This represents 00:13:07.720 --> 00:13:10.159 a massive infrastructural upgrade for the financial 00:13:10.159 --> 00:13:13.720 AI space. It finally provides a clean, standardized 00:13:13.720 --> 00:13:16.980 pipeline for the industry. Millions of complex, 00:13:17.179 --> 00:13:19.639 unstructured financial documents become high 00:13:19.639 --> 00:13:21.980 -quality training data. But how does retaining 00:13:21.980 --> 00:13:25.100 a simple text indentation actually change the 00:13:25.100 --> 00:13:28.549 AI's financial logic? Well, in accounting, indentation 00:13:28.549 --> 00:13:30.990 signifies vital parent -child relationships and 00:13:30.990 --> 00:13:33.669 revenue streams. If operating income isn't indented 00:13:33.669 --> 00:13:36.269 under gross profit, the underlying math breaks. 00:13:36.490 --> 00:13:39.549 Oh, I see. The AI needs that exact visual hierarchy 00:13:39.549 --> 00:13:42.149 to understand the financial reality. Clean data 00:13:42.149 --> 00:13:44.610 finally replaces messy guesswork in financial 00:13:44.610 --> 00:13:47.429 AI models. It truly is the foundation for the 00:13:47.429 --> 00:13:49.750 next generation of financial intelligence. Mid 00:13:49.750 --> 00:13:52.129 -roll sponsor, BreakMarker. We have covered a 00:13:52.129 --> 00:13:54.309 tremendous amount of ground today. Let's pull 00:13:54.309 --> 00:13:56.320 this deep dive together and look at the... big 00:13:56.320 --> 00:13:59.720 picture. We started with OpenAI becoming a comprehensive 00:13:59.720 --> 00:14:03.200 ad tech operating system. They are building the 00:14:03.200 --> 00:14:05.960 commercial infrastructure for the modern web. 00:14:06.159 --> 00:14:08.960 They're moving from a simple tool to the foundational 00:14:08.960 --> 00:14:12.049 layer. We saw how loop engineering is fundamentally 00:14:12.049 --> 00:14:15.590 changing our daily workflows. It is turning AI 00:14:15.590 --> 00:14:18.509 from a passive assistant into a persistent worker. 00:14:18.750 --> 00:14:21.970 Right. We explored the rumors of robust new foundation 00:14:21.970 --> 00:14:24.830 models arriving soon. Models with active memory 00:14:24.830 --> 00:14:27.490 large enough to understand entire corporate ecosystems. 00:14:27.970 --> 00:14:30.090 And finally, we looked at how researchers are 00:14:30.090 --> 00:14:32.970 fixing raw financial data. They are repairing 00:14:32.970 --> 00:14:35.370 the very foundation of financial training materials 00:14:35.370 --> 00:14:38.289 entirely. All these pieces are snapping together 00:14:38.289 --> 00:14:40.509 like Lego. blocks right now. We are building 00:14:40.509 --> 00:14:42.990 a fully autonomous digital economy from the ground 00:14:42.990 --> 00:14:45.669 up. Yeah. The persistent agents, the clean data, 00:14:45.769 --> 00:14:48.110 the monetization framework, it is all finally 00:14:48.110 --> 00:14:50.649 connecting. It is a genuinely fascinating time 00:14:50.649 --> 00:14:53.169 to be watching this technology evolve. I want 00:14:53.169 --> 00:14:55.299 to leave you with one final thought today. If 00:14:55.299 --> 00:14:58.440 AI can run your marketing campaigns at con automatically, 00:14:58.799 --> 00:15:01.500 if it can manage your inbox without ever asking 00:15:01.500 --> 00:15:05.279 permission, and if it flawlessly parses 550 billion 00:15:05.279 --> 00:15:08.059 tokens of complex SEC beta, what is the deeply 00:15:08.059 --> 00:15:09.820 human skill you need to be cultivating tomorrow? 00:15:10.179 --> 00:15:12.139 Beat. Thank you for taking this deep dive with 00:15:12.139 --> 00:15:13.139 us today. Out to your music.