WEBVTT 00:00:00.000 --> 00:00:02.779 So here's the big question. Can Google's new 00:00:02.779 --> 00:00:06.519 genius AI, Gemini 3, actually write pine script 00:00:06.519 --> 00:00:09.919 code that makes you serious money in the crypto 00:00:09.919 --> 00:00:12.160 market? That is the high stakes question, isn't 00:00:12.160 --> 00:00:14.740 it? And we're talking about real rigorous testing 00:00:14.740 --> 00:00:18.039 here. Right. We found one successful code conversion 00:00:18.039 --> 00:00:23.280 that led to just a staggering 2 ,497 % total 00:00:23.280 --> 00:00:26.379 profit on Bitcoin. Wow. But, and this is the 00:00:26.379 --> 00:00:28.480 part we have to talk about, that massive return 00:00:28.480 --> 00:00:32.240 came with a terrifying catch, a 46 % maximum 00:00:32.240 --> 00:00:34.380 drawdown. OK. And that risk is the first thing 00:00:34.380 --> 00:00:36.560 we absolutely have to unpack. Welcome back to 00:00:36.560 --> 00:00:39.479 the deep dive. Today, we are really trying to 00:00:39.479 --> 00:00:41.460 shortcut your path to being well -informed about 00:00:41.460 --> 00:00:44.479 AI and algorithmic trading. Our source material 00:00:44.479 --> 00:00:47.159 is this meticulous multi -test analysis, and 00:00:47.159 --> 00:00:49.380 it's focused solely on Gemini 3's capability 00:00:49.380 --> 00:00:51.560 with Pinescript. That's the core language for 00:00:51.560 --> 00:00:53.939 TradingView. Yeah, and our mission today is just 00:00:53.939 --> 00:00:56.079 clarity. We want to show you exactly where the 00:00:56.079 --> 00:00:59.039 AI's new deep -thinking abilities really shine, 00:00:59.439 --> 00:01:01.579 like converting a visual indicator into an automated 00:01:01.579 --> 00:01:04.400 backdustful strategy. It's amazing at that. And 00:01:04.400 --> 00:01:06.540 critically, we're going to show you where it 00:01:06.540 --> 00:01:09.719 fails. I mean where it feels... spectacularly 00:01:09.719 --> 00:01:12.400 when it tries to invent a new strategy from scratch 00:01:12.400 --> 00:01:15.159 or you know fix a fundamentally flawed financial 00:01:15.159 --> 00:01:17.180 idea. Okay let's start under the hood then. I 00:01:17.180 --> 00:01:20.420 mean the chatter is that Gemini 3 is a huge step 00:01:20.420 --> 00:01:23.180 up from older models. For someone who might just 00:01:23.180 --> 00:01:27.120 be using say standard chat GPT -4 what's the 00:01:27.120 --> 00:01:29.739 core shift here? What makes this new model so 00:01:29.739 --> 00:01:31.680 relevant? It's all about what they call reasoning 00:01:31.680 --> 00:01:34.819 abilities. The older AIs they essentially looked 00:01:34.819 --> 00:01:38.519 at a prompt and tried to jump straight to the 00:01:38.519 --> 00:01:40.319 answer. They were just guessing. They were guessing. 00:01:40.599 --> 00:01:43.239 And sometimes very accurately. Like a high school 00:01:43.239 --> 00:01:45.359 student skipping the work to just circle the 00:01:45.359 --> 00:01:47.760 multiple choice answer. Exactly. The new AI, 00:01:48.040 --> 00:01:51.500 Gemini 3, it actually pauses first. It writes 00:01:51.500 --> 00:01:54.260 an internal plan before it generates the final 00:01:54.260 --> 00:01:56.819 output. A plan? They call it deep thinking. It 00:01:56.819 --> 00:01:59.819 makes the AI agentic, which just means it constructs 00:01:59.819 --> 00:02:02.859 this precise step -by -step list of what it needs 00:02:02.859 --> 00:02:04.959 to do. And for something like Pinescript, where 00:02:04.959 --> 00:02:07.540 a single missed bracket can cost you thousands. 00:02:07.959 --> 00:02:10.400 That planning skill is not just critical. It's 00:02:10.400 --> 00:02:13.060 essential. So the plan is the difference. And 00:02:13.060 --> 00:02:15.139 you mentioned file handling, this ability to 00:02:15.139 --> 00:02:18.319 upload images, CSVs, and video all at the same 00:02:18.319 --> 00:02:21.340 time. How does that change the game when you're 00:02:21.340 --> 00:02:23.319 analyzing trading history, which is usually just 00:02:23.319 --> 00:02:26.060 a long, boring spreadsheet? Oh, it's a massive 00:02:26.060 --> 00:02:28.439 efficiency boost. I mean, imagine being able 00:02:28.439 --> 00:02:32.180 to upload years of raw price data, a screenshot 00:02:32.180 --> 00:02:34.860 of a chart setup you like, and the old code of 00:02:34.860 --> 00:02:38.039 an indicator. All at once. All at once. The AI 00:02:38.039 --> 00:02:40.500 saves you hours of trying to explain all these 00:02:40.500 --> 00:02:43.240 visual concepts just through text. Which brings 00:02:43.240 --> 00:02:46.120 us to the famous context window, the memory. 00:02:46.259 --> 00:02:48.500 Can you define that simply for us? The context 00:02:48.500 --> 00:02:51.479 window is basically the AI's short -term memory. 00:02:51.939 --> 00:02:53.699 It's how much information you can hold at one 00:02:53.699 --> 00:02:56.060 time in a conversation without forgetting what 00:02:56.060 --> 00:02:57.719 you said at the beginning. So if a competitor 00:02:57.719 --> 00:03:00.560 like Claude can hold maybe a medium -sized book's 00:03:00.560 --> 00:03:02.719 worth of info. Gemini 3 is carrying around a 00:03:02.719 --> 00:03:05.099 massive library. We're talking up to two million 00:03:05.099 --> 00:03:08.120 tokens. Two million. It is a truly staggering 00:03:08.120 --> 00:03:10.520 amount of information. Two million tokens, though. 00:03:10.520 --> 00:03:12.520 I mean, that sounds like massive overkill for 00:03:12.520 --> 00:03:15.599 trading data. Was that huge memory capacity actually 00:03:15.599 --> 00:03:18.039 necessary in these PineScript tests, or is it 00:03:18.039 --> 00:03:21.039 more of a flex? It's a great question. For short 00:03:21.039 --> 00:03:25.060 scripts, no, it's not strictly necessary. But 00:03:25.060 --> 00:03:27.280 what it fundamentally unlocks for traders is 00:03:27.280 --> 00:03:30.500 huge. You can feed the AI 10 years of minute 00:03:30.500 --> 00:03:32.379 -by -minute price data without it forgetting 00:03:32.379 --> 00:03:35.340 the start. And that's vital for spotting those 00:03:35.340 --> 00:03:37.620 long -term macro trends that really impact your 00:03:37.620 --> 00:03:40.000 profit. So if we connect this to the bigger picture, 00:03:40.159 --> 00:03:42.919 it just means the AI can absorb way more granular 00:03:42.919 --> 00:03:45.719 history to inform its coding decisions. Precisely. 00:03:45.860 --> 00:03:48.280 It gives the AI more data points to work with 00:03:48.280 --> 00:03:50.319 for that internal plan we talked about. Let's 00:03:50.319 --> 00:03:52.699 move to the success story then, test one, converting 00:03:52.699 --> 00:03:55.699 an indicator into a full strategy. First, what's 00:03:55.699 --> 00:03:57.740 the actual difference between an indicator and 00:03:57.740 --> 00:04:00.199 a strategy? It's pretty simple. An indicator 00:04:00.199 --> 00:04:02.280 is just informational. It shows you arrows or 00:04:02.280 --> 00:04:05.199 lines on the chart based on some math. It's visual 00:04:05.199 --> 00:04:07.879 guidance. OK. But a strategy lets you automate 00:04:07.879 --> 00:04:10.000 those decisions. And this is the crucial part. 00:04:10.180 --> 00:04:12.509 It lets you do back testing. running the system 00:04:12.509 --> 00:04:14.949 against historical data to see how it would have 00:04:14.949 --> 00:04:17.769 performed. The testing process here, it really 00:04:17.769 --> 00:04:20.069 highlights the rigor you need. It wasn't just 00:04:20.069 --> 00:04:23.620 dumping a file in. The prep was intense. You 00:04:23.620 --> 00:04:27.339 had to get clean CSV price data, use that mandatory 00:04:27.339 --> 00:04:30.920 ISO time format, and critically turn off every 00:04:30.920 --> 00:04:33.740 single other indicator on the chart before exporting. 00:04:33.980 --> 00:04:36.220 Oh, that level of rigor is just non -negotiable. 00:04:36.480 --> 00:04:39.019 If you have messy data going in, you get unusable 00:04:39.019 --> 00:04:41.579 code coming out. But what's fascinating is the 00:04:41.579 --> 00:04:43.959 prompt they used, they gave Gemini a specific 00:04:43.959 --> 00:04:46.839 role expert Pinescript coder, a really clear 00:04:46.839 --> 00:04:51.060 task, and extremely strict rules. And these rules 00:04:51.060 --> 00:04:53.500 in ensured a dose of realism, right? Setting 00:04:53.500 --> 00:04:56.660 initial equity at $1 ,000, mandating 100 % equity 00:04:56.660 --> 00:05:00.319 usage, and forcing in a realistic 0 .1 % commission 00:05:00.319 --> 00:05:02.860 fee on every trade. And the AI handled it all 00:05:02.860 --> 00:05:05.160 perfectly. It wrote a usable, compliant script. 00:05:05.399 --> 00:05:07.339 The numbers on Bitcoin were just immense. The 00:05:07.339 --> 00:05:10.980 BTC results showed a total profit of 2 ,497%. 00:05:10.980 --> 00:05:12.879 Yeah. You see that number, and you immediately 00:05:12.879 --> 00:05:14.279 think you've cracked the code. This is where 00:05:14.279 --> 00:05:16.680 we have to pause. The massive asterisk here is 00:05:16.680 --> 00:05:19.959 the max drawdown. It hit 46%. And we need to 00:05:19.959 --> 00:05:22.319 linger on that, because drawdown is that scary 00:05:22.319 --> 00:05:24.939 number. It's what separates academic success 00:05:24.939 --> 00:05:28.459 from real -world trading. Explain drawdown again, 00:05:28.540 --> 00:05:30.740 just really clearly. Drawdown is the largest 00:05:30.740 --> 00:05:33.540 peak to trough drop you've experienced. So if 00:05:33.540 --> 00:05:36.600 your $10 ,000 account hit a high point of, say, 00:05:36.600 --> 00:05:41.240 $20 ,000, a 46 % drawdown means that $20 ,000 00:05:41.240 --> 00:05:43.540 balance later dropped all the way down to $10 00:05:43.540 --> 00:05:46.879 ,800 before it recovered. That kind of volatility 00:05:46.879 --> 00:05:49.259 is just devastating for a person. And the risk 00:05:49.259 --> 00:05:51.899 profile got even more terrifying with something 00:05:51.899 --> 00:05:55.660 like Solana. The profit there soared to 50 ,000%, 00:05:55.660 --> 00:05:58.639 but the drawdown ballooned to 66%. Yeah. Wait, 00:05:58.980 --> 00:06:01.259 66 %? That's not a dip. That's blowing up most 00:06:01.259 --> 00:06:03.980 trading accounts. Is the AI just endorsing high 00:06:03.980 --> 00:06:06.779 -spakes gambling? Well, the AI is simply following 00:06:06.779 --> 00:06:09.139 the math of the indicator perfectly. It doesn't 00:06:09.139 --> 00:06:11.079 care about the emotional or practical consequences 00:06:11.079 --> 00:06:13.759 of that volatility. I still wrestle with prompt 00:06:13.759 --> 00:06:17.339 drift myself, and seeing that 66 % drawdown is 00:06:17.339 --> 00:06:19.439 a real -world wake -up call, even when the code 00:06:19.439 --> 00:06:22.519 itself is technically perfect. So beyond the 00:06:22.519 --> 00:06:25.139 success of the code itself, what lesson did that 00:06:25.139 --> 00:06:28.439 extreme drawdown teach us about relying on the 00:06:28.439 --> 00:06:31.829 AI? That the AI writes the code perfectly. But 00:06:31.829 --> 00:06:34.670 the human must still understand and manage that 00:06:34.670 --> 00:06:37.290 extreme account -crushing risk. That's the takeaway. 00:06:37.550 --> 00:06:40.170 Okay, let's pivot to the failure test, the flawed 00:06:40.170 --> 00:06:43.100 logic trap. The goal here was to take a working 00:06:43.100 --> 00:06:46.199 long -only strategy and try to improve it by 00:06:46.199 --> 00:06:48.740 adding shorting logic. You know, profiting when 00:06:48.740 --> 00:06:50.800 the price drops. Right. We specifically modified 00:06:50.800 --> 00:06:53.319 the Don Chin channel strategy. We asked Gemini3 00:06:53.319 --> 00:06:55.779 to short when the price closed below the lower 00:06:55.779 --> 00:06:58.060 band and then close that short when the price 00:06:58.060 --> 00:06:59.839 got back to the middle line. It sounds logical. 00:06:59.959 --> 00:07:02.579 It sounds totally logical. But the code executed 00:07:02.579 --> 00:07:05.060 and the results were disastrous. The original 00:07:05.060 --> 00:07:08.399 strategy was already delivering 2 ,700 % profit. 00:07:08.680 --> 00:07:11.160 And Gemini's improved version. It dropped down 00:07:11.160 --> 00:07:14.439 to a dismal 150 % profit. Just massive losses 00:07:14.439 --> 00:07:16.279 were introduced by this improvement. So we have 00:07:16.279 --> 00:07:19.139 to analyze the why here. This is where that agentic 00:07:19.139 --> 00:07:22.139 reasoning just failed. It completely failed to 00:07:22.139 --> 00:07:26.259 override a flawed financial premise. The AI wrote 00:07:26.259 --> 00:07:29.160 the code for those short signals perfectly. Flawlessly. 00:07:29.579 --> 00:07:32.000 But it failed to consider the overwhelming upward 00:07:32.000 --> 00:07:34.639 bias of the crypto market. It missed the domain 00:07:34.639 --> 00:07:38.060 expertise. It did. It prioritized coding execution 00:07:38.060 --> 00:07:40.819 over a sanity check of how the market actually 00:07:40.819 --> 00:07:44.680 works. Exactly. Crypto historically is long biased 00:07:44.680 --> 00:07:48.699 because of adoption, scarcity, inflation dynamics. 00:07:49.100 --> 00:07:51.740 Shorting is inherently a low probability bet 00:07:51.740 --> 00:07:54.860 over the long term. Gemini 3, despite all this 00:07:54.860 --> 00:07:56.699 deep thinking, did not act like a seasoned trader. 00:07:56.759 --> 00:07:58.949 It was just a brilliant code monkey. And it's 00:07:58.949 --> 00:08:00.949 important to note this wasn't just a Gemini failure. 00:08:01.470 --> 00:08:04.529 Grok and ChatGPT also failed this exact test. 00:08:04.870 --> 00:08:07.370 Even Claude couldn't beat the simple long -only 00:08:07.370 --> 00:08:09.250 version. Right, and it raises this really important 00:08:09.250 --> 00:08:11.990 question. Since the AI is supposed to be agentic, 00:08:12.370 --> 00:08:14.790 why didn't its internal plan include a step to 00:08:14.790 --> 00:08:16.430 challenge instructions that were mathematically 00:08:16.430 --> 00:08:18.970 guaranteed to lower profit? And the short answer 00:08:18.970 --> 00:08:22.649 is it prioritizes accurate code execution over 00:08:22.649 --> 00:08:25.410 challenging the user's potentially flawed financial 00:08:25.410 --> 00:08:27.990 premise. We're going to pause right there. When 00:08:27.990 --> 00:08:29.949 we come back, we'll look at why the ultimate 00:08:29.949 --> 00:08:32.509 dream prompt, the one everyone wants to run, 00:08:33.090 --> 00:08:36.470 failed spectacularly, and which AI tool is actually 00:08:36.470 --> 00:08:39.210 the undisputed champion for heavy -duty reliable 00:08:39.210 --> 00:08:43.190 coding. Welcome back to the deep dive. We're 00:08:43.190 --> 00:08:46.210 now diving into where Gemini 3 hits its computational 00:08:46.210 --> 00:08:49.649 and I guess ethical limitations. The first two 00:08:49.649 --> 00:08:51.909 tests were practical, but test 3 was aiming for 00:08:51.909 --> 00:08:54.659 the moon. It really was. The dream prompt Test 00:08:54.659 --> 00:08:57.580 3 was simple. Find a pattern. Make me a strategy 00:08:57.580 --> 00:09:00.440 that wins with drawdown under 30%. With no indicators 00:09:00.440 --> 00:09:02.440 mentioned. No indicators. Just the raw price 00:09:02.440 --> 00:09:04.779 data. This is what every trader wants. New alpha 00:09:04.779 --> 00:09:06.980 discovery. So detail the result. What happened 00:09:06.980 --> 00:09:09.059 when you asked it for a new profitable pattern? 00:09:09.360 --> 00:09:11.799 Gemini 3's deep thinking icon, it's bunned for 00:09:11.799 --> 00:09:13.919 a while. It started the analysis, trying to write 00:09:13.919 --> 00:09:16.279 an internal plan to process all this data. And 00:09:16.279 --> 00:09:19.220 then it just crashed. Time down. And you tried 00:09:19.220 --> 00:09:21.039 again. I tried again thinking it was a fluke. 00:09:21.039 --> 00:09:24.000 It crashed again. Why? I mean if it has all this 00:09:24.000 --> 00:09:26.639 memory and this capacity to run a step -by -step 00:09:26.639 --> 00:09:29.759 plan, why did it give up on the most fundamental 00:09:29.759 --> 00:09:32.259 task of a data analyst? Because the task was 00:09:32.259 --> 00:09:35.799 just too vast. It was too vague. For the AI to 00:09:35.799 --> 00:09:39.379 truly find a new profitable pattern, it has to 00:09:39.379 --> 00:09:42.419 check billions of permutations. Billions. It 00:09:42.419 --> 00:09:44.279 would need to test every combination of moving 00:09:44.279 --> 00:09:47.340 averages, every oscillator setting, every volume 00:09:47.340 --> 00:09:50.039 threshold. against the entire data set. Whoa! 00:09:50.279 --> 00:09:53.539 Imagine scaling that to a billion queries. That 00:09:53.539 --> 00:09:56.440 level of computation just exceeds the AI's time 00:09:56.440 --> 00:09:58.559 limit for a single response. It's usually capped 00:09:58.559 --> 00:10:01.529 at around 60 seconds. It's not that it can't 00:10:01.529 --> 00:10:03.970 do it in theory, but it can't finish it within 00:10:03.970 --> 00:10:05.750 the constraints of the chat model interface. 00:10:05.950 --> 00:10:08.169 It needs direction. So the lesson here is you 00:10:08.169 --> 00:10:10.710 can't just ask it to find a strategy. No, you 00:10:10.710 --> 00:10:12.350 have to provide boundaries. You have to guide 00:10:12.350 --> 00:10:15.009 it. Tell it, use the RSI indicator combined with 00:10:15.009 --> 00:10:17.950 a 50 -day moving average or look for specific 00:10:17.950 --> 00:10:20.830 breakout price action. Specificity prevents the 00:10:20.830 --> 00:10:23.110 algorithmic overload. And the timeout. Exactly. 00:10:23.409 --> 00:10:26.029 That's a crucial operational lesson. Let's look 00:10:26.029 --> 00:10:29.870 at the final test, the ethical barrier. Asking 00:10:29.870 --> 00:10:33.210 the AI to reverse engineer a closed source indicator. 00:10:33.730 --> 00:10:35.769 This is a really common desire in the trading 00:10:35.769 --> 00:10:38.809 community. People pay hundreds for indicators 00:10:38.809 --> 00:10:41.730 with locked code. Right. We asked Gemini3 to 00:10:41.730 --> 00:10:44.070 look at screenshots of the indicators lines and 00:10:44.070 --> 00:10:46.830 raw data values and then try to recreate the 00:10:46.830 --> 00:10:48.409 underlying PineScript math. And the immediate 00:10:48.409 --> 00:10:52.409 response was? An immediate refusal. Gemini3 said 00:10:52.409 --> 00:10:55.029 very politely, I am having a hard time fulfilling 00:10:55.029 --> 00:10:56.830 your request and then it just stopped trying. 00:10:56.990 --> 00:10:59.450 What were the reasons for that besides just the 00:10:59.450 --> 00:11:01.570 sheer complexity of it? It's primarily the safety 00:11:01.570 --> 00:11:04.330 filters that Google put in place. There's a real 00:11:04.330 --> 00:11:06.629 fear of intellectual property theft or, you know, 00:11:06.909 --> 00:11:09.250 similar nefarious uses. And recreating proprietary 00:11:09.250 --> 00:11:11.309 code would trigger that. It's exactly the kind 00:11:11.309 --> 00:11:13.330 of task that triggers those strong safety locks. 00:11:13.470 --> 00:11:15.549 What's so fascinating here is the contrast with 00:11:15.549 --> 00:11:18.100 the competition. Absolutely. Yeah. Claude has 00:11:18.100 --> 00:11:20.539 succeeded in performing similar reverse engineering 00:11:20.539 --> 00:11:23.580 tasks before. So this tells us Gemini 3 either 00:11:23.580 --> 00:11:26.620 has much stricter filters, or it just has less 00:11:26.620 --> 00:11:28.879 patience for that complexity once the IP filter 00:11:28.879 --> 00:11:31.159 gets triggered. So we have a pretty clear dividing 00:11:31.159 --> 00:11:33.820 line now. What is the core difference between 00:11:33.820 --> 00:11:36.840 a successful prompt, like the conversion, and 00:11:36.840 --> 00:11:39.240 a timeout or crash prompt, like finding a new 00:11:39.240 --> 00:11:41.899 strategy? Success requires specific variables 00:11:41.899 --> 00:11:44.299 and strict boundaries. Vagueness leads to the 00:11:44.299 --> 00:11:46.340 AI timing out because the computation just becomes 00:11:46.340 --> 00:11:48.480 too massive. Okay, let's summarize the final 00:11:48.480 --> 00:11:51.000 rankings. After all this testing conversion, 00:11:51.419 --> 00:11:54.889 flawed logic, vague tasks... Where do the three 00:11:54.889 --> 00:11:57.850 major AI players stand for Pinescript crypto 00:11:57.850 --> 00:12:00.289 coding? Yeah, we have a clear hierarchy for utility 00:12:00.289 --> 00:12:03.250 here. Start with Gemini 3, our focus today. Gemini 00:12:03.250 --> 00:12:05.789 3 is best for absolute beginners and for simple 00:12:05.789 --> 00:12:08.889 indicator conversions. The pros? It's fast, the 00:12:08.889 --> 00:12:10.970 interface is clean, and that multi -file upload 00:12:10.970 --> 00:12:13.429 is superb. And the cons? The cons. It gives up 00:12:13.429 --> 00:12:16.490 too easily on hard tasks and just have poor endurance 00:12:16.490 --> 00:12:19.389 for that long computational deep thinking. And 00:12:19.389 --> 00:12:21.490 Grok? Grok is the tool for analyzing real -time 00:12:21.490 --> 00:12:24.970 news. And that's simply because it has that great 00:12:24.970 --> 00:12:28.730 access to ex -Twitter data for sentiment analysis. 00:12:28.950 --> 00:12:31.289 But for coding? For pure coding rigor, it's average. 00:12:31.769 --> 00:12:34.049 The code can be pretty sloppy. Which leaves the 00:12:34.049 --> 00:12:36.110 undisputed winner for the heavy lifting, the 00:12:36.110 --> 00:12:38.669 one that acts like a true senior developer. That 00:12:38.669 --> 00:12:42.549 is Claude 3 .5 Sonnet or Opus. It is the winner 00:12:42.549 --> 00:12:45.299 for complex coding. When you give Claude a hard 00:12:45.299 --> 00:12:47.639 task, it just keeps iterating. It keeps working. 00:12:48.220 --> 00:12:51.259 It writes an incredibly long, clean code. And 00:12:51.259 --> 00:12:53.320 if you paste in an error message, it acts like 00:12:53.320 --> 00:12:56.059 a patient developer and just fixes it without 00:12:56.059 --> 00:12:58.299 you needing to start all over. The only real 00:12:58.299 --> 00:13:00.480 downside there is the lower message limit, which 00:13:00.480 --> 00:13:02.980 forces you to pay if you're doing a massive workload. 00:13:03.179 --> 00:13:06.360 Correct. But for high quality, complex code generation 00:13:06.360 --> 00:13:08.700 that requires that kind of perseverance, it's 00:13:08.700 --> 00:13:10.840 still superior. So the final practical advice 00:13:10.840 --> 00:13:14.379 is this. Use Gemini 3 to brainstorm ideas, maybe 00:13:14.379 --> 00:13:16.919 clean up your data, but then rely on Claude when 00:13:16.919 --> 00:13:19.539 you need that complex, final, robust code written 00:13:19.539 --> 00:13:22.340 for your TradingView bot. Exactly. And let's 00:13:22.340 --> 00:13:24.019 recap the core lesson for you, the listener. 00:13:24.700 --> 00:13:27.220 AI is an exceptional conversion tool for code. 00:13:27.259 --> 00:13:29.860 It can save you hours of manual work. But it 00:13:29.860 --> 00:13:32.399 is not capable of creating new profitable logic 00:13:32.399 --> 00:13:35.820 from scratch or fixing a fundamentally flawed 00:13:35.820 --> 00:13:38.879 financial idea that just contradicts basic market 00:13:38.879 --> 00:13:42.379 behavior. The core lesson is clear. AI is a tool, 00:13:42.519 --> 00:13:45.779 not a boss. You are the traitor. That's the mandate. 00:13:46.379 --> 00:13:48.639 Remember, always check that max drawdown number. 00:13:48.879 --> 00:13:51.259 Never trust the code blindly, even if the profit 00:13:51.259 --> 00:13:54.340 percentage looks amazing. And if the AI prioritizes 00:13:54.340 --> 00:13:56.659 perfect code execution over self -preservation, 00:13:56.980 --> 00:14:00.580 as we saw in that terrifying 66 % drawdown, what 00:14:00.580 --> 00:14:02.720 non -financial risks are we blindly automating 00:14:02.720 --> 00:14:04.799 in other complex fields just because the generated 00:14:04.799 --> 00:14:07.580 code works? What happens when the AI optimizes 00:14:07.580 --> 00:14:09.460 for a goal that we haven't properly checked for 00:14:09.460 --> 00:14:11.120 catastrophic failure? Something for you to mull 00:14:11.120 --> 00:14:12.500 over. Until next time. Take care.