PyPI - content-core - Versions diffs - 0.4.0__py3-none-any.whl → 0.5.1__py3-none-any.whl - Mend

content-core 0.4.0py3-none-any.whl → 0.5.1py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.

This version of content-core might be problematic. Click here for more details.

Files changed (15) hide show

content_core/__init__.py +6 -2
content_core/cc_config.yaml +35 -0
content_core/common/state.py +4 -0
content_core/config.py +23 -4
content_core/content/extraction/graph.py +15 -1
content_core/notebooks/docling.ipynb +27 -0
content_core/notebooks/run.ipynb +74 -58
content_core/processors/docling.py +72 -0
content_core/templated_message.py +16 -24
{content_core-0.4.0.dist-info → content_core-0.5.1.dist-info}/METADATA +56 -2
{content_core-0.4.0.dist-info → content_core-0.5.1.dist-info}/RECORD +14 -12
content_core/prompter.py +0 -159
{content_core-0.4.0.dist-info → content_core-0.5.1.dist-info}/WHEEL +0 -0
{content_core-0.4.0.dist-info → content_core-0.5.1.dist-info}/entry_points.txt +0 -0
{content_core-0.4.0.dist-info → content_core-0.5.1.dist-info}/licenses/LICENSE +0 -0

content_core/notebooks/run.ipynb CHANGED Viewed

@@ -56,7 +56,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [
     {
@@ -64,60 +64,7 @@
      "output_type": "stream",
      "text": [
       "title='How to Use Chat GPT For Finance in 2025 (Beginners Guide)' source_type='url' identified_type='youtube' identified_provider='' metadata={'video_id': 'lLprprtHfts', 'transcript': [{'text': 'this is going to be a completely free', 'start': 0.16, 'duration': 4.88}, {'text': 'course on how you can use JGPT if you', 'start': 2.399, 'duration': 4.96}, {'text': 'work in finance Over the last few years', 'start': 5.04, 'duration': 3.92}, {'text': \"I've taught thousands of finance\", 'start': 7.359, 'duration': 4.32}, {'text': 'professionals like you how to use AI to', 'start': 8.96, 'duration': 4.88}, {'text': 'save hours but also to become', 'start': 11.679, 'duration': 4.801}, {'text': 'irrepressible in this time of changes', 'start': 13.84, 'duration': 4.4}, {'text': 'especially with all of the pressure we', 'start': 16.48, 'duration': 3.84}, {'text': 'got in finance Now for the information', 'start': 18.24, 'duration': 4.48}, {'text': 'in this video I normally charge hundreds', 'start': 20.32, 'duration': 5.119}, {'text': 'of dollars but I decided to give it away', 'start': 22.72, 'duration': 5.2}, {'text': 'for free So close all of your open tabs', 'start': 25.439, 'duration': 4.401}, {'text': 'and pay full attention because I', 'start': 27.92, 'duration': 3.84}, {'text': 'guarantee that if you watch this video', 'start': 29.84, 'duration': 4.239}, {'text': 'until the end you will have all of the', 'start': 31.76, 'duration': 5.76}, {'text': 'foundations you need to start using CHPT', 'start': 34.079, 'duration': 5.601}, {'text': 'in your work In this course I will show', 'start': 37.52, 'duration': 5.12}, {'text': 'you how you can sign up to Chip and walk', 'start': 39.68, 'duration': 5.52}, {'text': 'you through all of the functionalities', 'start': 42.64, 'duration': 4.96}, {'text': 'that are the most important in Chip And', 'start': 45.2, 'duration': 4.879}, {'text': 'then we will cover the risks and', 'start': 47.6, 'duration': 5.52}, {'text': 'limitations of using CHP in your work', 'start': 50.079, 'duration': 5.281}, {'text': 'And finally you will learn how to write', 'start': 53.12, 'duration': 5.04}, {'text': 'AI prompts like a pro which is arguably', 'start': 55.36, 'duration': 5.12}, {'text': 'the most valuable skills you should have', 'start': 58.16, 'duration': 4.559}, {'text': 'right', 'start': 60.48, 'duration': 2.239}, {'text': 'now We are going to learn together how', 'start': 63.0, 'duration': 5.72}, {'text': 'you can sign up and also what are the', 'start': 66.159, 'duration': 5.121}, {'text': 'functionalities of CHP but really', 'start': 68.72, 'duration': 4.88}, {'text': 'important we will cover what are the', 'start': 71.28, 'duration': 5.04}, {'text': 'risks and limitations and then we you', 'start': 73.6, 'duration': 5.12}, {'text': 'will learn how to prompt the most', 'start': 76.32, 'duration': 4.799}, {'text': 'important skill to learn in this course', 'start': 78.72, 'duration': 4.399}, {'text': \"And finally you'll get an exercise to\", 'start': 81.119, 'duration': 4.241}, {'text': 'test your capabilities If you are new to', 'start': 83.119, 'duration': 5.401}, {'text': 'chatgptt you can go to', 'start': 85.36, 'duration': 5.28}, {'text': 'chat.openai.com and you will arrive to', 'start': 88.52, 'duration': 4.36}, {'text': 'this page On this page you will click', 'start': 90.64, 'duration': 4.799}, {'text': 'sign up They will ask you to create an', 'start': 92.88, 'duration': 4.64}, {'text': 'account You can either use your Google', 'start': 95.439, 'duration': 4.161}, {'text': 'credential or Microsoft credential or', 'start': 97.52, 'duration': 4.08}, {'text': 'even Apple or just give your email', 'start': 99.6, 'duration': 4.32}, {'text': \"address Let's imagine that I want to\", 'start': 101.6, 'duration': 4.4}, {'text': 'choose my email address Then I just need', 'start': 103.92, 'duration': 4.32}, {'text': 'to put a password', 'start': 106.0, 'duration': 4.0}, {'text': 'Now that I have entered my password I', 'start': 108.24, 'duration': 4.32}, {'text': 'will just need to validate my email You', 'start': 110.0, 'duration': 4.32}, {'text': 'should arrive to this email address', 'start': 112.56, 'duration': 3.28}, {'text': 'where you just click verify email', 'start': 114.32, 'duration': 4.399}, {'text': 'address And now your email is verified', 'start': 115.84, 'duration': 5.52}, {'text': 'You just need to login again Here OpenAI', 'start': 118.719, 'duration': 4.521}, {'text': 'will ask you some', 'start': 121.36, 'duration': 4.92}, {'text': 'details And you can enter also your', 'start': 123.24, 'duration': 5.799}, {'text': 'birthday And now we are successfully', 'start': 126.28, 'duration': 5.56}, {'text': \"inside Chpt for the first time Let's see\", 'start': 129.039, 'duration': 6.241}, {'text': 'now how you can use CH GPT Here is the', 'start': 131.84, 'duration': 5.28}, {'text': 'chat box where you are going to start', 'start': 135.28, 'duration': 4.72}, {'text': 'entering your questions and instructions', 'start': 137.12, 'duration': 5.92}, {'text': 'For example I can ask what are the top', 'start': 140.0, 'duration': 6.72}, {'text': 'three skills for an FPNA manager and how', 'start': 143.04, 'duration': 6.0}, {'text': \"to learn them Now I'm going to click\", 'start': 146.72, 'duration': 6.04}, {'text': 'here or you can press alt', 'start': 149.04, 'duration': 6.88}, {'text': 'enter and you get from chaptt your', 'start': 152.76, 'duration': 5.479}, {'text': \"answer Now let's imagine that I don't\", 'start': 155.92, 'duration': 5.36}, {'text': 'want three skills but five skills And I', 'start': 158.239, 'duration': 7.64}, {'text': 'can click this little pen to correct my', 'start': 161.28, 'duration': 6.88}, {'text': 'instruction And I can ask top five', 'start': 165.879, 'duration': 4.921}, {'text': 'skills And I can ask \"Show me the result', 'start': 168.16, 'duration': 5.12}, {'text': 'in a', 'start': 170.8, 'duration': 2.48}, {'text': 'table.\" And now I get the result in a', 'start': 174.04, 'duration': 5.479}, {'text': 'table What you can also do is press this', 'start': 176.319, 'duration': 5.161}, {'text': 'copy and paste', 'start': 179.519, 'duration': 4.321}, {'text': 'button And when you do that it copies', 'start': 181.48, 'duration': 4.679}, {'text': 'all of the text Here for example I will', 'start': 183.84, 'duration': 4.8}, {'text': 'just paste it and I have all of the', 'start': 186.159, 'duration': 4.961}, {'text': 'tests But as you can note the text is', 'start': 188.64, 'duration': 5.04}, {'text': 'not formatted So what you can do is', 'start': 191.12, 'duration': 6.36}, {'text': 'select the table like this and copy and', 'start': 193.68, 'duration': 7.199}, {'text': 'paste the table This is my favorite', 'start': 197.48, 'duration': 5.399}, {'text': 'function to copy and paste is just', 'start': 200.879, 'duration': 4.08}, {'text': 'selecting the text rather than using', 'start': 202.879, 'duration': 5.36}, {'text': 'this There is another function which is', 'start': 204.959, 'duration': 5.521}, {'text': \"regenerate Here if I'm not happy about\", 'start': 208.239, 'duration': 6.64}, {'text': 'the result I can just click regenerate', 'start': 210.48, 'duration': 8.0}, {'text': 'And JGPT will propose something which is', 'start': 214.879, 'duration': 5.92}, {'text': 'an alternative to my question And I can', 'start': 218.48, 'duration': 5.759}, {'text': 'come back and again here change and I', 'start': 220.799, 'duration': 7.841}, {'text': 'can say what are the top technical', 'start': 224.239, 'duration': 4.401}, {'text': 'skills and save and submit and you can', 'start': 229.319, 'duration': 5.881}, {'text': 'see what is happening I have now', 'start': 232.799, 'duration': 5.241}, {'text': 'technical skills but look at what is', 'start': 235.2, 'duration': 6.239}, {'text': 'here Here is a way to come back to all', 'start': 238.04, 'duration': 6.199}, {'text': 'of my previous instructions Like this I', 'start': 241.439, 'duration': 4.481}, {'text': 'can come back to the history If you want', 'start': 244.239, 'duration': 3.681}, {'text': 'to start a new chat then you will click', 'start': 245.92, 'duration': 5.679}, {'text': 'here And now I can ask tell me how is a', 'start': 247.92, 'duration': 9.239}, {'text': 'typical day for FPNA manager', 'start': 251.599, 'duration': 5.56}, {'text': \"It's really important to change\", 'start': 257.199, 'duration': 4.16}, {'text': 'discussions each time you are changing', 'start': 259.12, 'duration': 4.639}, {'text': \"subjects because if you don't do that\", 'start': 261.359, 'duration': 7.12}, {'text': 'then here if I ask what are the top', 'start': 263.759, 'duration': 4.72}, {'text': \"skills it will understand that I'm\", 'start': 269.32, 'duration': 5.0}, {'text': 'talking about FPNA manager and not FPNA', 'start': 271.52, 'duration': 4.64}, {'text': 'analyst because we started the', 'start': 274.32, 'duration': 4.56}, {'text': 'discussion by FPNA manager and if my', 'start': 276.16, 'duration': 5.039}, {'text': \"question is about FPNA analyst it's\", 'start': 278.88, 'duration': 4.96}, {'text': 'either better to change here and to say', 'start': 281.199, 'duration': 5.521}, {'text': 'for FPN analyst or directly to change a', 'start': 283.84, 'duration': 5.52}, {'text': 'new discussion and start from scratch if', 'start': 286.72, 'duration': 4.72}, {'text': 'you feel the two discussions are not', 'start': 289.36, 'duration': 4.72}, {'text': 'connected The last important part is', 'start': 291.44, 'duration': 3.56}, {'text': 'this', 'start': 294.08, 'duration': 4.48}, {'text': 'sidebar I will click to see the sidebar', 'start': 295.0, 'duration': 6.24}, {'text': 'and to come back to all of my previous', 'start': 298.56, 'duration': 7.0}, {'text': 'discussions And there I can click', 'start': 301.24, 'duration': 7.959}, {'text': 'to manage my settings For example I can', 'start': 305.56, 'duration': 6.12}, {'text': 'ask to save or not the chat history that', 'start': 309.199, 'duration': 4.641}, {'text': 'we used to at the left And also I can', 'start': 311.68, 'duration': 4.56}, {'text': 'activate or deactivate the training on', 'start': 313.84, 'duration': 4.96}, {'text': 'my data because for me I have subscribed', 'start': 316.24, 'duration': 6.799}, {'text': \"to CHP teams It's by default\", 'start': 318.8, 'duration': 4.239}, {'text': \"deactivated Finally let's look together\", 'start': 323.8, 'duration': 5.8}, {'text': 'at the most advanced functions of JPT', 'start': 326.4, 'duration': 6.96}, {'text': 'You will see that with JGPT you can also', 'start': 329.6, 'duration': 6.56}, {'text': 'upload a file So you have the button in', 'start': 333.36, 'duration': 5.92}, {'text': \"the bottom here attachments I'm going to\", 'start': 336.16, 'duration': 5.44}, {'text': 'upload a file that looks like this So', 'start': 339.28, 'duration': 5.68}, {'text': 'here is a file with my company sales', 'start': 341.6, 'duration': 7.039}, {'text': 'data and I am in a SAS industry So I', 'start': 344.96, 'duration': 5.76}, {'text': 'will want for example to do a cohort', 'start': 348.639, 'duration': 5.28}, {'text': \"analysis Let's see with JGPT if my\", 'start': 350.72, 'duration': 5.28}, {'text': \"company has a contract with JGPT and I'm\", 'start': 353.919, 'duration': 5.84}, {'text': \"allowed to upload data inside JGPT Let's\", 'start': 356.0, 'duration': 6.88}, {'text': 'see what I can do Just as a warning if', 'start': 359.759, 'duration': 4.641}, {'text': \"your company doesn't have a contract\", 'start': 362.88, 'duration': 4.24}, {'text': \"with OpenAI then check if you're allowed\", 'start': 364.4, 'duration': 5.359}, {'text': \"or not to upload data It's better to be\", 'start': 367.12, 'duration': 5.199}, {'text': 'sure than to do a mistake So we are back', 'start': 369.759, 'duration': 5.921}, {'text': \"in JGPT and I'm going to upload my file\", 'start': 372.319, 'duration': 4.88}, {'text': 'and you can see that I can connect to', 'start': 375.68, 'duration': 3.519}, {'text': 'Google Drive connect to Microsoft One', 'start': 377.199, 'duration': 4.801}, {'text': 'Drive or upload from my computer So here', 'start': 379.199, 'duration': 5.44}, {'text': \"is the file I'm uploading and I will ask\", 'start': 382.0, 'duration': 6.639}, {'text': 'can you create a cohort analysis for my', 'start': 384.639, 'duration': 8.041}, {'text': 'retention rate based on the month of the', 'start': 388.639, 'duration': 7.601}, {'text': 'subscriber starting doing business with', 'start': 392.68, 'duration': 6.359}, {'text': 'us show on top a visualization because I', 'start': 396.24, 'duration': 5.64}, {'text': 'like to visualize this', 'start': 399.039, 'duration': 6.16}, {'text': 'information So now first we can see that', 'start': 401.88, 'duration': 5.4}, {'text': 'we have here the table So that means', 'start': 405.199, 'duration': 5.201}, {'text': 'JGPT understood the table with exactly', 'start': 407.28, 'duration': 5.52}, {'text': 'the right columns It read those columns', 'start': 410.4, 'duration': 4.32}, {'text': \"using Python And it's really important\", 'start': 412.8, 'duration': 4.08}, {'text': 'to know that JGPT is not good at', 'start': 414.72, 'duration': 3.68}, {'text': \"calculating because it's not made to\", 'start': 416.88, 'duration': 3.92}, {'text': 'calculate but is really good at creating', 'start': 418.4, 'duration': 5.44}, {'text': 'the code to send them to Python And', 'start': 420.8, 'duration': 5.36}, {'text': 'Python is a good tool to calculate and', 'start': 423.84, 'duration': 4.32}, {'text': \"that's how you separate the work like\", 'start': 426.16, 'duration': 5.12}, {'text': 'you will do for yourself If you have to', 'start': 428.16, 'duration': 4.96}, {'text': 'calculate you will use a computer or you', 'start': 431.28, 'duration': 4.0}, {'text': 'will use a calculator Here is the same', 'start': 433.12, 'duration': 4.079}, {'text': 'you use the right tool to calculate', 'start': 435.28, 'duration': 5.44}, {'text': 'which is Python Now TGPT is coding the', 'start': 437.199, 'duration': 6.56}, {'text': 'code to create our cohort analysis and', 'start': 440.72, 'duration': 5.12}, {'text': 'you can see that now I have the mostly', 'start': 443.759, 'duration': 4.801}, {'text': 'retention rates and here it decided to', 'start': 445.84, 'duration': 5.28}, {'text': 'use red for something positive So what I', 'start': 448.56, 'duration': 5.12}, {'text': 'can do is even ask to change the color', 'start': 451.12, 'duration': 7.079}, {'text': 'Can you make 100% blue and 0%', 'start': 453.68, 'duration': 6.959}, {'text': \"red but you see that while it's\", 'start': 458.199, 'duration': 4.521}, {'text': 'calculating I want to show you how', 'start': 460.639, 'duration': 4.321}, {'text': 'powerful the tool is because this is', 'start': 462.72, 'duration': 5.039}, {'text': 'impossible to do so fast in Excel and in', 'start': 464.96, 'duration': 5.2}, {'text': 'Excel also you will have to work with', 'start': 467.759, 'duration': 4.56}, {'text': 'conditional formatting But here I will', 'start': 470.16, 'duration': 6.08}, {'text': 'have here my file with a proper cohort', 'start': 472.319, 'duration': 6.241}, {'text': 'analysis And what I can even do is', 'start': 476.24, 'duration': 4.88}, {'text': 'download the chart as a picture because', 'start': 478.56, 'duration': 5.12}, {'text': 'Python does it as a picture And I can', 'start': 481.12, 'duration': 5.28}, {'text': 'even here continue to analyze thanks to', 'start': 483.68, 'duration': 5.28}, {'text': 'these buttons So now for example I ask', 'start': 486.4, 'duration': 5.04}, {'text': 'to analyze the customer turn rate and it', 'start': 488.96, 'duration': 4.799}, {'text': 'will do this analysis for me And you can', 'start': 491.44, 'duration': 4.24}, {'text': \"see that now I have a graph and that's\", 'start': 493.759, 'duration': 3.761}, {'text': 'what I really like here is to continue', 'start': 495.68, 'duration': 3.76}, {'text': 'the discussion thanks to this button And', 'start': 497.52, 'duration': 4.16}, {'text': 'if you want to audit what happened you', 'start': 499.44, 'duration': 4.879}, {'text': 'can always look at the code and you can', 'start': 501.68, 'duration': 4.56}, {'text': \"always ask if you don't really\", 'start': 504.319, 'duration': 3.921}, {'text': 'understand the code you can ask JGP to', 'start': 506.24, 'duration': 3.6}, {'text': 'explain you the code like this You can', 'start': 508.24, 'duration': 5.12}, {'text': 'verify if the calculation is', 'start': 509.84, 'duration': 3.52}, {'text': 'correct This is a function that allows', 'start': 513.719, 'duration': 6.2}, {'text': 'you to customize and make your own GPT', 'start': 516.64, 'duration': 4.8}, {'text': 'Let me show you how you can do that So', 'start': 519.919, 'duration': 4.641}, {'text': 'we are back in JPT and I will go on the', 'start': 521.44, 'duration': 5.44}, {'text': 'side open the sidebar and here you', 'start': 524.56, 'duration': 5.76}, {'text': 'should see a button called explore GPTs', 'start': 526.88, 'duration': 6.959}, {'text': 'and GPTs are GPT versions that are', 'start': 530.32, 'duration': 6.16}, {'text': 'customized by either yourself or other', 'start': 533.839, 'duration': 6.241}, {'text': 'people This is the marketplace of GPS or', 'start': 536.48, 'duration': 6.96}, {'text': 'the app store of GPS So for example here', 'start': 540.08, 'duration': 6.96}, {'text': 'you can type US GAP and it will show you', 'start': 543.44, 'duration': 6.48}, {'text': 'for example here a US gap advisor where', 'start': 547.04, 'duration': 4.88}, {'text': 'you can click on you can check if there', 'start': 549.92, 'duration': 4.0}, {'text': 'are good ratings or not and if a lot of', 'start': 551.92, 'duration': 3.68}, {'text': 'people have used it you can see what it', 'start': 553.92, 'duration': 4.4}, {'text': 'does and what we can do is for example', 'start': 555.6, 'duration': 5.84}, {'text': 'start a chat and I can ask explain me', 'start': 558.32, 'duration': 5.12}, {'text': \"revenue recognition and I don't even\", 'start': 561.44, 'duration': 4.72}, {'text': 'have to explain that I want to get this', 'start': 563.44, 'duration': 4.959}, {'text': \"explanation based on US gap because I'm\", 'start': 566.16, 'duration': 4.08}, {'text': \"already in a US gap advisor So that's\", 'start': 568.399, 'duration': 4.481}, {'text': 'the big advantage This US gap advisor', 'start': 570.24, 'duration': 5.84}, {'text': 'was configured to unswear for US GAP and', 'start': 572.88, 'duration': 5.36}, {'text': 'to answer in a specific way And you can', 'start': 576.08, 'duration': 4.24}, {'text': 'see that here we got a really good', 'start': 578.24, 'duration': 4.4}, {'text': 'answer because it knows we are finance', 'start': 580.32, 'duration': 4.48}, {'text': 'people because who would want to go to a', 'start': 582.64, 'duration': 4.96}, {'text': 'US GAP advisor mostly finance people And', 'start': 584.8, 'duration': 5.2}, {'text': 'we have already somebody behind who did', 'start': 587.6, 'duration': 4.72}, {'text': 'the work to customize the elsewhere So', 'start': 590.0, 'duration': 4.24}, {'text': 'like this we save a lot of time So now', 'start': 592.32, 'duration': 4.4}, {'text': 'do you know that you can actually also', 'start': 594.24, 'duration': 4.96}, {'text': 'create your own GPT let me show you how', 'start': 596.72, 'duration': 5.04}, {'text': 'So you will go back to explore GPT to', 'start': 599.2, 'duration': 5.28}, {'text': 'the left and then upright you can create', 'start': 601.76, 'duration': 6.48}, {'text': 'your own GPT So I will click create and', 'start': 604.48, 'duration': 7.76}, {'text': 'now I am in the GPT configurator And', 'start': 608.24, 'duration': 6.08}, {'text': 'here you have two solution Either you', 'start': 612.24, 'duration': 4.719}, {'text': 'can create through a description or you', 'start': 614.32, 'duration': 5.12}, {'text': 'can configure yourself by giving a name', 'start': 616.959, 'duration': 5.281}, {'text': 'describing what your custom GPT does and', 'start': 619.44, 'duration': 4.72}, {'text': 'then giving some instruction And what', 'start': 622.24, 'duration': 4.48}, {'text': 'you can even do after is upload files to', 'start': 624.16, 'duration': 5.28}, {'text': 'augment the knowledge of this GPT Pay', 'start': 626.72, 'duration': 4.88}, {'text': 'attention that if you activate code', 'start': 629.44, 'duration': 4.72}, {'text': 'interpreter those files are downloadable', 'start': 631.6, 'duration': 5.44}, {'text': 'by the other users If your GPT is public', 'start': 634.16, 'duration': 4.72}, {'text': 'you can also keep your GPT for yourself', 'start': 637.04, 'duration': 3.76}, {'text': 'But just know that that this is', 'start': 638.88, 'duration': 4.079}, {'text': 'something to pay attention Also you have', 'start': 640.8, 'duration': 5.279}, {'text': 'the action function here is more if you', 'start': 642.959, 'duration': 4.961}, {'text': 'want to use code to get data from', 'start': 646.079, 'duration': 4.721}, {'text': 'outside through API and this is too', 'start': 647.92, 'duration': 4.88}, {'text': 'advanced to cover that in this course', 'start': 650.8, 'duration': 4.0}, {'text': \"But if you don't know how to start a fun\", 'start': 652.8, 'duration': 4.159}, {'text': 'way is to use the create view because', 'start': 654.8, 'duration': 4.64}, {'text': 'there you can just create your GPT', 'start': 656.959, 'duration': 4.161}, {'text': 'through a discussion So now let me show', 'start': 659.44, 'duration': 4.8}, {'text': 'you how to create a GPT So you will go', 'start': 661.12, 'duration': 5.6}, {'text': 'in the create view which is the easiest', 'start': 664.24, 'duration': 4.48}, {'text': \"one if you just started Let's imagine\", 'start': 666.72, 'duration': 4.88}, {'text': 'that we want a coach to be better at', 'start': 668.72, 'duration': 6.64}, {'text': 'FPNA I want to create a coach to get', 'start': 671.6, 'duration': 6.88}, {'text': 'better at FPNA I will just say that and', 'start': 675.36, 'duration': 5.68}, {'text': \"let's see the discussion Yeah I'm happy\", 'start': 678.48, 'duration': 5.68}, {'text': 'about this name So', 'start': 681.04, 'duration': 3.12}, {'text': 'yes and you can see that TGPT even', 'start': 687.16, 'duration': 6.76}, {'text': \"created a logo for our FPNA coach So I'm\", 'start': 690.56, 'duration': 6.64}, {'text': 'happy Yes I like', 'start': 693.92, 'duration': 3.28}, {'text': 'it So what should be emphasized so here', 'start': 697.959, 'duration': 6.521}, {'text': 'I want that this coach shows me the', 'start': 701.04, 'duration': 7.6}, {'text': 'technical aspect of being a FPNA pro and', 'start': 704.48, 'duration': 7.28}, {'text': 'shows me both the theory and practical', 'start': 708.64, 'duration': 5.16}, {'text': 'examples and step by', 'start': 711.76, 'duration': 5.36}, {'text': \"step So now that I've done that this is\", 'start': 713.8, 'duration': 6.039}, {'text': 'my instructions to improve this GPT and', 'start': 717.12, 'duration': 4.48}, {'text': \"like this I don't have to repeat after\", 'start': 719.839, 'duration': 5.041}, {'text': 'how this GPT will act So I want more', 'start': 721.6, 'duration': 6.16}, {'text': 'something direct and maybe I need also', 'start': 724.88, 'duration': 5.88}, {'text': 'motivational', 'start': 727.76, 'duration': 3.0}, {'text': 'So now the GPT is actually finished and', 'start': 732.079, 'duration': 5.2}, {'text': 'we can try it So for example I can try', 'start': 734.72, 'duration': 4.96}, {'text': 'how can I improve my budgeting process', 'start': 737.279, 'duration': 4.8}, {'text': 'So you see we have the answer and either', 'start': 739.68, 'duration': 4.24}, {'text': 'you are happy and you can create it or', 'start': 742.079, 'duration': 4.241}, {'text': 'you can even do something is you go to', 'start': 743.92, 'duration': 6.0}, {'text': 'configure you see how was this GPT', 'start': 746.32, 'duration': 5.44}, {'text': 'configurated and you can change', 'start': 749.92, 'duration': 4.88}, {'text': 'something for example you say use cables', 'start': 751.76, 'duration': 6.0}, {'text': 'to present information when possible and', 'start': 754.8, 'duration': 4.719}, {'text': 'you can even change your conversation', 'start': 757.76, 'duration': 5.44}, {'text': 'starters so now I will create the GPT', 'start': 759.519, 'duration': 6.481}, {'text': \"and that's where you can either have it\", 'start': 763.2, 'duration': 5.28}, {'text': 'on invite only or anyone in your', 'start': 766.0, 'duration': 5.279}, {'text': 'organization because I have a paid JGPT', 'start': 768.48, 'duration': 5.52}, {'text': 'version which is teams or for companies', 'start': 771.279, 'duration': 4.721}, {'text': 'they have enterprise So for me in my', 'start': 774.0, 'duration': 4.48}, {'text': 'organization I can decide who has access', 'start': 776.0, 'duration': 5.36}, {'text': 'or I can put it on the GPT store or if', 'start': 778.48, 'duration': 4.72}, {'text': 'you only want that specific people have', 'start': 781.36, 'duration': 3.68}, {'text': 'you can just share them the link like', 'start': 783.2, 'duration': 3.6}, {'text': \"this is not on the GPT store but it's\", 'start': 785.04, 'duration': 3.52}, {'text': 'still available for people with the link', 'start': 786.8, 'duration': 3.68}, {'text': 'So for now I just wanted to keep it for', 'start': 788.56, 'duration': 4.399}, {'text': 'myself So I will click update and then I', 'start': 790.48, 'duration': 5.039}, {'text': 'will view this GPT and this GPT will', 'start': 792.959, 'duration': 4.401}, {'text': 'also appear on the left with all of my', 'start': 795.519, 'duration': 4.56}, {'text': 'other GPT And now I can ask for example', 'start': 797.36, 'duration': 6.08}, {'text': 'how to do a cohort analysis I will enter', 'start': 800.079, 'duration': 5.521}, {'text': 'And this will help me as a coach And you', 'start': 803.44, 'duration': 4.079}, {'text': 'see that it follows the step by step', 'start': 805.6, 'duration': 4.239}, {'text': 'like I wanted Again you see how you save', 'start': 807.519, 'duration': 4.241}, {'text': \"time because I don't have to repeat that\", 'start': 809.839, 'duration': 4.961}, {'text': 'I want to have advices shown me in a', 'start': 811.76, 'duration': 4.879}, {'text': 'step by step It will always do it like', 'start': 814.8, 'duration': 4.64}, {'text': 'this And on top it combines theory with', 'start': 816.639, 'duration': 4.721}, {'text': 'practical advice and step-by-step', 'start': 819.44, 'duration': 4.92}, {'text': 'instructions', 'start': 821.36, 'duration': 3.0}, {'text': 'In this lesson we will explore Canvas a', 'start': 824.959, 'duration': 5.601}, {'text': 'powerful feature in TGPT that helps', 'start': 827.92, 'duration': 5.12}, {'text': 'finance professionals draft reports', 'start': 830.56, 'duration': 5.279}, {'text': 'commentaries and even modify code in', 'start': 833.04, 'duration': 5.52}, {'text': 'real time all in a seamless workspace', 'start': 835.839, 'duration': 4.8}, {'text': \"Let me show you how to use it Let's take\", 'start': 838.56, 'duration': 4.959}, {'text': 'the P&L of Nvidia as an example and', 'start': 840.639, 'duration': 5.601}, {'text': \"let's ask JGPT to draft commentaries on\", 'start': 843.519, 'duration': 5.801}, {'text': \"this P&L I'm going to take a\", 'start': 846.24, 'duration': 6.56}, {'text': \"screenshot I'll now back in GPT and I\", 'start': 849.32, 'duration': 5.959}, {'text': 'will paste the screenshot and I will say', 'start': 852.8, 'duration': 5.76}, {'text': 'draft the financial commentaries on the', 'start': 855.279, 'duration': 6.881}, {'text': 'P&L of this company And here I can', 'start': 858.56, 'duration': 5.92}, {'text': 'activate canvas Sometimes it activates', 'start': 862.16, 'duration': 4.239}, {'text': 'automatically but I really want to use', 'start': 864.48, 'duration': 3.44}, {'text': 'the canvas function and you will', 'start': 866.399, 'duration': 3.521}, {'text': \"understand why So let's launch it and\", 'start': 867.92, 'duration': 3.08}, {'text': 'see what is', 'start': 869.92, 'duration': 3.599}, {'text': 'happening Now we have a totally new view', 'start': 871.0, 'duration': 4.6}, {'text': 'that is coming because on the right side', 'start': 873.519, 'duration': 5.12}, {'text': 'my document is getting written by JGPT', 'start': 875.6, 'duration': 4.96}, {'text': 'and on the left side I have the', 'start': 878.639, 'duration': 3.681}, {'text': 'conversation that is still active where', 'start': 880.56, 'duration': 4.32}, {'text': 'you can see on the upper left the', 'start': 882.32, 'duration': 5.04}, {'text': 'screenshot and my prompt and here the', 'start': 884.88, 'duration': 5.68}, {'text': 'answer from JGPT So what can we do on', 'start': 887.36, 'duration': 6.4}, {'text': 'the right side with this document well', 'start': 890.56, 'duration': 4.719}, {'text': 'let me show you the different', 'start': 893.76, 'duration': 4.24}, {'text': 'functionalities on the bottom right If I', 'start': 895.279, 'duration': 5.601}, {'text': 'go over the pen I have different type of', 'start': 898.0, 'duration': 5.279}, {'text': \"edits I can add emojis which we don't\", 'start': 900.88, 'duration': 4.48}, {'text': 'really need for a factual commentary I', 'start': 903.279, 'duration': 4.24}, {'text': 'can polish it I can change the reading', 'start': 905.36, 'duration': 6.24}, {'text': \"level and the length Let's see here if I\", 'start': 907.519, 'duration': 6.641}, {'text': 'ask to polish it Push it means that it', 'start': 911.6, 'duration': 5.039}, {'text': 'will review the text and makes it even', 'start': 914.16, 'duration': 4.56}, {'text': \"better So you can see that it's changing\", 'start': 916.639, 'duration': 4.32}, {'text': 'slowly all of the text Now the position', 'start': 918.72, 'duration': 4.239}, {'text': \"is done Let's see what are the other\", 'start': 920.959, 'duration': 5.041}, {'text': \"function The reading level maybe let's\", 'start': 922.959, 'duration': 6.0}, {'text': 'make it more either easy or advanced', 'start': 926.0, 'duration': 5.04}, {'text': \"Let's put it graduate school to see if\", 'start': 928.959, 'duration': 4.161}, {'text': 'our wording is changing Okay so you can', 'start': 931.04, 'duration': 4.239}, {'text': 'see we are using different types of', 'start': 933.12, 'duration': 4.399}, {'text': 'wording and adjectives and you can', 'start': 935.279, 'duration': 4.641}, {'text': 'yourself see which one you like the most', 'start': 937.519, 'duration': 4.32}, {'text': \"for your style Okay now let's try\", 'start': 939.92, 'duration': 3.68}, {'text': \"something else Let's make it shorter\", 'start': 941.839, 'duration': 3.761}, {'text': \"because it's a bit too long So now I\", 'start': 943.6, 'duration': 5.32}, {'text': 'will really make it the shortest as', 'start': 945.6, 'duration': 5.84}, {'text': 'possible And you can see that now the', 'start': 948.92, 'duration': 5.64}, {'text': 'document is much shorter So this is', 'start': 951.44, 'duration': 5.519}, {'text': 'actually already good because we can', 'start': 954.56, 'duration': 4.079}, {'text': 'change really quickly But look what I', 'start': 956.959, 'duration': 3.761}, {'text': 'can also do here For example I can', 'start': 958.639, 'duration': 5.44}, {'text': 'select the text and I can ask CH GPT to', 'start': 960.72, 'duration': 7.039}, {'text': 'change it to add that the growth comes', 'start': 964.079, 'duration': 6.961}, {'text': \"from AI So let's add some flavor in the\", 'start': 967.759, 'duration': 6.121}, {'text': 'text from AI and', 'start': 971.04, 'duration': 5.599}, {'text': 'blockchain And we can see here that is', 'start': 973.88, 'duration': 5.16}, {'text': 'rewriting it adding that the growth is', 'start': 976.639, 'duration': 4.32}, {'text': 'currently driven by advancement in AI', 'start': 979.04, 'duration': 3.76}, {'text': 'and blockchain technology Of course this', 'start': 980.959, 'duration': 4.401}, {'text': 'part is you who is driving the seats You', 'start': 982.8, 'duration': 4.479}, {'text': \"can also say that oh I don't really like\", 'start': 985.36, 'duration': 5.12}, {'text': \"blockchain technology so let's just\", 'start': 987.279, 'duration': 5.841}, {'text': 'remove it and now I just delete it and I', 'start': 990.48, 'duration': 6.4}, {'text': 'can write especially in the last month', 'start': 993.12, 'duration': 5.6}, {'text': \"So that's what is good is now you are\", 'start': 996.88, 'duration': 3.92}, {'text': 'also in control and you can write over', 'start': 998.72, 'duration': 5.76}, {'text': 'it You can even highlight here in bold', 'start': 1000.8, 'duration': 7.279}, {'text': 'You can even here change the headings', 'start': 1004.48, 'duration': 6.96}, {'text': 'All of this you are in control The last', 'start': 1008.079, 'duration': 6.161}, {'text': 'important functionality of Canva is with', 'start': 1011.44, 'duration': 6.0}, {'text': 'coding So here I ask JGPT to create a', 'start': 1014.24, 'duration': 6.08}, {'text': 'code to illustrate a scenario analysis', 'start': 1017.44, 'duration': 5.839}, {'text': 'and you will see that if I activate', 'start': 1020.32, 'duration': 6.32}, {'text': 'again the Canva we will get the code', 'start': 1023.279, 'duration': 6.64}, {'text': 'written in Canva and I can go on and', 'start': 1026.64, 'duration': 6.08}, {'text': 'also modify this code But what is even', 'start': 1029.919, 'duration': 5.361}, {'text': 'better is there is the button run So I', 'start': 1032.72, 'duration': 6.719}, {'text': 'can run the code and it will for me test', 'start': 1035.28, 'duration': 7.519}, {'text': 'if the code is working or it will also', 'start': 1039.439, 'duration': 5.12}, {'text': 'tell me if there is any bug So for', 'start': 1042.799, 'duration': 3.76}, {'text': \"example here tells me that it doesn't\", 'start': 1044.559, 'duration': 4.721}, {'text': 'have this type of libraries So I can ask', 'start': 1046.559, 'duration': 5.921}, {'text': 'to fix the bugs and you can see that now', 'start': 1049.28, 'duration': 5.2}, {'text': 'that is fixing the bug is going through', 'start': 1052.48, 'duration': 4.079}, {'text': 'all of the code to improve the code to', 'start': 1054.48, 'duration': 5.199}, {'text': 'make it work So this is useful if you', 'start': 1056.559, 'duration': 5.761}, {'text': 'want either to change part of the code', 'start': 1059.679, 'duration': 4.481}, {'text': 'So because you can select a part and', 'start': 1062.32, 'duration': 4.0}, {'text': 'just debug the part that is not working', 'start': 1064.16, 'duration': 4.759}, {'text': 'or if you also want to test the code', 'start': 1066.32, 'duration': 5.68}, {'text': 'inside GPT directly So go and try it by', 'start': 1068.919, 'duration': 5.081}, {'text': 'yourself especially for the writing part', 'start': 1072.0, 'duration': 6.48}, {'text': \"and I'm sure you will love this new\", 'start': 1074.0, 'duration': 4.48}, {'text': 'functionality Let me show you the TGPT', 'start': 1079.08, 'duration': 5.4}, {'text': 'search functionality and how to use it', 'start': 1082.16, 'duration': 5.12}, {'text': \"in finance I'm back in TGPT and you can\", 'start': 1084.48, 'duration': 5.199}, {'text': 'see that here there is the button search', 'start': 1087.28, 'duration': 5.519}, {'text': 'So if you really want to get online and', 'start': 1089.679, 'duration': 6.161}, {'text': 'get sources then I will advise you to', 'start': 1092.799, 'duration': 5.841}, {'text': 'click to search the web But sometimes', 'start': 1095.84, 'duration': 6.48}, {'text': 'also without clicking it If you ask for', 'start': 1098.64, 'duration': 6.48}, {'text': 'example an information that is recent If', 'start': 1102.32, 'duration': 5.359}, {'text': 'I say what is the price of the Nvidia', 'start': 1105.12, 'duration': 4.72}, {'text': \"stock from last Friday then let's see\", 'start': 1107.679, 'duration': 3.841}, {'text': 'what is going to happen You can see that', 'start': 1109.84, 'duration': 5.44}, {'text': 'it automatically search the web from my', 'start': 1111.52, 'duration': 6.0}, {'text': \"question even if I didn't ask it And now\", 'start': 1115.28, 'duration': 5.759}, {'text': \"I have also February 14 2025 So that's\", 'start': 1117.52, 'duration': 5.68}, {'text': 'the date when I recorded this video And', 'start': 1121.039, 'duration': 4.721}, {'text': 'we can see here the price stock of', 'start': 1123.2, 'duration': 5.12}, {'text': 'Nvidia And what is also really good is', 'start': 1125.76, 'duration': 5.44}, {'text': 'you have the sources where you can find', 'start': 1128.32, 'duration': 5.28}, {'text': 'this information So like this you have', 'start': 1131.2, 'duration': 5.359}, {'text': 'one realtime information two you have', 'start': 1133.6, 'duration': 5.52}, {'text': \"sources Now let's look at a real use\", 'start': 1136.559, 'duration': 5.681}, {'text': 'case for us in finance When clicking', 'start': 1139.12, 'duration': 6.0}, {'text': \"search I'm going to ask specifically on\", 'start': 1142.24, 'duration': 5.04}, {'text': 'which website I want to get my', 'start': 1145.12, 'duration': 4.32}, {'text': 'information and what I want to get So if', 'start': 1147.28, 'duration': 3.6}, {'text': \"I click you'll see that it will\", 'start': 1149.44, 'duration': 3.84}, {'text': 'automatically search the web What is', 'start': 1150.88, 'duration': 4.88}, {'text': 'really good is that now I have the', 'start': 1153.28, 'duration': 4.399}, {'text': 'different types of publication on the', 'start': 1155.76, 'duration': 4.799}, {'text': 'IFS 15 I can try this one for example', 'start': 1157.679, 'duration': 5.36}, {'text': 'and now I am already inside the', 'start': 1160.559, 'duration': 4.761}, {'text': \"documentation So that's a great\", 'start': 1163.039, 'duration': 4.721}, {'text': 'timesaver and also what is really good', 'start': 1165.32, 'duration': 7.0}, {'text': 'is you have always the link of the', 'start': 1167.76, 'duration': 4.56}, {'text': 'publication Let me now present you the', 'start': 1172.919, 'duration': 6.041}, {'text': \"project functionalities from JPT I'm\", 'start': 1175.76, 'duration': 5.76}, {'text': 'back in Chip and here on the left side', 'start': 1178.96, 'duration': 5.2}, {'text': 'if you have a pro account or a teams or', 'start': 1181.52, 'duration': 4.159}, {'text': 'enterprise account you can see that', 'start': 1184.16, 'duration': 4.16}, {'text': 'there is the projects section here you', 'start': 1185.679, 'duration': 4.721}, {'text': 'will see the list of your projects and', 'start': 1188.32, 'duration': 4.08}, {'text': \"you can also create a new project that's\", 'start': 1190.4, 'duration': 4.32}, {'text': 'what we are going to do so I create a', 'start': 1192.4, 'duration': 4.24}, {'text': \"project let's imagine that I want to\", 'start': 1194.72, 'duration': 5.6}, {'text': \"work on the budget for 2025 let's create\", 'start': 1196.64, 'duration': 6.12}, {'text': 'the project so now I am inside my', 'start': 1200.32, 'duration': 5.599}, {'text': 'project I can close here the view and', 'start': 1202.76, 'duration': 6.12}, {'text': 'here for the project I can add file case', 'start': 1205.919, 'duration': 4.88}, {'text': 'So I have two files Let me show you I', 'start': 1208.88, 'duration': 4.96}, {'text': 'have here my P&L for 22 23 24 and the', 'start': 1210.799, 'duration': 5.681}, {'text': 'budget 24 And I have here also a file', 'start': 1213.84, 'duration': 4.719}, {'text': 'with my budget instructions and some', 'start': 1216.48, 'duration': 4.4}, {'text': \"deadlines for this year I've made a PDF\", 'start': 1218.559, 'duration': 4.561}, {'text': 'of this document And this PDF I will', 'start': 1220.88, 'duration': 4.799}, {'text': 'upload in my project To upload I just', 'start': 1223.12, 'duration': 5.36}, {'text': 'need to go and click add files and I', 'start': 1225.679, 'duration': 5.601}, {'text': 'will add my', 'start': 1228.48, 'duration': 2.8}, {'text': \"instructions Here I've added my budget\", 'start': 1232.12, 'duration': 5.4}, {'text': \"instructions and now I'm adding the P&L\", 'start': 1234.799, 'duration': 5.681}, {'text': 'Okay So we are set I can close and I can', 'start': 1237.52, 'duration': 5.6}, {'text': 'also add instructions here For example I', 'start': 1240.48, 'duration': 6.72}, {'text': 'can say this project is for the FPN team', 'start': 1243.12, 'duration': 6.88}, {'text': 'of a manufacturing company We are in the', 'start': 1247.2, 'duration': 6.24}, {'text': 'US and we work with Netswuite You can', 'start': 1250.0, 'duration': 5.84}, {'text': 'add any type of details You can also add', 'start': 1253.44, 'duration': 5.28}, {'text': 'the way you want JPT transfer Here I can', 'start': 1255.84, 'duration': 6.0}, {'text': 'say your responses should be direct and', 'start': 1258.72, 'duration': 6.88}, {'text': 'use bullet points and tables as much as', 'start': 1261.84, 'duration': 5.839}, {'text': \"possible So that's the advantage here is\", 'start': 1265.6, 'duration': 3.84}, {'text': 'you will see that those instructions', 'start': 1267.679, 'duration': 5.12}, {'text': 'they will be for any chat I start within', 'start': 1269.44, 'duration': 5.84}, {'text': \"this project So now let's imagine that I\", 'start': 1272.799, 'duration': 5.041}, {'text': 'want to draft an email to the sales team', 'start': 1275.28, 'duration': 6.48}, {'text': 'Draft an email to the sales team to ask', 'start': 1277.84, 'duration': 6.8}, {'text': 'their input for the budget and I can say', 'start': 1281.76, 'duration': 4.36}, {'text': 'mention the', 'start': 1284.64, 'duration': 4.159}, {'text': 'deadlines So here you can see that the', 'start': 1286.12, 'duration': 5.799}, {'text': \"deadline October 2025 I didn't have to\", 'start': 1288.799, 'duration': 5.12}, {'text': 'say which deadline it was because', 'start': 1291.919, 'duration': 4.481}, {'text': 'actually here in the instructions I', 'start': 1293.919, 'duration': 5.12}, {'text': 'already had my details and you can see', 'start': 1296.4, 'duration': 5.279}, {'text': 'here what we needed to have I can find', 'start': 1299.039, 'duration': 5.361}, {'text': 'it here back in the key areas of input', 'start': 1301.679, 'duration': 4.961}, {'text': \"So that's really practical because here\", 'start': 1304.4, 'duration': 4.56}, {'text': 'in this project there is a knowledge', 'start': 1306.64, 'duration': 4.32}, {'text': 'that you can use based on the document', 'start': 1308.96, 'duration': 3.92}, {'text': 'you uploaded but also based on the', 'start': 1310.96, 'duration': 4.4}, {'text': 'instructions Now I can go back to my', 'start': 1312.88, 'duration': 8.08}, {'text': 'budget 2025 and I can say draft PNL 2025', 'start': 1315.36, 'duration': 9.319}, {'text': 'format with the comparison with', 'start': 1320.96, 'duration': 7.04}, {'text': '2024 and again here it has the knowledge', 'start': 1324.679, 'duration': 6.12}, {'text': 'of the P&L of last year like this I can', 'start': 1328.0, 'duration': 4.88}, {'text': 'always reuse the information I opened it', 'start': 1330.799, 'duration': 4.321}, {'text': 'before So you can see here the answer', 'start': 1332.88, 'duration': 5.2}, {'text': 'where it assumed an increase of 10% We', 'start': 1335.12, 'duration': 4.88}, {'text': \"didn't ask that But what is interesting\", 'start': 1338.08, 'duration': 4.079}, {'text': 'is that we have already the table for', 'start': 1340.0, 'duration': 5.52}, {'text': '2024 and budget 2024 without having to', 'start': 1342.159, 'duration': 5.681}, {'text': 'give the information because it was in', 'start': 1345.52, 'duration': 5.039}, {'text': 'the Excel I provided before And now you', 'start': 1347.84, 'duration': 4.88}, {'text': 'can always go back when you are in the', 'start': 1350.559, 'duration': 5.521}, {'text': 'budget 2025 project You can go back in', 'start': 1352.72, 'duration': 5.92}, {'text': 'all of the charts that are made under', 'start': 1356.08, 'duration': 5.44}, {'text': \"this project And also let's imagine that\", 'start': 1358.64, 'duration': 5.36}, {'text': 'you had here another chart you can', 'start': 1361.52, 'duration': 5.519}, {'text': \"always add to the project 2025 So that's\", 'start': 1364.0, 'duration': 5.44}, {'text': 'also really advantageous because you can', 'start': 1367.039, 'duration': 5.281}, {'text': 'organize your chat based on some topics', 'start': 1369.44, 'duration': 5.119}, {'text': \"even if you don't add any files or even\", 'start': 1372.32, 'duration': 4.479}, {'text': \"if you don't add any instructions I have\", 'start': 1374.559, 'duration': 4.881}, {'text': 'some friends who are using it like this', 'start': 1376.799, 'duration': 4.721}, {'text': 'to basically organize their chat but', 'start': 1379.44, 'duration': 4.0}, {'text': \"they don't give any instructions or no\", 'start': 1381.52, 'duration': 3.92}, {'text': 'projects files inside And you might', 'start': 1383.44, 'duration': 3.76}, {'text': 'wonder what is the difference with', 'start': 1385.44, 'duration': 5.28}, {'text': 'custom GPT First here you can choose the', 'start': 1387.2, 'duration': 7.28}, {'text': 'model Here you can choose GPT4 or O1 or', 'start': 1390.72, 'duration': 6.88}, {'text': 'this for O mini or GPT for AC model and', 'start': 1394.48, 'duration': 5.52}, {'text': 'in GPS you cannot choose the model', 'start': 1397.6, 'duration': 4.64}, {'text': \"That's number one Number two here you\", 'start': 1400.0, 'duration': 4.64}, {'text': 'cannot really share this chat with', 'start': 1402.24, 'duration': 5.2}, {'text': 'somebody else but in GPS you can', 'start': 1404.64, 'duration': 4.96}, {'text': 'actually share the chat with your team', 'start': 1407.44, 'duration': 4.4}, {'text': 'and you can even also share your GPS', 'start': 1409.6, 'duration': 4.72}, {'text': 'with your team or with people outside of', 'start': 1411.84, 'duration': 4.48}, {'text': 'your organization if you decide to make', 'start': 1414.32, 'duration': 4.0}, {'text': 'it public to make project no project is', 'start': 1416.32, 'duration': 4.08}, {'text': 'only personal and here you see as well', 'start': 1418.32, 'duration': 4.4}, {'text': 'how easy it is to give your instructions', 'start': 1420.4, 'duration': 5.519}, {'text': 'and add the document for custom GPT you', 'start': 1422.72, 'duration': 6.0}, {'text': 'need to take a bit more time to', 'start': 1425.919, 'duration': 5.361}, {'text': 'customize your GPT So I will say really', 'start': 1428.72, 'duration': 6.16}, {'text': 'that project are different than GPS GPS', 'start': 1431.28, 'duration': 5.84}, {'text': 'is more like an agent They will have a', 'start': 1434.88, 'duration': 4.88}, {'text': 'specific behavior and project is more to', 'start': 1437.12, 'duration': 4.799}, {'text': 'chat in a closed environment where you', 'start': 1439.76, 'duration': 4.399}, {'text': 'have some information that are specific', 'start': 1441.919, 'duration': 4.801}, {'text': 'to the chat and you are going to create', 'start': 1444.159, 'duration': 4.88}, {'text': 'several chats about those information', 'start': 1446.72, 'duration': 4.319}, {'text': 'and those custom instructions One cool', 'start': 1449.039, 'duration': 4.561}, {'text': 'way also to use projects is to upload a', 'start': 1451.039, 'duration': 4.88}, {'text': 'lot of information about you or about a', 'start': 1453.6, 'duration': 5.28}, {'text': 'book or about a topic and then chat with', 'start': 1455.919, 'duration': 5.201}, {'text': 'this information Trade by yourself if', 'start': 1458.88, 'duration': 4.48}, {'text': 'you have a pro a team or enterprise', 'start': 1461.12, 'duration': 4.48}, {'text': 'version See how you can use it the best', 'start': 1463.36, 'duration': 5.52}, {'text': 'way for your own use', 'start': 1465.6, 'duration': 3.28}, {'text': \"cases Let's look now at the new\", 'start': 1469.32, 'duration': 5.4}, {'text': 'functionality from Chpt which is the', 'start': 1472.32, 'duration': 5.28}, {'text': 'operator Operator is like an agent that', 'start': 1474.72, 'duration': 6.4}, {'text': 'has access to your computer and can act', 'start': 1477.6, 'duration': 5.84}, {'text': 'on your behalf if you prompt it Operator', 'start': 1481.12, 'duration': 3.919}, {'text': 'is not yet available in all of the', 'start': 1483.44, 'duration': 3.44}, {'text': 'countries So let me show you a', 'start': 1485.039, 'duration': 4.561}, {'text': 'presentation of OpenAI showing how', 'start': 1486.88, 'duration': 4.72}, {'text': 'operator is working Here is the', 'start': 1489.6, 'duration': 4.24}, {'text': 'presentation from OpenAI showing the', 'start': 1491.6, 'duration': 5.84}, {'text': 'operator So now they show it mostly for', 'start': 1493.84, 'duration': 7.04}, {'text': 'individuals like shopping or doing a', 'start': 1497.44, 'duration': 6.08}, {'text': 'reservation for a table Here you can see', 'start': 1500.88, 'duration': 5.6}, {'text': 'that OpenAI is showing an example of how', 'start': 1503.52, 'duration': 6.48}, {'text': 'to use operator to book for you a table', 'start': 1506.48, 'duration': 6.079}, {'text': 'So here you see that once we prompted to', 'start': 1510.0, 'duration': 4.2}, {'text': 'book a dinner', 'start': 1512.559, 'duration': 5.36}, {'text': 'reservation operator is starting to act', 'start': 1514.2, 'duration': 6.2}, {'text': 'inside your browser On the left side you', 'start': 1517.919, 'duration': 4.801}, {'text': 'have the discussion you have with your', 'start': 1520.4, 'duration': 4.56}, {'text': \"operator and on the right side it's the\", 'start': 1522.72, 'duration': 4.72}, {'text': 'operator doing it everything by its own', 'start': 1524.96, 'duration': 4.0}, {'text': \"You can see that now it's selecting the\", 'start': 1527.44, 'duration': 4.08}, {'text': \"date It's selecting what type of food\", 'start': 1528.96, 'duration': 5.199}, {'text': \"and it's doing all of this on its own So\", 'start': 1531.52, 'duration': 4.88}, {'text': \"basically you don't control it You just\", 'start': 1534.159, 'duration': 4.961}, {'text': 'let it do on the side while looking at', 'start': 1536.4, 'duration': 4.8}, {'text': \"it to verify that it's doing what you\", 'start': 1539.12, 'duration': 3.679}, {'text': 'wanted And on the left side you can see', 'start': 1541.2, 'duration': 3.12}, {'text': 'all of the actions that were done So', 'start': 1542.799, 'duration': 3.281}, {'text': 'here you can see that the operator is', 'start': 1544.32, 'duration': 3.68}, {'text': 'telling you that the closest reservation', 'start': 1546.08, 'duration': 5.36}, {'text': 'is as from 7:45 You say \"Okay let\\'s go', 'start': 1548.0, 'duration': 5.6}, {'text': 'Let\\'s do it.\" And then the operator is', 'start': 1551.44, 'duration': 4.239}, {'text': 'going to book it for you You can see now', 'start': 1553.6, 'duration': 5.28}, {'text': 'you just need to give your details and', 'start': 1555.679, 'duration': 5.36}, {'text': 'then if you say yes the operator is', 'start': 1558.88, 'duration': 3.52}, {'text': 'going to click and then your reservation', 'start': 1561.039, 'duration': 3.281}, {'text': 'is done What does it mean for us well', 'start': 1562.4, 'duration': 3.84}, {'text': 'think about this report where you have', 'start': 1564.32, 'duration': 4.239}, {'text': 'to upload data manually You know like', 'start': 1566.24, 'duration': 6.08}, {'text': 'key in the URL then doing a copy and', 'start': 1568.559, 'duration': 6.081}, {'text': \"paste in some of the fields That's a\", 'start': 1572.32, 'duration': 4.479}, {'text': 'good use case for the operator Think', 'start': 1574.64, 'duration': 5.12}, {'text': 'also of this website where you have to', 'start': 1576.799, 'duration': 5.12}, {'text': 'go and have some steps in those websites', 'start': 1579.76, 'duration': 4.88}, {'text': \"to download invoices That's also a good\", 'start': 1581.919, 'duration': 5.601}, {'text': 'task for the operator Why because once', 'start': 1584.64, 'duration': 5.6}, {'text': 'you do the task with the operator it can', 'start': 1587.52, 'duration': 4.96}, {'text': 'memorize it and next time you just have', 'start': 1590.24, 'duration': 4.319}, {'text': 'to ask repeat this task and then the', 'start': 1592.48, 'duration': 4.48}, {'text': 'task will be done by itself So of course', 'start': 1594.559, 'duration': 4.641}, {'text': 'there are some confidentiality data to', 'start': 1596.96, 'duration': 4.4}, {'text': 'think about which access do you want to', 'start': 1599.2, 'duration': 4.16}, {'text': 'give to this operator also do you want', 'start': 1601.36, 'duration': 4.16}, {'text': 'to give your credit card details So', 'start': 1603.36, 'duration': 4.08}, {'text': \"that's something to think about But this\", 'start': 1605.52, 'duration': 4.56}, {'text': 'is really the start where your AI tool', 'start': 1607.44, 'duration': 4.64}, {'text': 'is going to be an agent for you and is', 'start': 1610.08, 'duration': 4.56}, {'text': 'going to be triggered by you or by', 'start': 1612.08, 'duration': 4.8}, {'text': 'scheduled task You could for example to', 'start': 1614.64, 'duration': 4.72}, {'text': 'ask every week to pull some data or', 'start': 1616.88, 'duration': 4.96}, {'text': 'every month to get some invoices or fill', 'start': 1619.36, 'duration': 4.319}, {'text': 'up some vendor forms where you have to', 'start': 1621.84, 'duration': 3.52}, {'text': 'give all of your details like banking', 'start': 1623.679, 'duration': 5.521}, {'text': 'details name tax ID etc etc And if you', 'start': 1625.36, 'duration': 5.679}, {'text': 'want to learn more I recommend this', 'start': 1629.2, 'duration': 4.32}, {'text': \"video from Ector Garcia where he's\", 'start': 1631.039, 'duration': 5.76}, {'text': 'showing us how operator can help on some', 'start': 1633.52, 'duration': 5.84}, {'text': 'QuickBooks task that you might do now', 'start': 1636.799, 'duration': 5.12}, {'text': 'manually But now with operator you can', 'start': 1639.36, 'duration': 5.04}, {'text': 'delegate this task to this agent Go and', 'start': 1641.919, 'duration': 4.64}, {'text': 'check', 'start': 1644.4, 'duration': 2.159}, {'text': 'it Let me now show you how you can', 'start': 1646.84, 'duration': 5.319}, {'text': 'customize Chip So on the right side you', 'start': 1649.12, 'duration': 5.12}, {'text': 'have this diamond icon You will click on', 'start': 1652.159, 'duration': 4.561}, {'text': 'it and you just have to click customize', 'start': 1654.24, 'duration': 5.2}, {'text': 'GPT And this is here where you can put', 'start': 1656.72, 'duration': 4.72}, {'text': 'your name what do you do what traits', 'start': 1659.44, 'duration': 5.359}, {'text': 'should chip have So if you like a formal', 'start': 1661.44, 'duration': 5.68}, {'text': 'tone or something a bit more casual and', 'start': 1664.799, 'duration': 4.641}, {'text': \"that's where here you can tell chip to\", 'start': 1667.12, 'duration': 4.159}, {'text': 'write like you if you have a specific', 'start': 1669.44, 'duration': 3.839}, {'text': 'tone And you see here you have already', 'start': 1671.279, 'duration': 3.921}, {'text': 'some suggestion Here is where you can', 'start': 1673.279, 'duration': 4.961}, {'text': 'explain chip more about you like where', 'start': 1675.2, 'duration': 6.32}, {'text': 'you work maybe your industry maybe your', 'start': 1678.24, 'duration': 5.12}, {'text': 'environment So if you are in the US or', 'start': 1681.52, 'duration': 3.519}, {'text': \"Europe that's important to explain And\", 'start': 1683.36, 'duration': 4.08}, {'text': 'if I click to advanced then I can also', 'start': 1685.039, 'duration': 5.201}, {'text': 'select the JGPT capabilities For example', 'start': 1687.44, 'duration': 4.64}, {'text': \"if you don't really like that it always\", 'start': 1690.24, 'duration': 3.52}, {'text': 'goes to the web search you can just', 'start': 1692.08, 'duration': 4.479}, {'text': 'deactivate it that now another way to', 'start': 1693.76, 'duration': 5.44}, {'text': 'get TGPT to get customized is to', 'start': 1696.559, 'duration': 5.36}, {'text': 'activate the memory So when I click to', 'start': 1699.2, 'duration': 5.52}, {'text': 'settings and I go to personalization I', 'start': 1701.919, 'duration': 5.281}, {'text': 'can ask to activate the memory And you', 'start': 1704.72, 'duration': 4.559}, {'text': 'can see what the memory can do is', 'start': 1707.2, 'duration': 4.8}, {'text': 'basically along your messages If you', 'start': 1709.279, 'duration': 5.841}, {'text': 'explain where you work if you explain a', 'start': 1712.0, 'duration': 5.2}, {'text': 'bit about you then it will start', 'start': 1715.12, 'duration': 4.559}, {'text': 'populate here the memories Here is a', 'start': 1717.2, 'duration': 4.64}, {'text': 'great way to let the AI personalize the', 'start': 1719.679, 'duration': 3.921}, {'text': 'experience for you And what is good is', 'start': 1721.84, 'duration': 3.839}, {'text': 'you can manage it yourself if you want', 'start': 1723.6, 'duration': 4.079}, {'text': 'or you can deactivate it if you prefer', 'start': 1725.679, 'duration': 4.48}, {'text': \"to have a neutral AI that doesn't know\", 'start': 1727.679, 'duration': 5.441}, {'text': 'about you or maybe just you decide when', 'start': 1730.159, 'duration': 4.721}, {'text': 'you want that it knows about you So you', 'start': 1733.12, 'duration': 3.76}, {'text': 'can do that in a custom GPT or in a', 'start': 1734.88, 'duration': 4.32}, {'text': 'project and you can decide in which chat', 'start': 1736.88, 'duration': 4.56}, {'text': 'where you want no information about you', 'start': 1739.2, 'duration': 3.68}, {'text': \"because you don't want to influence the\", 'start': 1741.44, 'duration': 5.359}, {'text': 'output So go ahead and check that for', 'start': 1742.88, 'duration': 3.919}, {'text': \"yourself Let's start in the meat of this\", 'start': 1747.32, 'duration': 8.12}, {'text': \"course how to prompt with CHP I'm going\", 'start': 1751.039, 'duration': 8.201}, {'text': 'to teach you my framework that works', 'start': 1755.44, 'duration': 6.4}, {'text': \"everywhere But first let's understand\", 'start': 1759.24, 'duration': 4.919}, {'text': 'why do we need a framework The first', 'start': 1761.84, 'duration': 5.199}, {'text': 'mistake we all do when we start using', 'start': 1764.159, 'duration': 6.081}, {'text': 'tool like JGPT is to use the Google', 'start': 1767.039, 'duration': 6.081}, {'text': 'method The Google method is simply what', 'start': 1770.24, 'duration': 5.76}, {'text': 'we have been doing the last 15 years is', 'start': 1773.12, 'duration': 5.919}, {'text': 'to research something based on keywords', 'start': 1776.0, 'duration': 6.48}, {'text': \"As we are humans we are lazy It's often\", 'start': 1779.039, 'duration': 5.601}, {'text': 'our favorite method because we have to', 'start': 1782.48, 'duration': 4.88}, {'text': 'write the list word to get the maximum', 'start': 1784.64, 'duration': 5.279}, {'text': \"output Let's see together what happens\", 'start': 1787.36, 'duration': 6.16}, {'text': 'if we are using JG GPT like Google', 'start': 1789.919, 'duration': 7.841}, {'text': 'Imagine that I want to issue reminder to', 'start': 1793.52, 'duration': 7.2}, {'text': \"one of my client who didn't pick I will\", 'start': 1797.76, 'duration': 5.32}, {'text': 'write inside', 'start': 1800.72, 'duration': 5.319}, {'text': \"JPT the name letter Let's see what\", 'start': 1803.08, 'duration': 7.479}, {'text': 'happens So we get something quite fast', 'start': 1806.039, 'duration': 7.0}, {'text': 'but we only get a description of what is', 'start': 1810.559, 'duration': 4.48}, {'text': 'the name letter', 'start': 1813.039, 'duration': 4.161}, {'text': \"The problem is that it doesn't really\", 'start': 1815.039, 'duration': 4.88}, {'text': 'help us because what we wanted is to', 'start': 1817.2, 'duration': 4.8}, {'text': 'send the letter We wanted to have the', 'start': 1819.919, 'duration': 4.721}, {'text': 'letter drafted So this is not working', 'start': 1822.0, 'duration': 5.36}, {'text': \"well because it doesn't really help us\", 'start': 1824.64, 'duration': 6.96}, {'text': \"with our work And that's where 90% of\", 'start': 1827.36, 'duration': 7.6}, {'text': 'the people will usually stop because', 'start': 1831.6, 'duration': 6.16}, {'text': 'they think tragic is smart but not', 'start': 1834.96, 'duration': 5.599}, {'text': 'helpful And imagine if you are working', 'start': 1837.76, 'duration': 6.0}, {'text': 'with somebody imagine that you want them', 'start': 1840.559, 'duration': 5.6}, {'text': 'to write the dunning letter for you Are', 'start': 1843.76, 'duration': 5.12}, {'text': 'you really going just to send them in', 'start': 1846.159, 'duration': 7.12}, {'text': 'teams in a chat dunning letter no you', 'start': 1848.88, 'duration': 6.0}, {'text': 'are going to explain what is the problem', 'start': 1853.279, 'duration': 4.241}, {'text': 'and what you want from them Well chat', 'start': 1854.88, 'duration': 5.48}, {'text': 'GPT is exactly the same way is like an', 'start': 1857.52, 'duration': 6.0}, {'text': 'assistant to whom you need to explain', 'start': 1860.36, 'duration': 5.64}, {'text': 'the context and be specific about what', 'start': 1863.52, 'duration': 5.72}, {'text': 'you want and also give a clear', 'start': 1866.0, 'duration': 5.76}, {'text': 'instruction to have clear expectation', 'start': 1869.24, 'duration': 6.28}, {'text': 'and have a clear output from chip This', 'start': 1871.76, 'duration': 7.76}, {'text': \"is why I'm bringing a team to solve a\", 'start': 1875.52, 'duration': 7.6}, {'text': 'crime which is called CSI With my', 'start': 1879.52, 'duration': 6.399}, {'text': 'framework called the CSI framework you', 'start': 1883.12, 'duration': 6.48}, {'text': 'are going to prompt using the CSI', 'start': 1885.919, 'duration': 7.201}, {'text': 'letters meaning C for context The', 'start': 1889.6, 'duration': 5.92}, {'text': 'context is basically explaining who you', 'start': 1893.12, 'duration': 4.799}, {'text': 'are and the context of your company E is', 'start': 1895.52, 'duration': 5.6}, {'text': 'being specific What is your problem i is', 'start': 1897.919, 'duration': 5.441}, {'text': 'to give a clear instruction because if', 'start': 1901.12, 'duration': 5.679}, {'text': 'you just stop at context and specific', 'start': 1903.36, 'duration': 5.84}, {'text': 'then you still let too much freedom to', 'start': 1906.799, 'duration': 4.801}, {'text': 'charge GPT to generate the output for', 'start': 1909.2, 'duration': 5.88}, {'text': \"you Now let's use our example of dueling\", 'start': 1911.6, 'duration': 8.799}, {'text': \"letter and let's see how we can use CSI\", 'start': 1915.08, 'duration': 8.8}, {'text': 'with I am an accountant being the', 'start': 1920.399, 'duration': 6.561}, {'text': 'context my client as an overview being', 'start': 1923.88, 'duration': 6.2}, {'text': 'the specificity and the problem and', 'start': 1926.96, 'duration': 7.12}, {'text': 'being specific about the instruction Now', 'start': 1930.08, 'duration': 7.439}, {'text': \"let's see if we use the CSI framework if\", 'start': 1934.08, 'duration': 6.319}, {'text': 'we have a better output from chip', 'start': 1937.519, 'duration': 6.081}, {'text': 'Already it starts to write the letter', 'start': 1940.399, 'duration': 8.481}, {'text': 'and we can see that we have exactly a', 'start': 1943.6, 'duration': 8.959}, {'text': 'letter on overdue on the outing balance', 'start': 1948.88, 'duration': 8.639}, {'text': 'and we ask nicely about the reminder and', 'start': 1952.559, 'duration': 8.161}, {'text': 'we also ask what is the status So we are', 'start': 1957.519, 'duration': 5.921}, {'text': 'quite gentle Now the problem is that', 'start': 1960.72, 'duration': 6.079}, {'text': \"this client didn't pay since a long time\", 'start': 1963.44, 'duration': 4.8}, {'text': \"and it's not the first time that we have\", 'start': 1966.799, 'duration': 4.321}, {'text': 'this problem So do you think this letter', 'start': 1968.24, 'duration': 6.08}, {'text': 'will be enough to get our money back no', 'start': 1971.12, 'duration': 4.799}, {'text': \"it's not enough because we are too\", 'start': 1974.32, 'duration': 4.64}, {'text': 'gentle So how are we going to solve this', 'start': 1975.919, 'duration': 5.681}, {'text': 'if we want to solve this problem we are', 'start': 1978.96, 'duration': 5.12}, {'text': 'going to bring a second team to make', 'start': 1981.6, 'duration': 5.12}, {'text': 'sure we get our money back and we are', 'start': 1984.08, 'duration': 7.36}, {'text': 'going to bring the FBI FBI stands for F', 'start': 1986.72, 'duration': 9.28}, {'text': 'the format B the blueprint and I', 'start': 1991.44, 'duration': 7.599}, {'text': 'identity If we get back our example', 'start': 1996.0, 'duration': 5.6}, {'text': 'where at the beginning we had this', 'start': 1999.039, 'duration': 4.801}, {'text': 'prompt I am an accountant My client has', 'start': 2001.6, 'duration': 4.24}, {'text': 'never due for two months Can you draft a', 'start': 2003.84, 'duration': 4.8}, {'text': 'communication to them we are going to', 'start': 2005.84, 'duration': 5.839}, {'text': 'add F the format So we are going to have', 'start': 2008.64, 'duration': 6.8}, {'text': 'a formal letter B my favorite part the', 'start': 2011.679, 'duration': 6.161}, {'text': 'blueprint We have something in mind but', 'start': 2015.44, 'duration': 4.8}, {'text': \"we don't really know how to say that So\", 'start': 2017.84, 'duration': 6.16}, {'text': 'we will let CHGPT formulate for us and', 'start': 2020.24, 'duration': 8.4}, {'text': 'just using words like tone to be hard or', 'start': 2024.0, 'duration': 8.159}, {'text': \"legal actions we know it's going to have\", 'start': 2028.64, 'duration': 5.84}, {'text': 'much more impact than our previous', 'start': 2032.159, 'duration': 6.721}, {'text': 'letter and finally I for identity where', 'start': 2034.48, 'duration': 8.72}, {'text': 'we will ask to act as the best lawyer', 'start': 2038.88, 'duration': 6.96}, {'text': \"and here identity you don't always need\", 'start': 2043.2, 'duration': 5.919}, {'text': 'it but it could be really useful in this', 'start': 2045.84, 'duration': 6.319}, {'text': \"type of situation Let's go into ch to\", 'start': 2049.119, 'duration': 4.321}, {'text': 'see how we are going to use this', 'start': 2052.159, 'duration': 4.24}, {'text': \"framework Let's see now the output once\", 'start': 2053.44, 'duration': 7.439}, {'text': 'we use the FBI formula on top of CSI You', 'start': 2056.399, 'duration': 6.801}, {'text': 'can see it started with a letter head', 'start': 2060.879, 'duration': 6.121}, {'text': 'and we have already in our title legal', 'start': 2063.2, 'duration': 6.56}, {'text': 'actions You will also notice that our', 'start': 2067.0, 'duration': 6.76}, {'text': 'tone has changed and we are using words', 'start': 2069.76, 'duration': 6.56}, {'text': 'to escalate the situation as we wanted', 'start': 2073.76, 'duration': 5.839}, {'text': 'We have words like lawsuit seeking a', 'start': 2076.32, 'duration': 6.24}, {'text': 'court judgment or engaging debt', 'start': 2079.599, 'duration': 3.881}, {'text': 'collection', 'start': 2082.56, 'duration': 4.72}, {'text': 'agencies So what do you think of this', 'start': 2083.48, 'duration': 6.119}, {'text': 'letter do you think this will have more', 'start': 2087.28, 'duration': 4.799}, {'text': 'impact than the previous one and here is', 'start': 2089.599, 'duration': 6.32}, {'text': 'an additional tip You can also ask can', 'start': 2092.079, 'duration': 7.241}, {'text': 'you list the legal', 'start': 2095.919, 'duration': 8.481}, {'text': 'action in a bullet point list and', 'start': 2099.32, 'duration': 8.6}, {'text': 'reddraft this letter i like to do it', 'start': 2104.4, 'duration': 6.88}, {'text': 'that way because you see here for a', 'start': 2107.92, 'duration': 6.4}, {'text': 'human is actually much easier to read in', 'start': 2111.28, 'duration': 5.92}, {'text': 'bullet points and we know that if', 'start': 2114.32, 'duration': 5.68}, {'text': 'somebody came across this letter they', 'start': 2117.2, 'duration': 6.08}, {'text': 'will first read that And I assure you if', 'start': 2120.0, 'duration': 6.4}, {'text': \"you read that I'm sure your client or\", 'start': 2123.28, 'duration': 6.0}, {'text': 'yourself you are going to react on this', 'start': 2126.4, 'duration': 5.04}, {'text': 'This is how you can use the framework', 'start': 2129.28, 'duration': 6.72}, {'text': 'CSI plus FBI And I recommend you to', 'start': 2131.44, 'duration': 8.44}, {'text': 'first try all of your prompts in CHP', 'start': 2136.0, 'duration': 7.2}, {'text': 'with at least CSI and when you want', 'start': 2139.88, 'duration': 6.92}, {'text': 'something more complex use FBI And the', 'start': 2143.2, 'duration': 6.24}, {'text': 'more you will get used to this framework', 'start': 2146.8, 'duration': 4.96}, {'text': 'the more you will see what works and the', 'start': 2149.44, 'duration': 4.639}, {'text': 'more you can take freedom from this', 'start': 2151.76, 'duration': 4.72}, {'text': 'framework and build your own frameworks', 'start': 2154.079, 'duration': 4.961}, {'text': 'based on your', 'start': 2156.48, 'duration': 2.56}, {'text': \"experience Let's start with an exercise\", 'start': 2159.72, 'duration': 6.28}, {'text': 'to practice the CSI and FBI formula', 'start': 2162.56, 'duration': 6.72}, {'text': 'Imagine that you are the FPNA analyst of', 'start': 2166.0, 'duration': 6.72}, {'text': 'a brewery and you have to create a', 'start': 2169.28, 'duration': 5.92}, {'text': 'weekly financial report You are going to', 'start': 2172.72, 'duration': 5.84}, {'text': 'use JG GPT using the CSI formula to', 'start': 2175.2, 'duration': 5.2}, {'text': \"generate this report That's the first\", 'start': 2178.56, 'duration': 4.0}, {'text': 'part When you are done you are going to', 'start': 2180.4, 'duration': 7.84}, {'text': 'use CSI and FBI to improve your prompt', 'start': 2182.56, 'duration': 7.76}, {'text': 'and also with that to improve your', 'start': 2188.24, 'duration': 4.32}, {'text': 'output Take five minutes to do this', 'start': 2190.32, 'duration': 5.16}, {'text': 'exercise and I see you in the', 'start': 2192.56, 'duration': 4.92}, {'text': 'solution', 'start': 2195.48, 'duration': 5.879}, {'text': 'Remember the exercise was to create a', 'start': 2197.48, 'duration': 6.599}, {'text': 'weekly financial report for your FPNA', 'start': 2201.359, 'duration': 5.041}, {'text': \"manager Let's see first what happens\", 'start': 2204.079, 'duration': 5.601}, {'text': 'when you use the CSI formula So here I', 'start': 2206.4, 'duration': 7.439}, {'text': 'wrote the context I am FPN analyst for a', 'start': 2209.68, 'duration': 7.08}, {'text': 'brewery Then I', 'start': 2213.839, 'duration': 6.0}, {'text': 'explain what is my specific problem I', 'start': 2216.76, 'duration': 4.599}, {'text': 'need to deliver the weekly financial', 'start': 2219.839, 'duration': 3.601}, {'text': 'report to my manager and I ask', 'start': 2221.359, 'duration': 5.041}, {'text': 'specifically to draft the report outline', 'start': 2223.44, 'duration': 4.2}, {'text': \"Let's see what\", 'start': 2226.4, 'duration': 5.36}, {'text': 'happens So as you can see we have the', 'start': 2227.64, 'duration': 7.16}, {'text': 'weekly financial report with adapted', 'start': 2231.76, 'duration': 6.16}, {'text': 'with the beer types Also we have', 'start': 2234.8, 'duration': 5.2}, {'text': \"production cost which you don't have for\", 'start': 2237.92, 'duration': 5.36}, {'text': 'a SAS and also inventory topics So it', 'start': 2240.0, 'duration': 6.68}, {'text': 'understood that we are a beer or brewery', 'start': 2243.28, 'duration': 6.079}, {'text': \"company with production cost So that's\", 'start': 2246.68, 'duration': 4.84}, {'text': 'really good because if you now say that', 'start': 2249.359, 'duration': 4.401}, {'text': \"you are a SAS company you'll have a\", 'start': 2251.52, 'duration': 4.72}, {'text': 'different type of report Now imagine', 'start': 2253.76, 'duration': 5.12}, {'text': 'that what you wanted is actually slides', 'start': 2256.24, 'duration': 5.119}, {'text': 'and this is not enough for slides', 'start': 2258.88, 'duration': 4.32}, {'text': \"because you'll have to readapt it So\", 'start': 2261.359, 'duration': 4.48}, {'text': 'either you can continue the discussion', 'start': 2263.2, 'duration': 7.8}, {'text': 'and say adapt it to', 'start': 2265.839, 'duration': 7.481}, {'text': 'create the', 'start': 2271.0, 'duration': 6.52}, {'text': 'slide structure from this', 'start': 2273.32, 'duration': 7.48}, {'text': 'And now we have the slide name for each', 'start': 2277.52, 'duration': 6.319}, {'text': \"of them Or I'll be more specific and I\", 'start': 2280.8, 'duration': 8.64}, {'text': 'can correct and say in each slide I want', 'start': 2283.839, 'duration': 8.041}, {'text': 'a', 'start': 2289.44, 'duration': 7.6}, {'text': 'catchy slide title a', 'start': 2291.88, 'duration': 5.16}, {'text': 'subtitle', 'start': 2297.64, 'duration': 6.92}, {'text': 'three bullet points and a nade on', 'start': 2299.16, 'duration': 7.72}, {'text': 'visuals', 'start': 2304.56, 'duration': 6.64}, {'text': 'And when you do that you get much deeper', 'start': 2306.88, 'duration': 7.36}, {'text': 'help because on top of of having the', 'start': 2311.2, 'duration': 7.0}, {'text': 'text you also get help on Hulk you could', 'start': 2314.24, 'duration': 6.96}, {'text': 'visualize the information that has to go', 'start': 2318.2, 'duration': 6.52}, {'text': 'on the slide And you see by just adding', 'start': 2321.2, 'duration': 7.2}, {'text': 'this we got much more value from the', 'start': 2324.72, 'duration': 8.16}, {'text': \"output of CH GPD Now let's try with the\", 'start': 2328.4, 'duration': 7.28}, {'text': 'CSI and FBI formula', 'start': 2332.88, 'duration': 8.4}, {'text': 'So I wrote the same CSI print but I', 'start': 2335.68, 'duration': 7.0}, {'text': 'added to', 'start': 2341.28, 'duration': 5.839}, {'text': 'it the F meaning the format the B the', 'start': 2342.68, 'duration': 6.439}, {'text': 'blueprint So the content needs to be', 'start': 2347.119, 'duration': 4.561}, {'text': 'concise and each section needs to have a', 'start': 2349.119, 'duration': 5.201}, {'text': 'key takeaway section and on top I asked', 'start': 2351.68, 'duration': 4.96}, {'text': 'to draft the outline like a financial', 'start': 2354.32, 'duration': 6.08}, {'text': \"expert Let's see the output of this\", 'start': 2356.64, 'duration': 7.76}, {'text': 'prompt Now you can see that I have those', 'start': 2360.4, 'duration': 7.439}, {'text': 'slides I have the key takeaway And if I', 'start': 2364.4, 'duration': 7.28}, {'text': 'wanted to have exactly the same output', 'start': 2367.839, 'duration': 6.321}, {'text': 'than what we tried in the beginning of', 'start': 2371.68, 'duration': 4.56}, {'text': 'the exercise I could just change here', 'start': 2374.16, 'duration': 4.36}, {'text': 'the blueprint and say the content is', 'start': 2376.24, 'duration': 8.8}, {'text': 'concise The title of each slide needs to', 'start': 2378.52, 'duration': 11.559}, {'text': 'be catchy add a subtitle in each slide', 'start': 2385.04, 'duration': 11.6}, {'text': 'and add a visual aid for each slide Now', 'start': 2390.079, 'duration': 9.0}, {'text': \"let's see if we have a better\", 'start': 2396.64, 'duration': 5.28}, {'text': 'output So here is really good to show', 'start': 2399.079, 'duration': 4.76}, {'text': 'you that because I made a mistake I', 'start': 2401.92, 'duration': 4.4}, {'text': 'forgot to mention the three bullet', 'start': 2403.839, 'duration': 7.441}, {'text': 'points So now I will just change and say', 'start': 2406.32, 'duration': 9.039}, {'text': 'add three bullet points on each slide', 'start': 2411.28, 'duration': 6.319}, {'text': 'because you could see here I only have', 'start': 2415.359, 'duration': 4.921}, {'text': 'the title the subtitle and the visual', 'start': 2417.599, 'duration': 6.921}, {'text': \"aid So let's try\", 'start': 2420.28, 'duration': 8.44}, {'text': 'again And now I have my slide my title', 'start': 2424.52, 'duration': 6.92}, {'text': 'my subtitle and the breakdown in three', 'start': 2428.72, 'duration': 4.639}, {'text': \"bullet points What I don't have though\", 'start': 2431.44, 'duration': 5.32}, {'text': 'is a catchy title So here I could', 'start': 2433.359, 'duration': 8.0}, {'text': 'ask reddraft it and change all the', 'start': 2436.76, 'duration': 6.64}, {'text': 'titles to', 'start': 2441.359, 'duration': 5.041}, {'text': 'catchy', 'start': 2443.4, 'duration': 3.0}, {'text': 'titles And now each slide has a catchy', 'start': 2446.44, 'duration': 5.8}, {'text': 'title This is how you improve your', 'start': 2449.68, 'duration': 5.28}, {'text': 'content also by iteration Now that you', 'start': 2452.24, 'duration': 5.52}, {'text': 'have seen how to prompt and understood', 'start': 2454.96, 'duration': 5.76}, {'text': \"the CSI and FBI formula let's go deeper\", 'start': 2457.76, 'duration': 7.079}, {'text': 'to get more from TDP', 'start': 2460.72, 'duration': 4.119}, {'text': \"Let's address an important topic The\", 'start': 2465.52, 'duration': 5.599}, {'text': 'risks and limitations of using CH GPT or', 'start': 2467.68, 'duration': 6.32}, {'text': 'any other LLMs that are open The first', 'start': 2471.119, 'duration': 5.2}, {'text': 'one as you may have heard if you are', 'start': 2474.0, 'duration': 5.28}, {'text': 'using something like CHPT you are using', 'start': 2476.319, 'duration': 5.441}, {'text': 'an external tool that is not part of', 'start': 2479.28, 'duration': 5.2}, {'text': 'your company and with whom probably your', 'start': 2481.76, 'duration': 5.04}, {'text': \"company didn't sign a contract So if\", 'start': 2484.48, 'duration': 4.879}, {'text': \"your company doesn't have THP enterprise\", 'start': 2486.8, 'duration': 4.48}, {'text': 'it means you are using it privately And', 'start': 2489.359, 'duration': 3.921}, {'text': 'if you are using it privately then this', 'start': 2491.28, 'duration': 5.76}, {'text': 'is like using your own Gmail inbox to', 'start': 2493.28, 'duration': 6.48}, {'text': 'send to yourself private document of', 'start': 2497.04, 'duration': 4.24}, {'text': \"your company This is something you're\", 'start': 2499.76, 'duration': 3.04}, {'text': 'you are not allowed and this is', 'start': 2501.28, 'duration': 4.4}, {'text': 'something your contract should prevent', 'start': 2502.8, 'duration': 4.88}, {'text': 'with a confidentiality close So', 'start': 2505.68, 'duration': 4.159}, {'text': 'confidentiality is really something you', 'start': 2507.68, 'duration': 4.48}, {'text': \"need to pay attention But everything I'm\", 'start': 2509.839, 'duration': 4.961}, {'text': 'going to show in this course is without', 'start': 2512.16, 'duration': 4.88}, {'text': 'this confidentiality problem because we', 'start': 2514.8, 'duration': 3.92}, {'text': 'will never say anything about the', 'start': 2517.04, 'duration': 5.12}, {'text': 'company We will never divate any numbers', 'start': 2518.72, 'duration': 6.879}, {'text': 'We will just use CHPT like a friend or', 'start': 2522.16, 'duration': 6.0}, {'text': 'an ex-colague with whom we exchange on', 'start': 2525.599, 'duration': 5.201}, {'text': 'best practice But we never tell anything', 'start': 2528.16, 'duration': 5.04}, {'text': 'about confidential information from the', 'start': 2530.8, 'duration': 3.6}, {'text': 'company And when you have these', 'start': 2533.2, 'duration': 4.24}, {'text': 'discussions if you are meeting your best', 'start': 2534.4, 'duration': 4.88}, {'text': 'friend at a barbecue and your best', 'start': 2537.44, 'duration': 4.399}, {'text': 'friend is also working in FPNA and you', 'start': 2539.28, 'duration': 4.799}, {'text': 'both talk about how to do a cohort', 'start': 2541.839, 'duration': 4.48}, {'text': 'analysis in Excel you can still learn a', 'start': 2544.079, 'duration': 5.361}, {'text': 'lot you can still help each other a lot', 'start': 2546.319, 'duration': 5.201}, {'text': 'and this is how we are going to use JG', 'start': 2549.44, 'duration': 4.48}, {'text': 'GPT like a good friend that you meet at', 'start': 2551.52, 'duration': 5.2}, {'text': 'a barbecue who is really really smart', 'start': 2553.92, 'duration': 6.72}, {'text': 'but who is also somebody you should not', 'start': 2556.72, 'duration': 6.08}, {'text': 'give any confidential information of', 'start': 2560.64, 'duration': 5.4}, {'text': 'your company the second part is about', 'start': 2562.8, 'duration': 6.96}, {'text': 'security GPT Gemini and Copilot and', 'start': 2566.04, 'duration': 7.0}, {'text': 'other LLMs are also external tool and', 'start': 2569.76, 'duration': 5.76}, {'text': 'anything you put inside is not part of', 'start': 2573.04, 'duration': 5.44}, {'text': 'your security protocol from your company', 'start': 2575.52, 'duration': 5.839}, {'text': \"So it's in the internet and because of\", 'start': 2578.48, 'duration': 5.68}, {'text': 'this you should check with your IT what', 'start': 2581.359, 'duration': 5.201}, {'text': 'you are allowed to do But again to avoid', 'start': 2584.16, 'duration': 4.8}, {'text': \"any risk don't put any confidential\", 'start': 2586.56, 'duration': 5.519}, {'text': 'information Another part is for yourself', 'start': 2588.96, 'duration': 6.159}, {'text': 'on your private data but also data about', 'start': 2592.079, 'duration': 5.28}, {'text': 'your colleagues or about your family or', 'start': 2595.119, 'duration': 4.24}, {'text': 'about friends Make sure you understand', 'start': 2597.359, 'duration': 5.601}, {'text': 'what OpenAI and others like Copilot', 'start': 2599.359, 'duration': 5.441}, {'text': 'Gemini what they are doing with your', 'start': 2602.96, 'duration': 4.08}, {'text': 'data And for this you can read the', 'start': 2604.8, 'duration': 6.08}, {'text': 'privacy policy of all of the LLMs to see', 'start': 2607.04, 'duration': 5.68}, {'text': 'what happens with your data and how you', 'start': 2610.88, 'duration': 4.16}, {'text': 'can protect yourself One of the other', 'start': 2612.72, 'duration': 6.16}, {'text': 'risk from THP is how good is the output', 'start': 2615.04, 'duration': 6.24}, {'text': 'You should think that THP is like a', 'start': 2618.88, 'duration': 5.6}, {'text': \"traininee If you don't coach a traininee\", 'start': 2621.28, 'duration': 5.68}, {'text': 'properly then the output of the trainy', 'start': 2624.48, 'duration': 4.639}, {'text': 'might not be as good as what you expect', 'start': 2626.96, 'duration': 4.56}, {'text': 'Your input in the instructions is really', 'start': 2629.119, 'duration': 4.48}, {'text': 'important But also it might be that your', 'start': 2631.52, 'duration': 3.839}, {'text': \"training doesn't know how to do the work\", 'start': 2633.599, 'duration': 3.281}, {'text': 'And because of this you should not', 'start': 2635.359, 'duration': 4.641}, {'text': 'expect that the quality of CH GPD is', 'start': 2636.88, 'duration': 5.84}, {'text': 'always at the best So you need to review', 'start': 2640.0, 'duration': 5.92}, {'text': 'and own the output and also improve it', 'start': 2642.72, 'duration': 4.8}, {'text': 'to make sure that you can use it for', 'start': 2645.92, 'duration': 3.28}, {'text': 'your work The last part is on', 'start': 2647.52, 'duration': 5.2}, {'text': 'hallucinations JPT Copilot Gemini and', 'start': 2649.2, 'duration': 7.04}, {'text': 'other LLMs are mostly generative AI', 'start': 2652.72, 'duration': 6.8}, {'text': 'meaning they generate output', 'start': 2656.24, 'duration': 4.96}, {'text': \"And it's really good because you can\", 'start': 2659.52, 'duration': 3.92}, {'text': 'create stories you can create', 'start': 2661.2, 'duration': 4.0}, {'text': 'presentations with it and content for', 'start': 2663.44, 'duration': 4.8}, {'text': 'your slides But the risk maybe it will', 'start': 2665.2, 'duration': 5.6}, {'text': \"create an accounting policy that doesn't\", 'start': 2668.24, 'duration': 5.68}, {'text': 'exist like a US GAP policy that is not', 'start': 2670.8, 'duration': 5.76}, {'text': 'in the books or an inventory method that', 'start': 2673.92, 'duration': 4.48}, {'text': \"doesn't make sense Because of this you\", 'start': 2676.56, 'duration': 4.0}, {'text': 'need to use your professional skepticism', 'start': 2678.4, 'duration': 4.959}, {'text': 'to review the work And each time you', 'start': 2680.56, 'duration': 5.6}, {'text': 'want to base a decision on the output of', 'start': 2683.359, 'duration': 5.76}, {'text': 'CHPT you need to review and make sure', 'start': 2686.16, 'duration': 5.28}, {'text': 'that if there are facts to review', 'start': 2689.119, 'duration': 5.041}, {'text': 'another source to see if the facts like', 'start': 2691.44, 'duration': 4.72}, {'text': 'for example accounting policies or US', 'start': 2694.16, 'duration': 4.8}, {'text': 'GAAP or if tax regulations make sure', 'start': 2696.16, 'duration': 5.84}, {'text': 'that those are the latest regulations', 'start': 2698.96, 'duration': 5.68}, {'text': 'the most updated one and also the right', 'start': 2702.0, 'duration': 5.92}, {'text': 'ones Again everything that you will use', 'start': 2704.64, 'duration': 6.32}, {'text': 'from those LLMs become yours And because', 'start': 2707.92, 'duration': 5.159}, {'text': \"it becomes yours it's your own\", 'start': 2710.96, 'duration': 4.879}, {'text': 'responsibility to review it and to own', 'start': 2713.079, 'duration': 4.841}, {'text': \"it and to improve it To finish let's\", 'start': 2715.839, 'duration': 4.641}, {'text': 'look at limitations First you need to', 'start': 2717.92, 'duration': 6.399}, {'text': 'understand that JGPT is not consistently', 'start': 2720.48, 'duration': 7.119}, {'text': 'updated It has some cut of date on which', 'start': 2724.319, 'duration': 5.841}, {'text': 'the model is trained And if you want', 'start': 2727.599, 'duration': 5.201}, {'text': 'something more actual you need to', 'start': 2730.16, 'duration': 5.439}, {'text': 'specifically ask to search the internet', 'start': 2732.8, 'duration': 5.36}, {'text': 'because it will not have the latest', 'start': 2735.599, 'duration': 5.041}, {'text': 'information that are after the cutff', 'start': 2738.16, 'duration': 4.24}, {'text': 'date And you can look at the link', 'start': 2740.64, 'duration': 4.32}, {'text': 'yourself to research when was the last', 'start': 2742.4, 'duration': 4.719}, {'text': \"cut of date as it's changing all the\", 'start': 2744.96, 'duration': 5.04}, {'text': 'time Second this model is probabilistic', 'start': 2747.119, 'duration': 5.601}, {'text': \"We don't have a model that calculates We\", 'start': 2750.0, 'duration': 6.079}, {'text': 'have a model that estimates what is the', 'start': 2752.72, 'duration': 5.599}, {'text': 'most probable answer And we know in', 'start': 2756.079, 'duration': 4.081}, {'text': 'finance we need something accurate So', 'start': 2758.319, 'duration': 5.04}, {'text': 'always have that in mind because TGPT', 'start': 2760.16, 'duration': 6.32}, {'text': 'will always show you different answers', 'start': 2763.359, 'duration': 5.041}, {'text': 'and you know that with figures is not', 'start': 2766.48, 'duration': 4.24}, {'text': 'really ideal So for this what I', 'start': 2768.4, 'duration': 6.48}, {'text': 'recommend is always ask JGPT to show its', 'start': 2770.72, 'duration': 7.04}, {'text': 'calculation and also as much as possible', 'start': 2774.88, 'duration': 5.92}, {'text': 'to let JGPT use Python to calculate', 'start': 2777.76, 'duration': 5.44}, {'text': 'because Python is a calculator and a', 'start': 2780.8, 'duration': 4.96}, {'text': 'calculation from Python You can audit it', 'start': 2783.2, 'duration': 4.399}, {'text': 'because you can look at the code and', 'start': 2785.76, 'duration': 3.92}, {'text': 'audit if it was the right code and if', 'start': 2787.599, 'duration': 4.321}, {'text': 'the code was done on the right data and', 'start': 2789.68, 'duration': 4.08}, {'text': 'like this you can verify if the', 'start': 2791.92, 'duration': 4.0}, {'text': 'information given is correct Other', 'start': 2793.76, 'duration': 4.8}, {'text': 'limitations are the incorrect or', 'start': 2795.92, 'duration': 5.12}, {'text': 'nonsensical answers So because it', 'start': 2798.56, 'duration': 4.24}, {'text': \"doesn't have all of the truth it might\", 'start': 2801.04, 'duration': 4.319}, {'text': 'be that you will get something that', 'start': 2802.8, 'duration': 4.16}, {'text': \"doesn't make sense and doesn't exist\", 'start': 2805.359, 'duration': 4.161}, {'text': 'Other limitations is that based on what', 'start': 2806.96, 'duration': 6.08}, {'text': 'you ask slightly change might change the', 'start': 2809.52, 'duration': 5.52}, {'text': 'answer Something you will notice as well', 'start': 2813.04, 'duration': 5.44}, {'text': 'is that chpt always sounds the same and', 'start': 2815.04, 'duration': 5.6}, {'text': 'this is because the model has been', 'start': 2818.48, 'duration': 5.04}, {'text': 'optimized to answer in a certain way So', 'start': 2820.64, 'duration': 5.199}, {'text': 'this is a limitation because it will not', 'start': 2823.52, 'duration': 6.24}, {'text': 'always adapt its own way of phrasing', 'start': 2825.839, 'duration': 6.081}, {'text': 'concepts based to you You will need to', 'start': 2829.76, 'duration': 4.48}, {'text': 'be really specific on what type of', 'start': 2831.92, 'duration': 4.64}, {'text': 'wording you want And the tip for that is', 'start': 2834.24, 'duration': 5.119}, {'text': 'to ask to write more directly and also', 'start': 2836.56, 'duration': 4.88}, {'text': 'to avoid some of the words that you see', 'start': 2839.359, 'duration': 3.681}, {'text': 'that are always coming back Like for', 'start': 2841.44, 'duration': 4.72}, {'text': 'example we are going to delve into a', 'start': 2843.04, 'duration': 5.36}, {'text': 'topic Delve is really something Chipt', 'start': 2846.16, 'duration': 4.08}, {'text': 'loves to repeat all the time Also', 'start': 2848.4, 'duration': 4.32}, {'text': 'something you need to understand Chpt is', 'start': 2850.24, 'duration': 4.72}, {'text': 'made to make you happy as a user And', 'start': 2852.72, 'duration': 4.72}, {'text': 'instead of making the experience really', 'start': 2854.96, 'duration': 4.72}, {'text': \"hard by asking you what do you want i'm\", 'start': 2857.44, 'duration': 4.08}, {'text': 'not sure Can you repeat again can you', 'start': 2859.68, 'duration': 3.76}, {'text': 'formulate again can you give me more', 'start': 2861.52, 'duration': 3.599}, {'text': 'information it will take a lot of', 'start': 2863.44, 'duration': 4.08}, {'text': 'assumptions if you are not specific', 'start': 2865.119, 'duration': 4.561}, {'text': 'enough So make sure that you understand', 'start': 2867.52, 'duration': 4.319}, {'text': 'that some of the response are based on', 'start': 2869.68, 'duration': 4.32}, {'text': 'assumptions and because of this you', 'start': 2871.839, 'duration': 4.721}, {'text': 'might have the quality that is not what', 'start': 2874.0, 'duration': 5.52}, {'text': 'you expect Finally the last limitation', 'start': 2876.56, 'duration': 5.68}, {'text': 'there are some policies inside the model', 'start': 2879.52, 'duration': 5.76}, {'text': 'that will stop you from asking and', 'start': 2882.24, 'duration': 4.879}, {'text': 'getting output especially if you ask', 'start': 2885.28, 'duration': 4.799}, {'text': 'something dangerous or if you ask', 'start': 2887.119, 'duration': 5.681}, {'text': 'something about somebody famous there is', 'start': 2890.079, 'duration': 4.881}, {'text': 'some limitations And there you have it a', 'start': 2892.8, 'duration': 5.12}, {'text': 'full tutorial on how to use CHGPT as a', 'start': 2894.96, 'duration': 4.48}, {'text': 'finance professional Now even though', 'start': 2897.92, 'duration': 4.88}, {'text': 'this was one of my most in-depth video I', 'start': 2899.44, 'duration': 5.2}, {'text': 'ever did on this channel and also I', 'start': 2902.8, 'duration': 4.16}, {'text': 'covered all of the fundamentals you need', 'start': 2904.64, 'duration': 4.56}, {'text': 'to get started this was actually just', 'start': 2906.96, 'duration': 5.359}, {'text': 'one of eight modules I have made in my', 'start': 2909.2, 'duration': 5.919}, {'text': 'JGPT for finance course So if you want', 'start': 2912.319, 'duration': 4.641}, {'text': 'to take this course this course takes', 'start': 2915.119, 'duration': 4.161}, {'text': 'you from complete beginner to a real pro', 'start': 2916.96, 'duration': 4.639}, {'text': 'to use AI better than all of your', 'start': 2919.28, 'duration': 4.16}, {'text': 'colleagues and be the most valuable', 'start': 2921.599, 'duration': 3.921}, {'text': 'person in your company So if you are', 'start': 2923.44, 'duration': 4.56}, {'text': 'serious about implementing AI into your', 'start': 2925.52, 'duration': 4.72}, {'text': 'work then click in the first link in the', 'start': 2928.0, 'duration': 4.4}, {'text': 'description to check out my full course', 'start': 2930.24, 'duration': 4.96}, {'text': 'It has everything I know about how to', 'start': 2932.4, 'duration': 5.52}, {'text': 'best leverage JGPT as a finance pro And', 'start': 2935.2, 'duration': 4.96}, {'text': 'if this free course was valuable to you', 'start': 2937.92, 'duration': 3.84}, {'text': 'then subscribe to the channel because I', 'start': 2940.16, 'duration': 2.959}, {'text': 'will make more and more of this video', 'start': 2941.76, 'duration': 4.64}, {'text': 'And I hope to see you in the next', 'start': 2943.119, 'duration': 3.281}]} content='this is going to be a completely free\\ncourse on how you can use JGPT if you\\nwork in finance Over the last few years\\nI\\'ve taught thousands of finance\\nprofessionals like you how to use AI to\\nsave hours but also to become\\nirrepressible in this time of changes\\nespecially with all of the pressure we\\ngot in finance Now for the information\\nin this video I normally charge hundreds\\nof dollars but I decided to give it away\\nfor free So close all of your open tabs\\nand pay full attention because I\\nguarantee that if you watch this video\\nuntil the end you will have all of the\\nfoundations you need to start using CHPT\\nin your work In this course I will show\\nyou how you can sign up to Chip and walk\\nyou through all of the functionalities\\nthat are the most important in Chip And\\nthen we will cover the risks and\\nlimitations of using CHP in your work\\nAnd finally you will learn how to write\\nAI prompts like a pro which is arguably\\nthe most valuable skills you should have\\nright\\nnow We are going to learn together how\\nyou can sign up and also what are the\\nfunctionalities of CHP but really\\nimportant we will cover what are the\\nrisks and limitations and then we you\\nwill learn how to prompt the most\\nimportant skill to learn in this course\\nAnd finally you\\'ll get an exercise to\\ntest your capabilities If you are new to\\nchatgptt you can go to\\nchat.openai.com and you will arrive to\\nthis page On this page you will click\\nsign up They will ask you to create an\\naccount You can either use your Google\\ncredential or Microsoft credential or\\neven Apple or just give your email\\naddress Let\\'s imagine that I want to\\nchoose my email address Then I just need\\nto put a password\\nNow that I have entered my password I\\nwill just need to validate my email You\\nshould arrive to this email address\\nwhere you just click verify email\\naddress And now your email is verified\\nYou just need to login again Here OpenAI\\nwill ask you some\\ndetails And you can enter also your\\nbirthday And now we are successfully\\ninside Chpt for the first time Let\\'s see\\nnow how you can use CH GPT Here is the\\nchat box where you are going to start\\nentering your questions and instructions\\nFor example I can ask what are the top\\nthree skills for an FPNA manager and how\\nto learn them Now I\\'m going to click\\nhere or you can press alt\\nenter and you get from chaptt your\\nanswer Now let\\'s imagine that I don\\'t\\nwant three skills but five skills And I\\ncan click this little pen to correct my\\ninstruction And I can ask top five\\nskills And I can ask \"Show me the result\\nin a\\ntable.\" And now I get the result in a\\ntable What you can also do is press this\\ncopy and paste\\nbutton And when you do that it copies\\nall of the text Here for example I will\\njust paste it and I have all of the\\ntests But as you can note the text is\\nnot formatted So what you can do is\\nselect the table like this and copy and\\npaste the table This is my favorite\\nfunction to copy and paste is just\\nselecting the text rather than using\\nthis There is another function which is\\nregenerate Here if I\\'m not happy about\\nthe result I can just click regenerate\\nAnd JGPT will propose something which is\\nan alternative to my question And I can\\ncome back and again here change and I\\ncan say what are the top technical\\nskills and save and submit and you can\\nsee what is happening I have now\\ntechnical skills but look at what is\\nhere Here is a way to come back to all\\nof my previous instructions Like this I\\ncan come back to the history If you want\\nto start a new chat then you will click\\nhere And now I can ask tell me how is a\\ntypical day for FPNA manager\\nIt\\'s really important to change\\ndiscussions each time you are changing\\nsubjects because if you don\\'t do that\\nthen here if I ask what are the top\\nskills it will understand that I\\'m\\ntalking about FPNA manager and not FPNA\\nanalyst because we started the\\ndiscussion by FPNA manager and if my\\nquestion is about FPNA analyst it\\'s\\neither better to change here and to say\\nfor FPN analyst or directly to change a\\nnew discussion and start from scratch if\\nyou feel the two discussions are not\\nconnected The last important part is\\nthis\\nsidebar I will click to see the sidebar\\nand to come back to all of my previous\\ndiscussions And there I can click\\nto manage my settings For example I can\\nask to save or not the chat history that\\nwe used to at the left And also I can\\nactivate or deactivate the training on\\nmy data because for me I have subscribed\\nto CHP teams It\\'s by default\\ndeactivated Finally let\\'s look together\\nat the most advanced functions of JPT\\nYou will see that with JGPT you can also\\nupload a file So you have the button in\\nthe bottom here attachments I\\'m going to\\nupload a file that looks like this So\\nhere is a file with my company sales\\ndata and I am in a SAS industry So I\\nwill want for example to do a cohort\\nanalysis Let\\'s see with JGPT if my\\ncompany has a contract with JGPT and I\\'m\\nallowed to upload data inside JGPT Let\\'s\\nsee what I can do Just as a warning if\\nyour company doesn\\'t have a contract\\nwith OpenAI then check if you\\'re allowed\\nor not to upload data It\\'s better to be\\nsure than to do a mistake So we are back\\nin JGPT and I\\'m going to upload my file\\nand you can see that I can connect to\\nGoogle Drive connect to Microsoft One\\nDrive or upload from my computer So here\\nis the file I\\'m uploading and I will ask\\ncan you create a cohort analysis for my\\nretention rate based on the month of the\\nsubscriber starting doing business with\\nus show on top a visualization because I\\nlike to visualize this\\ninformation So now first we can see that\\nwe have here the table So that means\\nJGPT understood the table with exactly\\nthe right columns It read those columns\\nusing Python And it\\'s really important\\nto know that JGPT is not good at\\ncalculating because it\\'s not made to\\ncalculate but is really good at creating\\nthe code to send them to Python And\\nPython is a good tool to calculate and\\nthat\\'s how you separate the work like\\nyou will do for yourself If you have to\\ncalculate you will use a computer or you\\nwill use a calculator Here is the same\\nyou use the right tool to calculate\\nwhich is Python Now TGPT is coding the\\ncode to create our cohort analysis and\\nyou can see that now I have the mostly\\nretention rates and here it decided to\\nuse red for something positive So what I\\ncan do is even ask to change the color\\nCan you make 100% blue and 0%\\nred but you see that while it\\'s\\ncalculating I want to show you how\\npowerful the tool is because this is\\nimpossible to do so fast in Excel and in\\nExcel also you will have to work with\\nconditional formatting But here I will\\nhave here my file with a proper cohort\\nanalysis And what I can even do is\\ndownload the chart as a picture because\\nPython does it as a picture And I can\\neven here continue to analyze thanks to\\nthese buttons So now for example I ask\\nto analyze the customer turn rate and it\\nwill do this analysis for me And you can\\nsee that now I have a graph and that\\'s\\nwhat I really like here is to continue\\nthe discussion thanks to this button And\\nif you want to audit what happened you\\ncan always look at the code and you can\\nalways ask if you don\\'t really\\nunderstand the code you can ask JGP to\\nexplain you the code like this You can\\nverify if the calculation is\\ncorrect This is a function that allows\\nyou to customize and make your own GPT\\nLet me show you how you can do that So\\nwe are back in JPT and I will go on the\\nside open the sidebar and here you\\nshould see a button called explore GPTs\\nand GPTs are GPT versions that are\\ncustomized by either yourself or other\\npeople This is the marketplace of GPS or\\nthe app store of GPS So for example here\\nyou can type US GAP and it will show you\\nfor example here a US gap advisor where\\nyou can click on you can check if there\\nare good ratings or not and if a lot of\\npeople have used it you can see what it\\ndoes and what we can do is for example\\nstart a chat and I can ask explain me\\nrevenue recognition and I don\\'t even\\nhave to explain that I want to get this\\nexplanation based on US gap because I\\'m\\nalready in a US gap advisor So that\\'s\\nthe big advantage This US gap advisor\\nwas configured to unswear for US GAP and\\nto answer in a specific way And you can\\nsee that here we got a really good\\nanswer because it knows we are finance\\npeople because who would want to go to a\\nUS GAP advisor mostly finance people And\\nwe have already somebody behind who did\\nthe work to customize the elsewhere So\\nlike this we save a lot of time So now\\ndo you know that you can actually also\\ncreate your own GPT let me show you how\\nSo you will go back to explore GPT to\\nthe left and then upright you can create\\nyour own GPT So I will click create and\\nnow I am in the GPT configurator And\\nhere you have two solution Either you\\ncan create through a description or you\\ncan configure yourself by giving a name\\ndescribing what your custom GPT does and\\nthen giving some instruction And what\\nyou can even do after is upload files to\\naugment the knowledge of this GPT Pay\\nattention that if you activate code\\ninterpreter those files are downloadable\\nby the other users If your GPT is public\\nyou can also keep your GPT for yourself\\nBut just know that that this is\\nsomething to pay attention Also you have\\nthe action function here is more if you\\nwant to use code to get data from\\noutside through API and this is too\\nadvanced to cover that in this course\\nBut if you don\\'t know how to start a fun\\nway is to use the create view because\\nthere you can just create your GPT\\nthrough a discussion So now let me show\\nyou how to create a GPT So you will go\\nin the create view which is the easiest\\none if you just started Let\\'s imagine\\nthat we want a coach to be better at\\nFPNA I want to create a coach to get\\nbetter at FPNA I will just say that and\\nlet\\'s see the discussion Yeah I\\'m happy\\nabout this name So\\nyes and you can see that TGPT even\\ncreated a logo for our FPNA coach So I\\'m\\nhappy Yes I like\\nit So what should be emphasized so here\\nI want that this coach shows me the\\ntechnical aspect of being a FPNA pro and\\nshows me both the theory and practical\\nexamples and step by\\nstep So now that I\\'ve done that this is\\nmy instructions to improve this GPT and\\nlike this I don\\'t have to repeat after\\nhow this GPT will act So I want more\\nsomething direct and maybe I need also\\nmotivational\\nSo now the GPT is actually finished and\\nwe can try it So for example I can try\\nhow can I improve my budgeting process\\nSo you see we have the answer and either\\nyou are happy and you can create it or\\nyou can even do something is you go to\\nconfigure you see how was this GPT\\nconfigurated and you can change\\nsomething for example you say use cables\\nto present information when possible and\\nyou can even change your conversation\\nstarters so now I will create the GPT\\nand that\\'s where you can either have it\\non invite only or anyone in your\\norganization because I have a paid JGPT\\nversion which is teams or for companies\\nthey have enterprise So for me in my\\norganization I can decide who has access\\nor I can put it on the GPT store or if\\nyou only want that specific people have\\nyou can just share them the link like\\nthis is not on the GPT store but it\\'s\\nstill available for people with the link\\nSo for now I just wanted to keep it for\\nmyself So I will click update and then I\\nwill view this GPT and this GPT will\\nalso appear on the left with all of my\\nother GPT And now I can ask for example\\nhow to do a cohort analysis I will enter\\nAnd this will help me as a coach And you\\nsee that it follows the step by step\\nlike I wanted Again you see how you save\\ntime because I don\\'t have to repeat that\\nI want to have advices shown me in a\\nstep by step It will always do it like\\nthis And on top it combines theory with\\npractical advice and step-by-step\\ninstructions\\nIn this lesson we will explore Canvas a\\npowerful feature in TGPT that helps\\nfinance professionals draft reports\\ncommentaries and even modify code in\\nreal time all in a seamless workspace\\nLet me show you how to use it Let\\'s take\\nthe P&L of Nvidia as an example and\\nlet\\'s ask JGPT to draft commentaries on\\nthis P&L I\\'m going to take a\\nscreenshot I\\'ll now back in GPT and I\\nwill paste the screenshot and I will say\\ndraft the financial commentaries on the\\nP&L of this company And here I can\\nactivate canvas Sometimes it activates\\nautomatically but I really want to use\\nthe canvas function and you will\\nunderstand why So let\\'s launch it and\\nsee what is\\nhappening Now we have a totally new view\\nthat is coming because on the right side\\nmy document is getting written by JGPT\\nand on the left side I have the\\nconversation that is still active where\\nyou can see on the upper left the\\nscreenshot and my prompt and here the\\nanswer from JGPT So what can we do on\\nthe right side with this document well\\nlet me show you the different\\nfunctionalities on the bottom right If I\\ngo over the pen I have different type of\\nedits I can add emojis which we don\\'t\\nreally need for a factual commentary I\\ncan polish it I can change the reading\\nlevel and the length Let\\'s see here if I\\nask to polish it Push it means that it\\nwill review the text and makes it even\\nbetter So you can see that it\\'s changing\\nslowly all of the text Now the position\\nis done Let\\'s see what are the other\\nfunction The reading level maybe let\\'s\\nmake it more either easy or advanced\\nLet\\'s put it graduate school to see if\\nour wording is changing Okay so you can\\nsee we are using different types of\\nwording and adjectives and you can\\nyourself see which one you like the most\\nfor your style Okay now let\\'s try\\nsomething else Let\\'s make it shorter\\nbecause it\\'s a bit too long So now I\\nwill really make it the shortest as\\npossible And you can see that now the\\ndocument is much shorter So this is\\nactually already good because we can\\nchange really quickly But look what I\\ncan also do here For example I can\\nselect the text and I can ask CH GPT to\\nchange it to add that the growth comes\\nfrom AI So let\\'s add some flavor in the\\ntext from AI and\\nblockchain And we can see here that is\\nrewriting it adding that the growth is\\ncurrently driven by advancement in AI\\nand blockchain technology Of course this\\npart is you who is driving the seats You\\ncan also say that oh I don\\'t really like\\nblockchain technology so let\\'s just\\nremove it and now I just delete it and I\\ncan write especially in the last month\\nSo that\\'s what is good is now you are\\nalso in control and you can write over\\nit You can even highlight here in bold\\nYou can even here change the headings\\nAll of this you are in control The last\\nimportant functionality of Canva is with\\ncoding So here I ask JGPT to create a\\ncode to illustrate a scenario analysis\\nand you will see that if I activate\\nagain the Canva we will get the code\\nwritten in Canva and I can go on and\\nalso modify this code But what is even\\nbetter is there is the button run So I\\ncan run the code and it will for me test\\nif the code is working or it will also\\ntell me if there is any bug So for\\nexample here tells me that it doesn\\'t\\nhave this type of libraries So I can ask\\nto fix the bugs and you can see that now\\nthat is fixing the bug is going through\\nall of the code to improve the code to\\nmake it work So this is useful if you\\nwant either to change part of the code\\nSo because you can select a part and\\njust debug the part that is not working\\nor if you also want to test the code\\ninside GPT directly So go and try it by\\nyourself especially for the writing part\\nand I\\'m sure you will love this new\\nfunctionality Let me show you the TGPT\\nsearch functionality and how to use it\\nin finance I\\'m back in TGPT and you can\\nsee that here there is the button search\\nSo if you really want to get online and\\nget sources then I will advise you to\\nclick to search the web But sometimes\\nalso without clicking it If you ask for\\nexample an information that is recent If\\nI say what is the price of the Nvidia\\nstock from last Friday then let\\'s see\\nwhat is going to happen You can see that\\nit automatically search the web from my\\nquestion even if I didn\\'t ask it And now\\nI have also February 14 2025 So that\\'s\\nthe date when I recorded this video And\\nwe can see here the price stock of\\nNvidia And what is also really good is\\nyou have the sources where you can find\\nthis information So like this you have\\none realtime information two you have\\nsources Now let\\'s look at a real use\\ncase for us in finance When clicking\\nsearch I\\'m going to ask specifically on\\nwhich website I want to get my\\ninformation and what I want to get So if\\nI click you\\'ll see that it will\\nautomatically search the web What is\\nreally good is that now I have the\\ndifferent types of publication on the\\nIFS 15 I can try this one for example\\nand now I am already inside the\\ndocumentation So that\\'s a great\\ntimesaver and also what is really good\\nis you have always the link of the\\npublication Let me now present you the\\nproject functionalities from JPT I\\'m\\nback in Chip and here on the left side\\nif you have a pro account or a teams or\\nenterprise account you can see that\\nthere is the projects section here you\\nwill see the list of your projects and\\nyou can also create a new project that\\'s\\nwhat we are going to do so I create a\\nproject let\\'s imagine that I want to\\nwork on the budget for 2025 let\\'s create\\nthe project so now I am inside my\\nproject I can close here the view and\\nhere for the project I can add file case\\nSo I have two files Let me show you I\\nhave here my P&L for 22 23 24 and the\\nbudget 24 And I have here also a file\\nwith my budget instructions and some\\ndeadlines for this year I\\'ve made a PDF\\nof this document And this PDF I will\\nupload in my project To upload I just\\nneed to go and click add files and I\\nwill add my\\ninstructions Here I\\'ve added my budget\\ninstructions and now I\\'m adding the P&L\\nOkay So we are set I can close and I can\\nalso add instructions here For example I\\ncan say this project is for the FPN team\\nof a manufacturing company We are in the\\nUS and we work with Netswuite You can\\nadd any type of details You can also add\\nthe way you want JPT transfer Here I can\\nsay your responses should be direct and\\nuse bullet points and tables as much as\\npossible So that\\'s the advantage here is\\nyou will see that those instructions\\nthey will be for any chat I start within\\nthis project So now let\\'s imagine that I\\nwant to draft an email to the sales team\\nDraft an email to the sales team to ask\\ntheir input for the budget and I can say\\nmention the\\ndeadlines So here you can see that the\\ndeadline October 2025 I didn\\'t have to\\nsay which deadline it was because\\nactually here in the instructions I\\nalready had my details and you can see\\nhere what we needed to have I can find\\nit here back in the key areas of input\\nSo that\\'s really practical because here\\nin this project there is a knowledge\\nthat you can use based on the document\\nyou uploaded but also based on the\\ninstructions Now I can go back to my\\nbudget 2025 and I can say draft PNL 2025\\nformat with the comparison with\\n2024 and again here it has the knowledge\\nof the P&L of last year like this I can\\nalways reuse the information I opened it\\nbefore So you can see here the answer\\nwhere it assumed an increase of 10% We\\ndidn\\'t ask that But what is interesting\\nis that we have already the table for\\n2024 and budget 2024 without having to\\ngive the information because it was in\\nthe Excel I provided before And now you\\ncan always go back when you are in the\\nbudget 2025 project You can go back in\\nall of the charts that are made under\\nthis project And also let\\'s imagine that\\nyou had here another chart you can\\nalways add to the project 2025 So that\\'s\\nalso really advantageous because you can\\norganize your chat based on some topics\\neven if you don\\'t add any files or even\\nif you don\\'t add any instructions I have\\nsome friends who are using it like this\\nto basically organize their chat but\\nthey don\\'t give any instructions or no\\nprojects files inside And you might\\nwonder what is the difference with\\ncustom GPT First here you can choose the\\nmodel Here you can choose GPT4 or O1 or\\nthis for O mini or GPT for AC model and\\nin GPS you cannot choose the model\\nThat\\'s number one Number two here you\\ncannot really share this chat with\\nsomebody else but in GPS you can\\nactually share the chat with your team\\nand you can even also share your GPS\\nwith your team or with people outside of\\nyour organization if you decide to make\\nit public to make project no project is\\nonly personal and here you see as well\\nhow easy it is to give your instructions\\nand add the document for custom GPT you\\nneed to take a bit more time to\\ncustomize your GPT So I will say really\\nthat project are different than GPS GPS\\nis more like an agent They will have a\\nspecific behavior and project is more to\\nchat in a closed environment where you\\nhave some information that are specific\\nto the chat and you are going to create\\nseveral chats about those information\\nand those custom instructions One cool\\nway also to use projects is to upload a\\nlot of information about you or about a\\nbook or about a topic and then chat with\\nthis information Trade by yourself if\\nyou have a pro a team or enterprise\\nversion See how you can use it the best\\nway for your own use\\ncases Let\\'s look now at the new\\nfunctionality from Chpt which is the\\noperator Operator is like an agent that\\nhas access to your computer and can act\\non your behalf if you prompt it Operator\\nis not yet available in all of the\\ncountries So let me show you a\\npresentation of OpenAI showing how\\noperator is working Here is the\\npresentation from OpenAI showing the\\noperator So now they show it mostly for\\nindividuals like shopping or doing a\\nreservation for a table Here you can see\\nthat OpenAI is showing an example of how\\nto use operator to book for you a table\\nSo here you see that once we prompted to\\nbook a dinner\\nreservation operator is starting to act\\ninside your browser On the left side you\\nhave the discussion you have with your\\noperator and on the right side it\\'s the\\noperator doing it everything by its own\\nYou can see that now it\\'s selecting the\\ndate It\\'s selecting what type of food\\nand it\\'s doing all of this on its own So\\nbasically you don\\'t control it You just\\nlet it do on the side while looking at\\nit to verify that it\\'s doing what you\\nwanted And on the left side you can see\\nall of the actions that were done So\\nhere you can see that the operator is\\ntelling you that the closest reservation\\nis as from 7:45 You say \"Okay let\\'s go\\nLet\\'s do it.\" And then the operator is\\ngoing to book it for you You can see now\\nyou just need to give your details and\\nthen if you say yes the operator is\\ngoing to click and then your reservation\\nis done What does it mean for us well\\nthink about this report where you have\\nto upload data manually You know like\\nkey in the URL then doing a copy and\\npaste in some of the fields That\\'s a\\ngood use case for the operator Think\\nalso of this website where you have to\\ngo and have some steps in those websites\\nto download invoices That\\'s also a good\\ntask for the operator Why because once\\nyou do the task with the operator it can\\nmemorize it and next time you just have\\nto ask repeat this task and then the\\ntask will be done by itself So of course\\nthere are some confidentiality data to\\nthink about which access do you want to\\ngive to this operator also do you want\\nto give your credit card details So\\nthat\\'s something to think about But this\\nis really the start where your AI tool\\nis going to be an agent for you and is\\ngoing to be triggered by you or by\\nscheduled task You could for example to\\nask every week to pull some data or\\nevery month to get some invoices or fill\\nup some vendor forms where you have to\\ngive all of your details like banking\\ndetails name tax ID etc etc And if you\\nwant to learn more I recommend this\\nvideo from Ector Garcia where he\\'s\\nshowing us how operator can help on some\\nQuickBooks task that you might do now\\nmanually But now with operator you can\\ndelegate this task to this agent Go and\\ncheck\\nit Let me now show you how you can\\ncustomize Chip So on the right side you\\nhave this diamond icon You will click on\\nit and you just have to click customize\\nGPT And this is here where you can put\\nyour name what do you do what traits\\nshould chip have So if you like a formal\\ntone or something a bit more casual and\\nthat\\'s where here you can tell chip to\\nwrite like you if you have a specific\\ntone And you see here you have already\\nsome suggestion Here is where you can\\nexplain chip more about you like where\\nyou work maybe your industry maybe your\\nenvironment So if you are in the US or\\nEurope that\\'s important to explain And\\nif I click to advanced then I can also\\nselect the JGPT capabilities For example\\nif you don\\'t really like that it always\\ngoes to the web search you can just\\ndeactivate it that now another way to\\nget TGPT to get customized is to\\nactivate the memory So when I click to\\nsettings and I go to personalization I\\ncan ask to activate the memory And you\\ncan see what the memory can do is\\nbasically along your messages If you\\nexplain where you work if you explain a\\nbit about you then it will start\\npopulate here the memories Here is a\\ngreat way to let the AI personalize the\\nexperience for you And what is good is\\nyou can manage it yourself if you want\\nor you can deactivate it if you prefer\\nto have a neutral AI that doesn\\'t know\\nabout you or maybe just you decide when\\nyou want that it knows about you So you\\ncan do that in a custom GPT or in a\\nproject and you can decide in which chat\\nwhere you want no information about you\\nbecause you don\\'t want to influence the\\noutput So go ahead and check that for\\nyourself Let\\'s start in the meat of this\\ncourse how to prompt with CHP I\\'m going\\nto teach you my framework that works\\neverywhere But first let\\'s understand\\nwhy do we need a framework The first\\nmistake we all do when we start using\\ntool like JGPT is to use the Google\\nmethod The Google method is simply what\\nwe have been doing the last 15 years is\\nto research something based on keywords\\nAs we are humans we are lazy It\\'s often\\nour favorite method because we have to\\nwrite the list word to get the maximum\\noutput Let\\'s see together what happens\\nif we are using JG GPT like Google\\nImagine that I want to issue reminder to\\none of my client who didn\\'t pick I will\\nwrite inside\\nJPT the name letter Let\\'s see what\\nhappens So we get something quite fast\\nbut we only get a description of what is\\nthe name letter\\nThe problem is that it doesn\\'t really\\nhelp us because what we wanted is to\\nsend the letter We wanted to have the\\nletter drafted So this is not working\\nwell because it doesn\\'t really help us\\nwith our work And that\\'s where 90% of\\nthe people will usually stop because\\nthey think tragic is smart but not\\nhelpful And imagine if you are working\\nwith somebody imagine that you want them\\nto write the dunning letter for you Are\\nyou really going just to send them in\\nteams in a chat dunning letter no you\\nare going to explain what is the problem\\nand what you want from them Well chat\\nGPT is exactly the same way is like an\\nassistant to whom you need to explain\\nthe context and be specific about what\\nyou want and also give a clear\\ninstruction to have clear expectation\\nand have a clear output from chip This\\nis why I\\'m bringing a team to solve a\\ncrime which is called CSI With my\\nframework called the CSI framework you\\nare going to prompt using the CSI\\nletters meaning C for context The\\ncontext is basically explaining who you\\nare and the context of your company E is\\nbeing specific What is your problem i is\\nto give a clear instruction because if\\nyou just stop at context and specific\\nthen you still let too much freedom to\\ncharge GPT to generate the output for\\nyou Now let\\'s use our example of dueling\\nletter and let\\'s see how we can use CSI\\nwith I am an accountant being the\\ncontext my client as an overview being\\nthe specificity and the problem and\\nbeing specific about the instruction Now\\nlet\\'s see if we use the CSI framework if\\nwe have a better output from chip\\nAlready it starts to write the letter\\nand we can see that we have exactly a\\nletter on overdue on the outing balance\\nand we ask nicely about the reminder and\\nwe also ask what is the status So we are\\nquite gentle Now the problem is that\\nthis client didn\\'t pay since a long time\\nand it\\'s not the first time that we have\\nthis problem So do you think this letter\\nwill be enough to get our money back no\\nit\\'s not enough because we are too\\ngentle So how are we going to solve this\\nif we want to solve this problem we are\\ngoing to bring a second team to make\\nsure we get our money back and we are\\ngoing to bring the FBI FBI stands for F\\nthe format B the blueprint and I\\nidentity If we get back our example\\nwhere at the beginning we had this\\nprompt I am an accountant My client has\\nnever due for two months Can you draft a\\ncommunication to them we are going to\\nadd F the format So we are going to have\\na formal letter B my favorite part the\\nblueprint We have something in mind but\\nwe don\\'t really know how to say that So\\nwe will let CHGPT formulate for us and\\njust using words like tone to be hard or\\nlegal actions we know it\\'s going to have\\nmuch more impact than our previous\\nletter and finally I for identity where\\nwe will ask to act as the best lawyer\\nand here identity you don\\'t always need\\nit but it could be really useful in this\\ntype of situation Let\\'s go into ch to\\nsee how we are going to use this\\nframework Let\\'s see now the output once\\nwe use the FBI formula on top of CSI You\\ncan see it started with a letter head\\nand we have already in our title legal\\nactions You will also notice that our\\ntone has changed and we are using words\\nto escalate the situation as we wanted\\nWe have words like lawsuit seeking a\\ncourt judgment or engaging debt\\ncollection\\nagencies So what do you think of this\\nletter do you think this will have more\\nimpact than the previous one and here is\\nan additional tip You can also ask can\\nyou list the legal\\naction in a bullet point list and\\nreddraft this letter i like to do it\\nthat way because you see here for a\\nhuman is actually much easier to read in\\nbullet points and we know that if\\nsomebody came across this letter they\\nwill first read that And I assure you if\\nyou read that I\\'m sure your client or\\nyourself you are going to react on this\\nThis is how you can use the framework\\nCSI plus FBI And I recommend you to\\nfirst try all of your prompts in CHP\\nwith at least CSI and when you want\\nsomething more complex use FBI And the\\nmore you will get used to this framework\\nthe more you will see what works and the\\nmore you can take freedom from this\\nframework and build your own frameworks\\nbased on your\\nexperience Let\\'s start with an exercise\\nto practice the CSI and FBI formula\\nImagine that you are the FPNA analyst of\\na brewery and you have to create a\\nweekly financial report You are going to\\nuse JG GPT using the CSI formula to\\ngenerate this report That\\'s the first\\npart When you are done you are going to\\nuse CSI and FBI to improve your prompt\\nand also with that to improve your\\noutput Take five minutes to do this\\nexercise and I see you in the\\nsolution\\nRemember the exercise was to create a\\nweekly financial report for your FPNA\\nmanager Let\\'s see first what happens\\nwhen you use the CSI formula So here I\\nwrote the context I am FPN analyst for a\\nbrewery Then I\\nexplain what is my specific problem I\\nneed to deliver the weekly financial\\nreport to my manager and I ask\\nspecifically to draft the report outline\\nLet\\'s see what\\nhappens So as you can see we have the\\nweekly financial report with adapted\\nwith the beer types Also we have\\nproduction cost which you don\\'t have for\\na SAS and also inventory topics So it\\nunderstood that we are a beer or brewery\\ncompany with production cost So that\\'s\\nreally good because if you now say that\\nyou are a SAS company you\\'ll have a\\ndifferent type of report Now imagine\\nthat what you wanted is actually slides\\nand this is not enough for slides\\nbecause you\\'ll have to readapt it So\\neither you can continue the discussion\\nand say adapt it to\\ncreate the\\nslide structure from this\\nAnd now we have the slide name for each\\nof them Or I\\'ll be more specific and I\\ncan correct and say in each slide I want\\na\\ncatchy slide title a\\nsubtitle\\nthree bullet points and a nade on\\nvisuals\\nAnd when you do that you get much deeper\\nhelp because on top of of having the\\ntext you also get help on Hulk you could\\nvisualize the information that has to go\\non the slide And you see by just adding\\nthis we got much more value from the\\noutput of CH GPD Now let\\'s try with the\\nCSI and FBI formula\\nSo I wrote the same CSI print but I\\nadded to\\nit the F meaning the format the B the\\nblueprint So the content needs to be\\nconcise and each section needs to have a\\nkey takeaway section and on top I asked\\nto draft the outline like a financial\\nexpert Let\\'s see the output of this\\nprompt Now you can see that I have those\\nslides I have the key takeaway And if I\\nwanted to have exactly the same output\\nthan what we tried in the beginning of\\nthe exercise I could just change here\\nthe blueprint and say the content is\\nconcise The title of each slide needs to\\nbe catchy add a subtitle in each slide\\nand add a visual aid for each slide Now\\nlet\\'s see if we have a better\\noutput So here is really good to show\\nyou that because I made a mistake I\\nforgot to mention the three bullet\\npoints So now I will just change and say\\nadd three bullet points on each slide\\nbecause you could see here I only have\\nthe title the subtitle and the visual\\naid So let\\'s try\\nagain And now I have my slide my title\\nmy subtitle and the breakdown in three\\nbullet points What I don\\'t have though\\nis a catchy title So here I could\\nask reddraft it and change all the\\ntitles to\\ncatchy\\ntitles And now each slide has a catchy\\ntitle This is how you improve your\\ncontent also by iteration Now that you\\nhave seen how to prompt and understood\\nthe CSI and FBI formula let\\'s go deeper\\nto get more from TDP\\nLet\\'s address an important topic The\\nrisks and limitations of using CH GPT or\\nany other LLMs that are open The first\\none as you may have heard if you are\\nusing something like CHPT you are using\\nan external tool that is not part of\\nyour company and with whom probably your\\ncompany didn\\'t sign a contract So if\\nyour company doesn\\'t have THP enterprise\\nit means you are using it privately And\\nif you are using it privately then this\\nis like using your own Gmail inbox to\\nsend to yourself private document of\\nyour company This is something you\\'re\\nyou are not allowed and this is\\nsomething your contract should prevent\\nwith a confidentiality close So\\nconfidentiality is really something you\\nneed to pay attention But everything I\\'m\\ngoing to show in this course is without\\nthis confidentiality problem because we\\nwill never say anything about the\\ncompany We will never divate any numbers\\nWe will just use CHPT like a friend or\\nan ex-colague with whom we exchange on\\nbest practice But we never tell anything\\nabout confidential information from the\\ncompany And when you have these\\ndiscussions if you are meeting your best\\nfriend at a barbecue and your best\\nfriend is also working in FPNA and you\\nboth talk about how to do a cohort\\nanalysis in Excel you can still learn a\\nlot you can still help each other a lot\\nand this is how we are going to use JG\\nGPT like a good friend that you meet at\\na barbecue who is really really smart\\nbut who is also somebody you should not\\ngive any confidential information of\\nyour company the second part is about\\nsecurity GPT Gemini and Copilot and\\nother LLMs are also external tool and\\nanything you put inside is not part of\\nyour security protocol from your company\\nSo it\\'s in the internet and because of\\nthis you should check with your IT what\\nyou are allowed to do But again to avoid\\nany risk don\\'t put any confidential\\ninformation Another part is for yourself\\non your private data but also data about\\nyour colleagues or about your family or\\nabout friends Make sure you understand\\nwhat OpenAI and others like Copilot\\nGemini what they are doing with your\\ndata And for this you can read the\\nprivacy policy of all of the LLMs to see\\nwhat happens with your data and how you\\ncan protect yourself One of the other\\nrisk from THP is how good is the output\\nYou should think that THP is like a\\ntraininee If you don\\'t coach a traininee\\nproperly then the output of the trainy\\nmight not be as good as what you expect\\nYour input in the instructions is really\\nimportant But also it might be that your\\ntraining doesn\\'t know how to do the work\\nAnd because of this you should not\\nexpect that the quality of CH GPD is\\nalways at the best So you need to review\\nand own the output and also improve it\\nto make sure that you can use it for\\nyour work The last part is on\\nhallucinations JPT Copilot Gemini and\\nother LLMs are mostly generative AI\\nmeaning they generate output\\nAnd it\\'s really good because you can\\ncreate stories you can create\\npresentations with it and content for\\nyour slides But the risk maybe it will\\ncreate an accounting policy that doesn\\'t\\nexist like a US GAP policy that is not\\nin the books or an inventory method that\\ndoesn\\'t make sense Because of this you\\nneed to use your professional skepticism\\nto review the work And each time you\\nwant to base a decision on the output of\\nCHPT you need to review and make sure\\nthat if there are facts to review\\nanother source to see if the facts like\\nfor example accounting policies or US\\nGAAP or if tax regulations make sure\\nthat those are the latest regulations\\nthe most updated one and also the right\\nones Again everything that you will use\\nfrom those LLMs become yours And because\\nit becomes yours it\\'s your own\\nresponsibility to review it and to own\\nit and to improve it To finish let\\'s\\nlook at limitations First you need to\\nunderstand that JGPT is not consistently\\nupdated It has some cut of date on which\\nthe model is trained And if you want\\nsomething more actual you need to\\nspecifically ask to search the internet\\nbecause it will not have the latest\\ninformation that are after the cutff\\ndate And you can look at the link\\nyourself to research when was the last\\ncut of date as it\\'s changing all the\\ntime Second this model is probabilistic\\nWe don\\'t have a model that calculates We\\nhave a model that estimates what is the\\nmost probable answer And we know in\\nfinance we need something accurate So\\nalways have that in mind because TGPT\\nwill always show you different answers\\nand you know that with figures is not\\nreally ideal So for this what I\\nrecommend is always ask JGPT to show its\\ncalculation and also as much as possible\\nto let JGPT use Python to calculate\\nbecause Python is a calculator and a\\ncalculation from Python You can audit it\\nbecause you can look at the code and\\naudit if it was the right code and if\\nthe code was done on the right data and\\nlike this you can verify if the\\ninformation given is correct Other\\nlimitations are the incorrect or\\nnonsensical answers So because it\\ndoesn\\'t have all of the truth it might\\nbe that you will get something that\\ndoesn\\'t make sense and doesn\\'t exist\\nOther limitations is that based on what\\nyou ask slightly change might change the\\nanswer Something you will notice as well\\nis that chpt always sounds the same and\\nthis is because the model has been\\noptimized to answer in a certain way So\\nthis is a limitation because it will not\\nalways adapt its own way of phrasing\\nconcepts based to you You will need to\\nbe really specific on what type of\\nwording you want And the tip for that is\\nto ask to write more directly and also\\nto avoid some of the words that you see\\nthat are always coming back Like for\\nexample we are going to delve into a\\ntopic Delve is really something Chipt\\nloves to repeat all the time Also\\nsomething you need to understand Chpt is\\nmade to make you happy as a user And\\ninstead of making the experience really\\nhard by asking you what do you want i\\'m\\nnot sure Can you repeat again can you\\nformulate again can you give me more\\ninformation it will take a lot of\\nassumptions if you are not specific\\nenough So make sure that you understand\\nthat some of the response are based on\\nassumptions and because of this you\\nmight have the quality that is not what\\nyou expect Finally the last limitation\\nthere are some policies inside the model\\nthat will stop you from asking and\\ngetting output especially if you ask\\nsomething dangerous or if you ask\\nsomething about somebody famous there is\\nsome limitations And there you have it a\\nfull tutorial on how to use CHGPT as a\\nfinance professional Now even though\\nthis was one of my most in-depth video I\\never did on this channel and also I\\ncovered all of the fundamentals you need\\nto get started this was actually just\\none of eight modules I have made in my\\nJGPT for finance course So if you want\\nto take this course this course takes\\nyou from complete beginner to a real pro\\nto use AI better than all of your\\ncolleagues and be the most valuable\\nperson in your company So if you are\\nserious about implementing AI into your\\nwork then click in the first link in the\\ndescription to check out my full course\\nIt has everything I know about how to\\nbest leverage JGPT as a finance pro And\\nif this free course was valuable to you\\nthen subscribe to the channel because I\\nwill make more and more of this video\\nAnd I hope to see you in the next'\n",
-      "-----\n",
-      "This is going to be a completely free course on how you can use JGPT if you work in finance. Over the last few years, I've taught thousands of finance professionals how to use AI to save hours and become indispensable during these times of change, especially with the pressure in finance. Normally, I charge hundreds of dollars for this information, but I decided to give it away for free. So close all of your open tabs and pay full attention because I guarantee that if you watch this video until the end, you will have all the foundations you need to start using JGPT in your work.\n",
-      "\n",
-      "In this course, I will show you how to sign up for JGPT and walk you through the most important functionalities. We will cover the risks and limitations of using JGPT in your work, and finally, you will learn how to write AI prompts like a pro, which is arguably the most valuable skill you should have right now. We will learn together how to sign up and what the functionalities of JGPT are, and we will cover the risks and limitations. You will also learn how to prompt, the most important skill in this course, and finally, you'll get an exercise to test your capabilities.\n",
-      "\n",
-      "If you are new to JGPT, go to chat.openai.com and you will arrive at this page. Click \"Sign Up.\" You can create an account using your Google, Microsoft, or Apple credentials, or just provide your email address. After entering your email and password, you will need to validate your email. Click \"Verify Email Address\" in the email you receive, and now your email is verified. Log in again, and OpenAI will ask you for some details, including your birthday. Now you are successfully inside JGPT for the first time.\n",
-      "\n",
-      "Here is the chat box where you will start entering your questions and instructions. For example, you can ask, \"What are the top three skills for an FP&A manager and how to learn them?\" Click here or press Alt + Enter to get your answer. If you want five skills instead of three, click the little pen to correct your instruction and ask for the \"top five skills\" and \"show me the result in a table.\" You can also press the copy and paste button to copy all of the text. However, the text may not be formatted, so you can select the table and copy and paste it.\n",
-      "\n",
-      "If you are not happy with the result, you can click \"Regenerate,\" and JGPT will propose an alternative answer. You can change your question and ask about the top technical skills. You can also come back to all of your previous instructions by checking your history. If you want to start a new chat, click here and ask, \"Tell me how is a typical day for an FP&A manager.\" \n",
-      "\n",
-      "It's important to change discussions each time you change subjects. If you don't, JGPT may misunderstand your question. You can manage your settings in the sidebar, including saving or not saving chat history and activating or deactivating training on your data.\n",
-      "\n",
-      "Now let's look at the most advanced functions of JGPT. You can upload a file by clicking the attachments button. For example, if you have a file with your company sales data, you can upload it and ask JGPT to create a cohort analysis for your retention rate based on the month the subscriber started doing business with you. JGPT will read the columns using Python, which is good for calculations. \n",
-      "\n",
-      "You can ask JGPT to change the color of the visualization or analyze the customer churn rate. You can also audit the code generated by JGPT to ensure the calculations are correct. \n",
-      "\n",
-      "Let me show you how to create your own GPT. Go to the sidebar and click \"Explore GPTs.\" You can search for customized GPT versions or create your own. You can give it a name, describe what it does, and upload files to augment its knowledge. \n",
-      "\n",
-      "Now, let’s create a GPT. Click \"Create\" and use the create view to make it easier. For example, if you want a coach to help you get better at FP&A, you can describe what you want it to do. JGPT will even create a logo for your coach. \n",
-      "\n",
-      "In this lesson, we will explore Canvas, a powerful feature in JGPT that helps finance professionals draft reports and modify code in real time. For example, you can take a screenshot of a P&L statement and ask JGPT to draft commentaries on it. \n",
-      "\n",
-      "You can activate Canvas to see your document being written on one side while the conversation remains active on the other. You can edit the document, change the reading level, and make it shorter or longer. You can also add specific content, such as mentioning that growth comes from AI and blockchain.\n",
-      "\n",
-      "The last important functionality of Canvas is with coding. You can ask JGPT to create code for a scenario analysis, run the code, and debug it if necessary. \n",
-      "\n",
-      "Let me show you the JGPT search functionality. If you want to get online sources, click \"Search the Web.\" For example, if you ask for the price of Nvidia stock from last Friday, JGPT will automatically search the web for you. \n",
-      "\n",
-      "Now, let’s look at the project functionalities in JGPT. If you have a pro or enterprise account, you can create a new project. For example, if you want to work on the budget for 2025, you can upload files and add instructions for the project. \n",
-      "\n",
-      "You can draft emails or ask for specific reports based on the information you uploaded. This allows you to reuse information and organize your chats based on topics. \n",
-      "\n",
-      "Projects are different from custom GPTs. In projects, you can choose the model and give specific instructions, while custom GPTs are more like agents with specific behaviors. \n",
-      "\n",
-      "Now, let’s look at the new functionality called Operator. Operator acts on your behalf and can perform tasks for you. For example, it can book a table for you by interacting with your browser. \n",
-      "\n",
-      "To customize JGPT, click on the diamond icon and provide your preferences, such as tone and traits. You can also activate memory to allow JGPT to remember information about you. \n",
-      "\n",
-      "Let’s start with the core of this course: how to prompt with JGPT. I will teach you a framework called CSI, which stands for Context, Specificity, and Instruction. This framework helps you provide clear expectations for the output you want from JGPT. \n",
-      "\n",
-      "For example, if you want to draft a reminder letter for a client, you need to provide context about your role, specify the problem, and give clear instructions. \n",
-      "\n",
-      "You can also use the FBI framework, which stands for Format, Blueprint, and Identity, to enhance your prompts. This helps you get more impactful outputs. \n",
-      "\n",
-      "Now, let’s practice the CSI and FBI formulas. Imagine you are an FP&A analyst for a brewery and need to create a weekly financial report. Use the CSI formula to generate the report outline, and then improve it using the FBI formula. \n",
-      "\n",
-      "Remember, the risks and limitations of using JGPT include confidentiality issues, security concerns, and the quality of the output. Always review the information and ensure it meets your standards. \n",
-      "\n",
-      "JGPT is not consistently updated, and it may generate incorrect or nonsensical answers. Be specific in your prompts to get the best results. \n",
-      "\n",
-      "This concludes the tutorial on how to use JGPT as a finance professional. If you found this course valuable, consider subscribing for more content.\n",
-      "-----\n",
-      "This free course teaches finance professionals how to effectively use ChatGPT (CHPT) to enhance productivity and adaptability in a rapidly changing financial landscape. Key components include signing up for CHPT, understanding its functionalities, and recognizing its risks and limitations. The course emphasizes the importance of crafting effective AI prompts using the CSI (Context, Specificity, Instruction) and FBI (Format, Blueprint, Identity) frameworks to achieve precise outputs. Users learn to navigate CHPT's interface, utilize advanced features like file uploads for data analysis, and leverage the Canvas tool for drafting reports and modifying code in real-time. The course also covers the importance of data confidentiality, security, and the inherent limitations of AI outputs, such as potential inaccuracies and the need for user oversight. Participants are encouraged to create customized GPTs for specific tasks and projects, enhancing their efficiency in financial analysis and reporting. The course concludes with a call to explore further modules for comprehensive mastery of AI in finance.\n"
+      "-----\n"
      ]
     }
    ],
@@ -131,9 +78,9 @@
     "yt = await extract_content(dict(url=\"https://www.youtube.com/watch?v=lLprprtHfts\"))\n",
     "print(yt)\n",
     "print(\"-----\")\n",
-    "print(await cleanup_content(yt.content))\n",
-    "print(\"-----\")\n",
-    "print(await summarize(yt.content, None))"
+    "# print(await cleanup_content(yt.content))\n",
+    "# print(\"-----\")\n",
+    "# print(await summarize(yt.content, None))"
    ]
   },
   {
@@ -327,6 +274,75 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "ProcessSourceOutput(title='tmp2gizs_hk.09869', source_type='url', identified_type='application/pdf', identified_provider='', metadata={}, content='Docling Technical Report\\nVersion 1.0\\nChristoph Auer\\nMaksym Lysak\\nAhmed Nassar\\nMichele Dolfi\\nNikolaos Livathinos\\nPanos Vagenas\\nCesar Berrospi Ramis\\nMatteo Omenetti\\nFabian Lindlbauer\\nKasper Dinkla\\nLokesh Mishra\\nYusik Kim\\nShubham Gupta\\nRafael Teixeira de Lima\\nValery Weber\\nLucas Morin\\nIngmar Meijer\\nViktor Kuropiatnyk\\nPeter W. J. Staar\\nAI4K Group, IBM Research\\nR ̈uschlikon, Switzerland\\nAbstract\\nThis technical report introduces Docling, an easy to use, self-contained, MITlicensed open-source package for PDF document conversion. It is powered by\\nstate-of-the-art specialized AI models for layout analysis (DocLayNet) and table\\nstructure recognition (TableFormer), and runs efficiently on commodity hardware\\nin a small resource budget. The code interface allows for easy extensibility and\\naddition of new features and models.\\n1\\nIntroduction\\nConverting PDF documents back into a machine-processable format has been a major challenge\\nfor decades due to their huge variability in formats, weak standardization and printing-optimized\\ncharacteristic, which discards most structural features and metadata. With the advent of LLMs\\nand popular application patterns such as retrieval-augmented generation (RAG), leveraging the rich\\ncontent embedded in PDFs has become ever more relevant. In the past decade, several powerful\\ndocument understanding solutions have emerged on the market, most of which are commercial software, cloud offerings [3] and most recently, multi-modal vision-language models. As of today, only\\na handful of open-source tools cover PDF conversion, leaving a significant feature and quality gap\\nto proprietary solutions.\\nWith Docling, we open-source a very capable and efficient document conversion tool which builds\\non the powerful, specialized AI models and datasets for layout analysis and table structure recognition we developed and presented in the recent past [12, 13, 9]. Docling is designed as a simple,\\nself-contained python library with permissive license, running entirely locally on commodity hardware. Its code architecture allows for easy extensibility and addition of new features and models.\\nDocling Technical Report\\n1\\narXiv:2408.09869v5 [cs.CL] 9 Dec 2024\\nHere is what Docling delivers today:\\n* Converts PDF documents to JSON or Markdown format, stable and lightning fast\\n* Understands detailed page layout, reading order, locates figures and recovers table structures\\n* Extracts metadata from the document, such as title, authors, references and language\\n* Optionally applies OCR, e.g. for scanned PDFs\\n* Can be configured to be optimal for batch-mode (i.e high throughput, low time-to-solution)\\nor interactive mode (compromise on efficiency, low time-to-solution)\\n* Can leverage different accelerators (GPU, MPS, etc).\\n2\\nGetting Started\\nTo use Docling, you can simply install the docling package from PyPI. Documentation and examples\\nare available in our GitHub repository at github.com/DS4SD/docling. All required model assets1 are\\ndownloaded to a local huggingface datasets cache on first use, unless you choose to pre-install the\\nmodel assets in advance.\\nDocling provides an easy code interface to convert PDF documents from file system, URLs or binary\\nstreams, and retrieve the output in either JSON or Markdown format. For convenience, separate\\nmethods are offered to convert single documents or batches of documents. A basic usage example\\nis illustrated below. Further examples are available in the Doclign code repository.\\nfrom\\ndocling. document_converter\\nimport\\nDocumentConverter\\nsource = \"https:// arxiv.org/pdf /2206.01062\"\\n# PDF path or URL\\nconverter = DocumentConverter ()\\nresult = converter. convert_single (source)\\nprint(result. render_as_markdown ())\\n# output: \"##\\nDocLayNet: A Large\\nHuman -Annotated\\nDataset\\nfor Document -Layout\\nAnalysis\\n[...]\"\\nOptionally, you can configure custom pipeline features and runtime options, such as turning on or\\noff features (e.g. OCR, table structure recognition), enforcing limits on the input document size, and\\ndefining the budget of CPU threads. Advanced usage examples and options are documented in the\\nREADME file. Docling also provides a Dockerfile to demonstrate how to install and run it inside a\\ncontainer.\\n3\\nProcessing pipeline\\nDocling implements a linear pipeline of operations, which execute sequentially on each given document (see Fig. 1). Each document is first parsed by a PDF backend, which retrieves the programmatic\\ntext tokens, consisting of string content and its coordinates on the page, and also renders a bitmap\\nimage of each page to support downstream operations. Then, the standard model pipeline applies a\\nsequence of AI models independently on every page in the document to extract features and content,\\nsuch as layout and table structures. Finally, the results from all pages are aggregated and passed\\nthrough a post-processing stage, which augments metadata, detects the document language, infers\\nreading-order and eventually assembles a typed document object which can be serialized to JSON\\nor Markdown.\\n3.1\\nPDF backends\\nTwo basic requirements to process PDF documents in our pipeline are a) to retrieve all text content\\nand their geometric coordinates on each page and b) to render the visual representation of each\\npage as it would appear in a PDF viewer. Both these requirements are encapsulated in Docling’s\\nPDF backend interface. While there are several open-source PDF parsing libraries available for\\npython, we faced major obstacles with all of them for different reasons, among which were restrictive\\n1see huggingface.co/ds4sd/docling-models/\\n2\\nLayout\\nAnalysis\\nSerialize as\\nJSON\\nor Markdown\\n{;}\\nParse\\nPDF pages\\nTable\\nStructure\\nOCR\\nModel Pipeline\\nAssemble results,\\nApply document\\npost-processing\\nFigure 1: Sketch of Docling’s default processing pipeline. The inner part of the model pipeline is\\neasily customizable and extensible.\\nlicensing (e.g. pymupdf [7]), poor speed or unrecoverable quality issues, such as merged text cells\\nacross far-apart text tokens or table columns (pypdfium, PyPDF) [15, 14].\\nWe therefore decided to provide multiple backend choices, and additionally open-source a custombuilt PDF parser, which is based on the low-level qpdf[4] library. It is made available in a separate\\npackage named docling-parse and powers the default PDF backend in Docling. As an alternative,\\nwe provide a PDF backend relying on pypdfium, which may be a safe backup choice in certain cases,\\ne.g. if issues are seen with particular font encodings.\\n3.2\\nAI models\\nAs part of Docling, we initially release two highly capable AI models to the open-source community,\\nwhich have been developed and published recently by our team. The first model is a layout analysis\\nmodel, an accurate object-detector for page elements [13]. The second model is TableFormer [12, 9],\\na state-of-the-art table structure recognition model. We provide the pre-trained weights (hosted on\\nhuggingface) and a separate package for the inference code as docling-ibm-models. Both models\\nare also powering the open-access deepsearch-experience, our cloud-native service for knowledge\\nexploration tasks.\\nLayout Analysis Model\\nOur layout analysis model is an object-detector which predicts the bounding-boxes and classes of\\nvarious elements on the image of a given page. Its architecture is derived from RT-DETR [16] and\\nre-trained on DocLayNet [13], our popular human-annotated dataset for document-layout analysis,\\namong other proprietary datasets. For inference, our implementation relies on the onnxruntime [5].\\nThe Docling pipeline feeds page images at 72 dpi resolution, which can be processed on a single\\nCPU with sub-second latency. All predicted bounding-box proposals for document elements are\\npost-processed to remove overlapping proposals based on confidence and size, and then intersected\\nwith the text tokens in the PDF to group them into meaningful and complete units such as paragraphs,\\nsection titles, list items, captions, figures or tables.\\nTable Structure Recognition\\nThe TableFormer model [12], first published in 2022 and since refined with a custom structure token\\nlanguage [9], is a vision-transformer model for table structure recovery. It can predict the logical\\nrow and column structure of a given table based on an input image, and determine which table\\ncells belong to column headers, row headers or the table body. Compared to earlier approaches,\\nTableFormer handles many characteristics of tables, such as partial or no borderlines, empty cells,\\nrows or columns, cell spans and hierarchy both on column-heading or row-heading level, tables with\\ninconsistent indentation or alignment and other complexities. For inference, our implementation\\nrelies on PyTorch [2].\\n3\\nThe Docling pipeline feeds all table objects detected in the layout analysis to the TableFormer model,\\nby providing an image-crop of the table and the included text cells. TableFormer structure predictions are matched back to the PDF cells in post-processing to avoid expensive re-transcription text\\nin the table image. Typical tables require between 2 and 6 seconds to be processed on a standard\\nCPU, strongly depending on the amount of included table cells.\\nOCR\\nDocling provides optional support for OCR, for example to cover scanned PDFs or content in\\nbitmaps images embedded on a page. In our initial release, we rely on EasyOCR [1], a popular thirdparty OCR library with support for many languages. Docling, by default, feeds a high-resolution\\npage image (216 dpi) to the OCR engine, to allow capturing small print detail in decent quality.\\nWhile EasyOCR delivers reasonable transcription quality, we observe that it runs fairly slow on\\nCPU (upwards of 30 seconds per page).\\nWe are actively seeking collaboration from the open-source community to extend Docling with\\nadditional OCR backends and speed improvements.\\n3.3\\nAssembly\\nIn the final pipeline stage, Docling assembles all prediction results produced on each page into a\\nwell-defined datatype that encapsulates a converted document, as defined in the auxiliary package\\ndocling-core. The generated document object is passed through a post-processing model which\\nleverages several algorithms to augment features, such as detection of the document language, correcting the reading order, matching figures with captions and labelling metadata such as title, authors\\nand references. The final output can then be serialized to JSON or transformed into a Markdown\\nrepresentation at the users request.\\n3.4\\nExtensibility\\nDocling provides a straight-forward interface to extend its capabilities, namely the model pipeline.\\nA model pipeline constitutes the central part in the processing, following initial document parsing\\nand preceding output assembly, and can be fully customized by sub-classing from an abstract baseclass (BaseModelPipeline) or cloning the default model pipeline. This effectively allows to fully\\ncustomize the chain of models, add or replace models, and introduce additional pipeline configuration parameters. To use a custom model pipeline, the custom pipeline class to instantiate can\\nbe provided as an argument to the main document conversion methods. We invite everyone in the\\ncommunity to propose additional or alternative models and improvements.\\nImplementations of model classes must satisfy the python Callable interface. The __call__ method\\nmust accept an iterator over page objects, and produce another iterator over the page objects which\\nwere augmented with the additional features predicted by the model, by extending the provided\\nPagePredictions data model accordingly.\\n4\\nPerformance\\nIn this section, we establish some reference numbers for the processing speed of Docling and the\\nresource budget it requires. All tests in this section are run with default options on our standard test\\nset distributed with Docling, which consists of three papers from arXiv and two IBM Redbooks,\\nwith a total of 225 pages. Measurements were taken using both available PDF backends on two\\ndifferent hardware systems: one MacBook Pro M3 Max, and one bare-metal server running Ubuntu\\n20.04 LTS on an Intel Xeon E5-2690 CPU. For reproducibility, we fixed the thread budget (through\\nsetting OMP NUM THREADS environment variable) once to 4 (Docling default) and once to 16\\n(equal to full core count on the test hardware). All results are shown in Table 1.\\nIf you need to run Docling in very low-resource environments, please consider configuring the\\npypdfium backend. While it is faster and more memory efficient than the default docling-parse\\nbackend, it will come at the expense of worse quality results, especially in table structure recovery.\\nEstablishing GPU acceleration support for the AI models is currently work-in-progress and largely\\nuntested, but may work implicitly when CUDA is available and discovered by the onnxruntime and\\n4\\ntorch runtimes backing the Docling pipeline. We will deliver updates on this topic at in a future\\nversion of this report.\\nTable 1: Runtime characteristics of Docling with the standard model pipeline and settings, on our\\ntest dataset of 225 pages, on two different systems. OCR is disabled. We show the time-to-solution\\n(TTS), computed throughput in pages per second, and the peak memory used (resident set size) for\\nboth the Docling-native PDF backend and for the pypdfium backend, using 4 and 16 threads.\\nCPU\\nThread\\nbudget\\nnative backend\\npypdfium backend\\nTTS\\nPages/s\\nMem\\nTTS\\nPages/s\\nMem\\nApple M3 Max\\n(16 cores)\\n4\\n177 s\\n1.27\\n6.20 GB\\n103 s\\n2.18\\n2.56 GB\\n16\\n167 s\\n1.34\\n92 s\\n2.45\\nIntel(R) Xeon\\nE5-2690\\n(16 cores)\\n4\\n375 s\\n0.60\\n6.16 GB\\n239 s\\n0.94\\n2.42 GB\\n16\\n244 s\\n0.92\\n143 s\\n1.57\\n5\\nApplications\\nThanks to the high-quality, richly structured document conversion achieved by Docling, its output qualifies for numerous downstream applications. For example, Docling can provide a base\\nfor detailed enterprise document search, passage retrieval or classification use-cases, or support\\nknowledge extraction pipelines, allowing specific treatment of different structures in the document,\\nsuch as tables, figures, section structure or references. For popular generative AI application patterns, such as retrieval-augmented generation (RAG), we provide quackling, an open-source package\\nwhich capitalizes on Docling’s feature-rich document output to enable document-native optimized\\nvector embedding and chunking. It plugs in seamlessly with LLM frameworks such as LlamaIndex [8]. Since Docling is fast, stable and cheap to run, it also makes for an excellent choice to build\\ndocument-derived datasets. With its powerful table structure recognition, it provides significant benefit to automated knowledge-base construction [11, 10]. Docling is also integrated within the open\\nIBM data prep kit [6], which implements scalable data transforms to build large-scale multi-modal\\ntraining datasets.\\n6\\nFuture work and contributions\\nDocling is designed to allow easy extension of the model library and pipelines. In the future, we\\nplan to extend Docling with several more models, such as a figure-classifier model, an equationrecognition model, a code-recognition model and more. This will help improve the quality of conversion for specific types of content, as well as augment extracted document metadata with additional information. Further investment into testing and optimizing GPU acceleration as well as\\nimproving the Docling-native PDF backend are on our roadmap, too.\\nWe encourage everyone to propose or implement additional features and models, and will\\ngladly take your inputs and contributions under review. The codebase of Docling is open for use\\nand contribution, under the MIT license agreement and in alignment with our contributing guidelines\\nincluded in the Docling repository. If you use Docling in your projects, please consider citing this\\ntechnical report.\\nReferences\\n[1] J. AI. Easyocr: Ready-to-use ocr with 80+ supported languages. https://github.com/\\nJaidedAI/EasyOCR, 2024. Version: 1.7.0.\\n[2] J. Ansel, E. Yang, H. He, N. Gimelshein, A. Jain, M. Voznesensky, B. Bao, P. Bell, D. Berard,\\nE. Burovski, G. Chauhan, A. Chourdia, W. Constable, A. Desmaison, Z. DeVito, E. Ellison,\\nW. Feng, J. Gong, M. Gschwind, B. Hirsh, S. Huang, K. Kalambarkar, L. Kirsch, M. Lazos, M. Lezcano, Y. Liang, J. Liang, Y. Lu, C. Luk, B. Maher, Y. Pan, C. Puhrsch, M. Reso,\\nM. Saroufim, M. Y. Siraichi, H. Suk, M. Suo, P. Tillet, E. Wang, X. Wang, W. Wen, S. Zhang,\\nX. Zhao, K. Zhou, R. Zou, A. Mathews, G. Chanan, P. Wu, and S. Chintala. Pytorch 2: Faster\\n5\\nmachine learning through dynamic python bytecode transformation and graph compilation.\\nIn Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2 (ASPLOS ’24). ACM, 4 2024. doi:\\n10.1145/3620665.3640366. URL https://pytorch.org/assets/pytorch2-2.pdf.\\n[3] C. Auer, M. Dolfi, A. Carvalho, C. B. Ramis, and P. W. Staar. Delivering document conversion\\nas a cloud service with high throughput and responsiveness. In 2022 IEEE 15th International\\nConference on Cloud Computing (CLOUD), pages 363-373. IEEE, 2022.\\n[4] J. Berkenbilt. Qpdf: A content-preserving pdf document transformer, 2024. URL https:\\n//github.com/qpdf/qpdf.\\n[5] O. R. developers. Onnx runtime. https://onnxruntime.ai/, 2024. Version: 1.18.1.\\n[6] IBM.\\nData Prep Kit: a community project to democratize and accelerate unstructured\\ndata preparation for LLM app developers, 2024.\\nURL https://github.com/IBM/\\ndata-prep-kit.\\n[7] A. S. Inc. PyMuPDF, 2024. URL https://github.com/pymupdf/PyMuPDF.\\n[8] J. Liu. LlamaIndex, 11 2022. URL https://github.com/jerryjliu/llama_index.\\n[9] M. Lysak, A. Nassar, N. Livathinos, C. Auer, and P. Staar. Optimized Table Tokenization\\nfor Table Structure Recognition. In Document Analysis and Recognition - ICDAR 2023: 17th\\nInternational Conference, San Jos ́e, CA, USA, August 21-26, 2023, Proceedings, Part II, pages\\n37-50, Berlin, Heidelberg, Aug. 2023. Springer-Verlag. ISBN 978-3-031-41678-1. doi: 10.\\n1007/978-3-031-41679-8 3. URL https://doi.org/10.1007/978-3-031-41679-8_3.\\n[10] L. Mishra, S. Dhibi, Y. Kim, C. Berrospi Ramis, S. Gupta, M. Dolfi, and P. Staar. Statements: Universal information extraction from tables with large language models for ESG\\nKPIs. In D. Stammbach, J. Ni, T. Schimanski, K. Dutia, A. Singh, J. Bingler, C. Christiaen, N. Kushwaha, V. Muccione, S. A. Vaghefi, and M. Leippold, editors, Proceedings of the\\n1st Workshop on Natural Language Processing Meets Climate Change (ClimateNLP 2024),\\npages 193-214, Bangkok, Thailand, Aug. 2024. Association for Computational Linguistics.\\nURL https://aclanthology.org/2024.climatenlp-1.15.\\n[11] L. Morin, V. Weber, G. I. Meijer, F. Yu, and P. W. J. Staar. Patcid: an open-access dataset of\\nchemical structures in patent documents. Nature Communications, 15(1):6532, August 2024.\\nISSN 2041-1723. doi: 10.1038/s41467-024-50779-y. URL https://doi.org/10.1038/\\ns41467-024-50779-y.\\n[12] A. Nassar, N. Livathinos, M. Lysak, and P. Staar. Tableformer: Table structure understanding\\nwith transformers.\\nIn Proceedings of the IEEE/CVF Conference on Computer Vision and\\nPattern Recognition, pages 4614-4623, 2022.\\n[13] B. Pfitzmann, C. Auer, M. Dolfi, A. S. Nassar, and P. Staar.\\nDoclaynet: a large humanannotated dataset for document-layout segmentation. pages 3743-3751, 2022.\\n[14] pypdf Maintainers. pypdf: A Pure-Python PDF Library, 2024. URL https://github.com/\\npy-pdf/pypdf.\\n[15] P. Team. PyPDFium2: Python bindings for PDFium, 2024. URL https://github.com/\\npypdfium2-team/pypdfium2.\\n[16] Y. Zhao, W. Lv, S. Xu, J. Wei, G. Wang, Q. Dang, Y. Liu, and J. Chen. Detrs beat yolos on\\nreal-time object detection, 2023.\\n6\\nAppendix\\nIn this section, we illustrate a few examples of Docling’s output in Markdown and JSON.\\nBirgit Pfitzmann IBM Research Rueschlikon, Switzerland bpf@zurich.ibm.com\\nChristoph Auer IBM Research Rueschlikon, Switzerland cau@zurich.ibm.com\\nMichele Dolfi IBM Research Rueschlikon, Switzerland dol@zurich.ibm.com\\nAhmed S. Nassar IBM Research Rueschlikon, Switzerland ahn@zurich.ibm.com\\nPeter Staar IBM Research Rueschlikon, Switzerland taa@zurich.ibm.com\\nAccurate document layout analysis is a key requirement for highquality PDF document conversion. With the recent availability of public, large groundtruth datasets such as PubLayNet and DocBank, deep-learning models have proven to be very effective at layout detection and segmentation. While\\nthese datasets are of adequate size to train such models, they severely lack in layout variability since they are sourced from scientific article\\nrepositories such as PubMed and arXiv only. Consequently, the accuracy of the layout segmentation drops significantly when these models are\\napplied on more challenging and diverse layouts. In this paper, we present DocLayNet, a new, publicly available, document-layout annotation dataset\\nin COCO format. It contains 80863 manually annotated pages from diverse data sources to represent a wide variability in layouts. For each PDF\\npage, the layout annotations provide labelled bounding-boxes with a choice of 11 distinct classes. DocLayNet also provides a subset of double- and\\ntriple-annotated pages to determine the inter-annotator agreement. In multiple experiments, we provide baseline accuracy scores (in mAP) for a set\\nof popular object detection models. We also demonstrate that these models fall approximately 10% behind the inter-annotator agreement.\\nFurthermore, we provide evidence that DocLayNet is of sufficient size. Lastly, we compare models trained on PubLayNet, DocBank and DocLayNet,\\nshowing that layout predictions of the DocLayNettrained models are more robust and thus the preferred choice for general-purpose document-layout\\nanalysis.\\n· Information systems → Document structure; · Applied computing → Document analysis; · Computing methodologies → Machine learning;\\nComputer vision; Object detection;\\nPermission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not\\nmade or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party\\ncomponents of this work must be honored. For all other uses, contact the owner/author(s).\\nKDD \\'22, August 14-18, 2022, Washington, DC, USA (c) 2022 Copyright held by the owner/author(s). ACM ISBN 978-1-4503-9385-0/22/08.\\nhttps://doi.org/10.1145/3534678.3539043\\nFigure 1: Four examples of complex page layouts across different document categories\\nPDF document conversion, layout segmentation, object-detection, data set, Machine Learning\\nBirgit Pfitzmann, Christoph Auer, Michele Dolfi, Ahmed S. Nassar, and Peter Staar. 2022. DocLayNet: A Large Human-Annotated Dataset for\\nDocumentLayout Analysis. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD \\'22), August 14-18,\\n2022, Washington, DC, USA. ACM, New York, NY, USA, 9 pages. https://doi.org/10.1145/ 3534678.3539043\\nDocLayNet: A Large Human-Annotated Dataset for Document-Layout Analysis\\nABSTRACT\\nCCS CONCEPTS\\nKEYWORDS\\nACM Reference Format:\\nDocLayNet: A Large Human-Annotated Dataset for\\nDocument-Layout Analysis\\nBirgit Ptzmann\\nIBM Research\\nRueschlikon, Switzerland\\nbpf@zurich.ibm.com\\nChristoph Auer\\nIBM Research\\nRueschlikon, Switzerland\\ncau@zurich.ibm.com\\nMichele Dol\\nIBM Research\\nRueschlikon, Switzerland\\ndol@zurich.ibm.com\\nAhmed S. Nassar\\nIBM Research\\nRueschlikon, Switzerland\\nahn@zurich.ibm.com\\nPeter Staar\\nIBM Research\\nRueschlikon, Switzerland\\ntaa@zurich.ibm.com\\nABSTRACT\\nAccurate document layout analysis is a key requirement for highquality PDF document conversion. With the recent availability of\\npublic, large ground-truth datasets such as PubLayNet and DocBank,\\ndeep-learning models have proven to be very eective at layout\\ndetection and segmentation. While these datasets are of adequate\\nsize to train such models, they severely lack in layout variability\\nsince they are sourced from scientic article repositories such as\\nPubMed and arXiv only. Consequently, the accuracy of the layout\\nsegmentation drops signicantly when these models are applied\\non more challenging and diverse layouts. In this paper, we present\\nDocLayNet, a new, publicly available, document-layout annotation\\ndataset in COCO format. It contains 80863 manually annotated\\npages from diverse data sources to represent a wide variability in\\nlayouts. For each PDF page, the layout annotations provide labelled\\nbounding-boxes with a choice of 11 distinct classes. DocLayNet\\nalso provides a subset of double- and triple-annotated pages to\\ndetermine the inter-annotator agreement. In multiple experiments,\\nwe provide baseline accuracy scores (in mAP) for a set of popular\\nobject detection models. We also demonstrate that these models\\nfall approximately 10% behind the inter-annotator agreement. Furthermore, we provide evidence that DocLayNet is of sucient size.\\nLastly, we compare models trained on PubLayNet, DocBank and\\nDocLayNet, showing that layout predictions of the DocLayNettrained models are more robust and thus the preferred choice for\\ngeneral-purpose document-layout analysis.\\nCCS CONCEPTS\\n* Information systems →Document structure; * Applied computing →Document analysis; * Computing methodologies\\n→Machine learning; Computer vision; Object detection;\\nPermission to make digital or hard copies of part or all of this work for personal or\\nclassroom use is granted without fee provided that copies are not made or distributed\\nfor prot or commercial advantage and that copies bear this notice and the full citation\\non the rst page. Copyrights for third-party components of this work must be honored.\\nFor all other uses, contact the owner/author(s).\\nKDD ’22, August 14-18, 2022, Washington, DC, USA\\n(c) 2022 Copyright held by the owner/author(s).\\nACM ISBN 978-1-4503-9385-0/22/08.\\nhttps://doi.org/10.1145/3534678.3539043\\n13\\nUSING THE VERTICAL TUBE -\\nMODELS AY11230/11234\\n1. The vertical tube can be used for\\n\\ninstructional viewing or to photograph\\n\\nthe image with a digital camera or a\\n\\nmicro TV unit\\n\\n2. Loosen the retention screw, then rotate\\n\\nthe adjustment ring to change the\\n\\nlength of the vertical tube.\\n3. Make sure that both the images in\\n\\nOPERATION (cont.)\\nSELECTING OBJECTIVE\\nMAGNIFICATION\\n\\n1. There are two objectives. The lower\\nmagnification objective has a greater\\ndepth of field and view.\\n2. In order to observe the specimen\\neasily use the lower magnification\\nobjective first. Then, by rotating the\\ncase, the magnification can be\\nchanged.\\nCHANGING THE INTERPUPILLARY\\nDISTANCE\\n1. The distance between the observer\\'s\\npupils is the interpupillary distance.\\n2. To adjust the interpupillary distance\\nrotate the prism caps until both eyes\\ncoincide with the image in the\\neyepiece.\\n\\nFOCUSING\\n1. Remove the lens protective cover.\\n2. Place the specimen on the working\\nstage.\\n3. Focus the specimen with the left eye\\nfirst while turning the focus knob until\\nthe image appears clear and sharp.\\n4. Rotate the right eyepiece ring until the\\nimages in each eyepiece coincide and\\nare sharp and clear.\\nCHANGING THE BULB\\n1. Disconnect the power cord.\\n2. When the bulb is cool, remove the\\noblique illuminator cap and remove\\nthe halogen bulb with cap.\\n3. Replace with a new halogen bulb.\\n4. Open the window in the base plate and\\nreplace the halogen lamp or\\n\\nfluorescent lamp of transmitted\\nilluminator.\\nFOCUSING\\n1. Turn the focusing knob away or toward\\nyou until a clear image is viewed.\\n2. If the image is unclear, adjust the\\nheight of the elevator up or down,\\nthen turn the focusing knob again.\\nZOOM MAGNIFICATION\\n1. Turn the zoom magnification knob to\\nthe desired magnification and field of\\nview.\\n2. In most situations, it is recommended\\nthat you focus at the lowest\\n\\nmagnification, then move to a higher\\nmagnification and re-focus as\\n\\nnecessary.\\n3. If the image is not clear to both eyes\\nat the same time, the diopter ring may\\nneed adjustment.\\nDIOPTER RING ADJUSTMENT\\n1. To adjust the eyepiece for viewing with\\nor without eyeglasses and for\\n\\ndifferences in acuity between the right\\nand left eyes, follow the following\\nsteps:\\na. Observe an image through the left\\neyepiece and bring a specific point\\ninto focus using the focus knob.\\nb. By turning the diopter ring\\n\\nadjustment for the left eyepiece,\\nbring the same point into sharp\\nfocus.\\nc.Then bring the same point into\\nfocus through the right eyepiece\\nby turning the right diopter ring.\\nd.With more than one viewer, each\\nviewer should note their own\\ndiopter ring position for the left\\nand right eyepieces, then before\\nviewing set the diopter ring\\nadjustments to that setting.\\nCHANGING THE BULB\\n1. Disconnect the power cord from the\\nelectrical outlet.\\n2. When the bulb is cool, remove the\\noblique illuminator cap and remove\\nthe halogen bulb with cap.\\n3. Replace with a new halogen bulb.\\n4. Open the window in the base plate\\nand replace the halogen lamp or\\nfluorescent lamp of transmitted\\nilluminator.\\n\\nModel AY11230\\nModel AY11234\\n14\\nObjectives\\nRevolving Turret\\nCoarse\\nAdjustment\\nKnob\\nMODEL AY11236\\nMICROSCOPE USAGE\\nBARSKA Model AY11236 is a powerful fixed power compound\\nmicroscope designed for biological studies such as specimen\\nexamination. It can also be used for examining bacteria and\\nfor general clinical and medical studies and other scientific uses.\\nCONSTRUCTION\\nBARSKA Model AY11236 is a fixed power compound microscope.\\nIt is constructed with two optical paths at the same angle. It is\\nequipped with transmitted illumination. By using this instrument,\\nthe user can observe specimens at magnification from 40x to\\n1000x by selecting the desired objective lens. Coarse and fine\\nfocus adjustments provide accuracy and image detail. The rotating\\nhead allows the user to position the eyepieces for maximum\\nviewing comfort and easy access to all adjustment knobs.\\nModel AY11236\\nFine\\nAdjustment\\nKnob\\nStage\\nCondenser\\nFocusing\\nKnob\\nEyepiece\\nStand\\nLamp\\nOn/Off\\nSwitch\\nLamp\\nPower\\nCord\\nRotating Head\\nStage Clip\\nAdjustment\\nInterpupillary Slide Adjustment\\nCircling Minimums\\n7KHUHZDVDFKDQJHWRWKH7(536FULWHULDLQWKDWDႇHFWVFLUFOLQJDUHDGLPHQVLRQE\\\\H[SDQGLQJWKHDUHDVWRSURYLGH\\nimproved obstacle protection. To indicate that the new criteria had been applied to a given procedure, a\\nis placed on\\nthe circling line of minimums. The new circling tables and explanatory information is located in the Legend of the TPP.\\n7KHDSSURDFKHVXVLQJVWDQGDUGFLUFOLQJDSSURDFKDUHDVFDQEHLGHQWL¿HGE\\\\WKHDEVHQFHRIWKH\\non the circling line of\\nminima.\\n$SSO\\\\6WDQGDUG&LUFOLQJ$SSURDFK0DQHXYHULQJ5DGLXV7DEOH\\n$SSO\\\\([SDQGHG&LUFOLQJ$SSURDFK0DQHXYHULQJ$LUVSDFH5DGLXV\\nTable\\nAIRPORT SKETCH\\n\\nThe airport sketch is a depiction of the airport with emphasis on runway pattern and related\\ninformation, positioned in either the lower left or lower right corner of the chart to aid pilot recognition of the airport from the air and to provide some information to aid on ground\\nnavigation of the airport. The runways are drawn to scale and oriented to true north. Runway\\ndimensions (length and width) are shown for all active runways.\\nRunway(s) are depicted based on what type and construction of the runway.\\nHard Surface\\nOther Than\\nHard Surface\\nMetal Surface\\nClosed Runway\\nUnder Construction\\nStopways,\\nTaxiways, Parking Areas\\nDisplaced\\nThreshold\\nClosed\\nPavement\\nWater Runway\\nTaxiways and aprons are shaded grey. Other runway features that may be shown are runway numbers, runway dimensions, runway slope, arresting gear, and displaced threshold.\\n2WKHULQIRUPDWLRQFRQFHUQLQJOLJKWLQJ¿QDODSSURDFKEHDULQJVDLUSRUWEHDFRQREVWDFOHVFRQWUROWRZHU1$9$,\\'VKHOLpads may also be shown.\\n$LUSRUW(OHYDWLRQDQG7RXFKGRZQ=RQH(OHYDWLRQ\\nThe airport elevation is shown enclosed within a box in the upper left corner of the sketch box and the touchdown zone\\nelevation (TDZE) is shown in the upper right corner of the sketch box. The airport elevation is the highest point of an\\nDLUSRUW¶VXVDEOHUXQZD\\\\VPHDVXUHGLQIHHWIURPPHDQVHDOHYHO7KH7\\'=(LVWKHKLJKHVWHOHYDWLRQLQWKH¿UVWIHHWRI\\nthe landing surface. Circling only approaches will not show a TDZE.\\n114\\nFAA Chart Users’ Guide - Terminal Procedures Publication (TPP) - Terms\\nAGL 2013 Financial Calendar\\n22 August 2012\\n2012 full year result and final dividend announced\\n30 August 2012\\nEx-dividend trading commences\\n5 September 2012\\nRecord date for 2012 final dividend\\n27 September 2012\\nFinal dividend payable\\n23 October 2012\\nAnnual General Meeting\\n27 February 20131\\n2013 interim result and interim dividend announced\\n28 August 20131\\n2013 full year results and final dividend announced\\n1 Indicative dates only, subject to change/Board confirmation\\nAGL’s Annual General Meeting will be held at the City Recital Hall, Angel Place, Sydney\\ncommencing at 10.30am on Tuesday 23 October 2012.\\nYesterday\\nEstablished in Sydney in 1837, and then\\nknown as The Australian Gas Light Company,\\nthe AGL business has an established history\\nand reputation for serving the gas and\\nelectricity needs of Australian households.\\nIn 1841, when AGL supplied the gas to light\\nthe first public street lamp, it was reported\\nin the Sydney Gazette as a “wonderful\\nachievement of scientific knowledge, assisted\\nby mechanical ingenuity.” Within two years,\\n165 gas lamps were lighting the City of Sydney.\\nLooking back on\\n175 years of\\nlooking forward.\\nAGL Energy Limited ABN 74 115 061 375\\n29\\nsigns, signals and road markings\\n3\\nIn chapter 2, you and your vehicle, you learned about\\nsome of the controls in your vehicle. This chapter is a handy\\nreference section that gives examples of the most common\\nsigns, signals and road markings that keep traffic organized\\nand flowing smoothly.\\nSigns\\nThere are three ways to read signs: by their shape, colour and\\nthe messages printed on them. Understanding these three ways\\nof classifying signs will help you figure out the meaning of signs\\nthat are new to you.\\nStop\\nYield the right-of-way\\nShows driving\\nregulations\\nExplains lane use\\nSchool zone signs\\nare fluorescent\\nyellow-green\\nTells about motorist\\nservices\\nShows a permitted\\naction\\nShows an action that\\nis not permitted\\nWarns of hazards\\nahead\\nWarns of\\nconstruction zones\\nRailway crossing\\nShows distance and\\ndirection\\n* Signs\\n- regulatory signs\\n- school,\\nplayground and\\ncrosswalk signs\\n- lane use signs\\n- turn control signs\\n- parking signs\\n- reserved lane\\nsigns\\n- warning signs\\n- object markers\\n- construction\\nsigns\\n- information and\\ndestination signs\\n- railway signs\\n* Signals\\n- lane control\\nsignals\\n- traffic lights\\n* Road markings\\n- yellow lines\\n- white lines\\n- reserved lane\\nmarkings\\n- other markings\\nin this chapter\\nFigure 1: Four examples of complex page layouts across different document categories\\nKEYWORDS\\nPDF document conversion, layout segmentation, object-detection,\\ndata set, Machine Learning\\nACM Reference Format:\\nBirgit Ptzmann, Christoph Auer, Michele Dol, Ahmed S. Nassar, and Peter\\nStaar. 2022. DocLayNet: A Large Human-Annotated Dataset for DocumentLayout Analysis. In Proceedings of the 28th ACM SIGKDD Conference on\\nKnowledge Discovery and Data Mining (KDD ’22), August 14-18, 2022, Washington, DC, USA. ACM, New York, NY, USA, 9 pages. https://doi.org/10.1145/\\n3534678.3539043\\narXiv:2206.01062v1 [cs.CV] 2 Jun 2022\\nFigure 2: Title page of the DocLayNet paper (arxiv.org/pdf/2206.01062) - left PDF, right rendered\\nMarkdown. If recognized, metadata such as authors are appearing first under the title. Text content\\ninside figures is currently dropped, the caption is retained and linked to the figure in the JSON\\nrepresentation (not shown).\\n7\\nKDD ’22, August 14-18, 2022, Washington, DC, USA Birgit Pfitzmann, Christoph Auer, Michele Dolfi, Ahmed S. Nassar, and Peter Staar\\nTable 2: Prediction performance (mAP@0.5-0.95) of object\\ndetection networks on DocLayNet test set. The MRCNN\\n(Mask R-CNN) and FRCNN (Faster R-CNN) models with\\nResNet-50 or ResNet-101 backbone were trained based on\\nthe network architectures from the detectron2 model zoo\\n(Mask R-CNN R50, R101-FPN 3x, Faster R-CNN R101-FPN\\n3x), with default congurations. The YOLO implementation\\nutilized was YOLOv5x6 [13]. All models were initialised using pre-trained weights from the COCO 2017 dataset.\\nhuman\\nMRCNN\\nFRCNN\\nYOLO\\nR50\\nR101\\nR101\\nv5x6\\nCaption\\n84-89\\n68.4\\n71.5\\n70.1\\n77.7\\nFootnote\\n83-91\\n70.9\\n71.8\\n73.7\\n77.2\\nFormula\\n83-85\\n60.1\\n63.4\\n63.5\\n66.2\\nList-item\\n87-88\\n81.2\\n80.8\\n81.0\\n86.2\\nPage-footer\\n93-94\\n61.6\\n59.3\\n58.9\\n61.1\\nPage-header\\n85-89\\n71.9\\n70.0\\n72.0\\n67.9\\nPicture\\n69-71\\n71.7\\n72.7\\n72.0\\n77.1\\nSection-header\\n83-84\\n67.6\\n69.3\\n68.4\\n74.6\\nTable\\n77-81\\n82.2\\n82.9\\n82.2\\n86.3\\nText\\n84-86\\n84.6\\n85.8\\n85.4\\n88.1\\nTitle\\n60-72\\n76.7\\n80.4\\n79.9\\n82.7\\nAll\\n82-83\\n72.4\\n73.5\\n73.4\\n76.8\\nto avoid this at any cost in order to have clear, unbiased baseline\\nnumbers for human document-layout annotation. Third, we introduced the feature of snapping boxes around text segments to\\nobtain a pixel-accurate annotation and again reduce time and eort.\\nThe CCS annotation tool automatically shrinks every user-drawn\\nbox to the minimum bounding-box around the enclosed text-cells\\nfor all purely text-based segments, which excludes only Table and\\nPicture. For the latter, we instructed annotation stato minimise\\ninclusion of surrounding whitespace while including all graphical\\nlines. A downside of snapping boxes to enclosed text cells is that\\nsome wrongly parsed PDF pages cannot be annotated correctly and\\nneed to be skipped. Fourth, we established a way to ag pages as\\nrejected for cases where no valid annotation according to the label\\nguidelines could be achieved. Example cases for this would be PDF\\npages that render incorrectly or contain layouts that are impossible\\nto capture with non-overlapping rectangles. Such rejected pages are\\nnot contained in the nal dataset. With all these measures in place,\\nexperienced annotation stamanaged to annotate a single page in\\na typical timeframe of 20s to 60s, depending on its complexity.\\n5\\nEXPERIMENTS\\nThe primary goal of DocLayNet is to obtain high-quality ML models\\ncapable of accurate document-layout analysis on a wide variety\\nof challenging layouts. As discussed in Section 2, object detection\\nmodels are currently the easiest to use, due to the standardisation\\nof ground-truth data in COCO format [16] and the availability of\\ngeneral frameworks such as detectron2 [17]. Furthermore, baseline\\nnumbers in PubLayNet and DocBank were obtained using standard\\nobject detection models such as Mask R-CNN and Faster R-CNN.\\nAs such, we will relate to these object detection methods in this\\nFigure 5: Prediction performance (mAP@0.5-0.95) of a Mask\\nR-CNN network with ResNet50 backbone trained on increasing fractions of the DocLayNet dataset. The learning curve\\nattens around the 80% mark, indicating that increasing the\\nsize of the DocLayNet dataset with similar data will not yield\\nsignicantly better predictions.\\npaper and leave the detailed evaluation of more recent methods\\nmentioned in Section 2 for future work.\\nIn this section, we will present several aspects related to the\\nperformance of object detection models on DocLayNet. Similarly\\nas in PubLayNet, we will evaluate the quality of their predictions\\nusing mean average precision (mAP) with 10 overlaps that range\\nfrom 0.5 to 0.95 in steps of 0.05 (mAP@0.5-0.95). These scores are\\ncomputed by leveraging the evaluation code provided by the COCO\\nAPI [16].\\nBaselines for Object Detection\\nIn Table 2, we present baseline experiments (given in mAP) on Mask\\nR-CNN [12], Faster R-CNN [11], and YOLOv5 [13]. Both training\\nand evaluation were performed on RGB images with dimensions of\\n1025⇥1025 pixels. For training, we only used one annotation in case\\nof redundantly annotated pages. As one can observe, the variation\\nin mAP between the models is rather low, but overall between 6\\nand 10% lower than the mAP computed from the pairwise human\\nannotations on triple-annotated pages. This gives a good indication\\nthat the DocLayNet dataset poses a worthwhile challenge for the\\nresearch community to close the gap between human recognition\\nand ML approaches. It is interesting to see that Mask R-CNN and\\nFaster R-CNN produce very comparable mAP scores, indicating\\nthat pixel-based image segmentation derived from bounding-boxes\\ndoes not help to obtain better predictions. On the other hand, the\\nmore recent Yolov5x model does very well and even out-performs\\nhumans on selected labels such as Text, Table and Picture. This is\\nnot entirely surprising, as Text, Table and Picture are abundant and\\nthe most visually distinctive in a document.\\nFigure 3: Page 6 of the DocLayNet paper. If recognized, metadata such as authors are appearing\\nfirst under the title. Elements recognized as page headers or footers are suppressed in Markdown to\\ndeliver uninterrupted content in reading order. Tables are inserted in reading order. The paragraph\\nin ”5. Experiments” wrapping over the column end is broken up in two and interrupted by the table.\\n8\\n% of\\nTotal\\n% of\\nTotal\\n% of\\nTotal\\ntriple\\ninterannotator\\nmAP @\\n0.5-0.95\\n(%)\\ntriple\\ninterannotator\\nmAP @\\n0.5-0.95\\n(%)\\ntriple\\ninterannotator\\nmAP @\\n0.5-0.95\\n(%)\\ntriple\\ninterannotator\\nmAP @\\n0.5-0.95\\n(%)\\ntriple\\ninterannotator\\nmAP @\\n0.5-0.95\\n(%)\\ntriple\\ninterannotator\\nmAP @\\n0.5-0.95\\n(%)\\ntriple\\ninterannotator\\nmAP @\\n0.5-0.95\\n(%)\\nclass\\nlabel\\nCount\\nTrain\\nTest\\nVal\\nAll\\nFin\\nMan\\nSci\\nLaw\\nPat\\nTen\\nCaption\\n22524\\n2.04\\n1.77\\n2.32\\n84-89\\n40-61\\n86-92\\n94-99\\n95-99\\n69-78\\nn/a\\nFootnote\\n6318\\n0.60\\n0.31\\n0.58\\n83-91\\nn/a\\n100\\n62-88\\n85-94\\nn/a\\n82-97\\nFormula\\n25027\\n2.25\\n1.90\\n2.96\\n83-85\\nn/a\\nn/a\\n84-87\\n86-96\\nn/a\\nn/a\\nList-item\\n185660\\n17.19\\n13.34\\n15.82\\n87-88\\n74-83\\n90-92\\n97-97\\n81-85\\n75-88\\n93-95\\nPagefooter\\n70878\\n6.51\\n5.58\\n6.00\\n93-94\\n88-90\\n95-96\\n100\\n92-97\\n100\\n96-98\\nPageheader\\n58022\\n5.10\\n6.70\\n5.06\\n85-89\\n66-76\\n90-94\\n98-100\\n91-92\\n97-99\\n81-86\\nPicture\\n45976\\n4.21\\n2.78\\n5.31\\n69-71\\n56-59\\n82-86\\n69-82\\n80-95\\n66-71\\n59-76\\nSectionheader\\n142884\\n12.60\\n15.77\\n12.85\\n83-84\\n76-81\\n90-92\\n94-95\\n87-94\\n69-73\\n78-86\\nTable\\n34733\\n3.20\\n2.27\\n3.60\\n77-81\\n75-80\\n83-86\\n98-99\\n58-80\\n79-84\\n70-85\\nText\\n510377\\n45.82\\n49.28\\n45.00\\n84-86\\n81-86\\n88-93\\n89-93\\n87-92\\n71-79\\n87-95\\nTitle\\n5071\\n0.47\\n0.30\\n0.50\\n60-72\\n24-63\\n50-63\\n94-100\\n82-96\\n68-79\\n24-56\\nTotal\\n1107470\\n941123\\n99816\\n66531\\n82-83\\n71-74\\n79-81\\n89-94\\n86-91\\n71-76\\n68-85\\n% of Total\\ntriple inter-annotator mAP @ 0.5-0.95 (%)\\nclass label\\nCount\\nTrain\\nTest\\nVal\\nAll\\nFin\\nMan\\nSci\\nLaw\\nPat\\nTen\\nCaption\\n22524\\n2.04\\n1.77\\n2.32\\n84-89\\n40-61\\n86-92\\n94-99\\n95-99\\n69-78\\nn/a\\nFootnote\\n6318\\n0.60\\n0.31\\n0.58\\n83-91\\nn/a\\n100\\n62-88\\n85-94\\nn/a\\n82-97\\nFormula\\n25027\\n2.25\\n1.90\\n2.96\\n83-85\\nn/a\\nn/a\\n84-87\\n86-96\\nn/a\\nn/a\\nList-item\\n185660\\n17.19\\n13.34\\n15.82\\n87-88\\n74-83\\n90-92\\n97-97\\n81-85\\n75-88\\n93-95\\nPage-footer\\n70878\\n6.51\\n5.58\\n6.00\\n93-94\\n88-90\\n95-96\\n100\\n92-97\\n100\\n96-98\\nPage-header\\n58022\\n5.10\\n6.70\\n5.06\\n85-89\\n66-76\\n90-94\\n98-100\\n91-92\\n97-99\\n81-86\\nPicture\\n45976\\n4.21\\n2.78\\n5.31\\n69-71\\n56-59\\n82-86\\n69-82\\n80-95\\n66-71\\n59-76\\nSection-header\\n142884\\n12.60\\n15.77\\n12.85\\n83-84\\n76-81\\n90-92\\n94-95\\n87-94\\n69-73\\n78-86\\nTable\\n34733\\n3.20\\n2.27\\n3.60\\n77-81\\n75-80\\n83-86\\n98-99\\n58-80\\n79-84\\n70-85\\nText\\n510377\\n45.82\\n49.28\\n45.00\\n84-86\\n81-86\\n88-93\\n89-93\\n87-92\\n71-79\\n87-95\\nTitle\\n5071\\n0.47\\n0.30\\n0.50\\n60-72\\n24-63\\n50-63\\n94-100\\n82-96\\n68-79\\n24-56\\nTotal\\n1107470\\n941123\\n99816\\n66531\\n82-83\\n71-74\\n79-81\\n89-94\\n86-91\\n71-76\\n68-85\\nA\\nB\\nC\\nFigure 4: Table 1 from the DocLayNet paper in the original PDF (A), as rendered Markdown (B)\\nand in JSON representation (C). Spanning table cells, such as the multi-column header ”triple interannotator mAP@0.5-0.95 (%)”, is repeated for each column in the Markdown representation (B),\\nwhich guarantees that every data point can be traced back to row and column headings only by its\\ngrid coordinates in the table. In the JSON representation, the span information is reflected in the\\nfields of each table cell (C).\\n9')"
+      ]
+     },
+     "execution_count": 1,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "from content_core.content.extraction import extract_content\n",
+    "await extract_content(dict(url=\"https://arxiv.org/pdf/2408.09869\"))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Docling"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "## 3-Day Itinerary for Buenos Aires\n",
+      "\n",
+      "## Day 1: Exploring the Heart of the City\n",
+      "\n",
+      "## Morning:\n",
+      "\n",
+      "- - Breakfast at Café Tortoni (8:00 AM - 9:30 AM)\n",
+      "- Start your day at this iconic café with medialunas and coffee.\n",
+      "- - Plaza de Mayo (10:00 AM - 11:00 AM) Visit the historic square, see Casa Rosada, and snap photos.\n",
+      "- - Catedral Metropolitana** (11:15 AM - 12:00 PM)\n",
+      "\n",
+      "Explore the cathedral and its stunning architecture.\n"
+     ]
+    }
+   ],
+   "source": [
+    "# from content_core.config import set_extraction_engine, set_docling_output_format\n",
+    "from content_core.content.extraction import extract_content\n",
+    "\n",
+    "# # 2) Turn on Docling\n",
+    "# set_extraction_engine(\"docling\")\n",
+    "\n",
+    "# # 3) (Optionally) pick your format – markdown, html or json\n",
+    "# set_docling_output_format(\"markdown\")\n",
+    "\n",
+    "# 4) Now extract exactly as before:\n",
+    "result = await extract_content({\n",
+    "    \"file_path\": \"../../../tests/input_content/file.pdf\", \"engine\": \"docling\", \"output_format\": \"markdown\"\n",
+    "})\n",
+    "print(result.content)"
+   ]
   }
  ],
  "metadata": {

content-core 0.4.0__py3-none-any.whl → 0.5.1__py3-none-any.whl

Potentially problematic release.

content-core 0.4.0py3-none-any.whl → 0.5.1py3-none-any.whl