WEBVTT 00:00:00.000 --> 00:00:02.240 Interpretable context methodology. 00:00:02.240 --> 00:00:09.920 That's the name of my paper that I published about three months ago, and now thousands of people and hundreds of GitHub stars as well over 30,000 00:00:09.920 --> 00:00:15.605 people in the community are all using my methodology to build their own systems with AI. 00:00:15.685 --> 00:00:19.445 But they're not building multi agentic frameworks and all these absurdities. 00:00:19.525 --> 00:00:23.525 They're building folders and markdown files on their computer and getting 00:00:23.685 --> 00:00:46.755 huge results from it. Now I have quite a few videos on that, but today, I wanted to talk less about the methodology and more about what I've been building, what me and my team have been building as a future thing. What can you actually build using this methodology, and what happens when you bring in some traditional software programming into it? Now for those of you who are new to my channel, essentially, my methodology is about structuring 00:00:47.075 --> 00:00:53.395 folders, structuring markdown files. It's very similar to the LLM Wiki that Kaparthi, the founder of OpenAI 00:00:53.710 --> 00:01:05.310 and, uh, currently just moved to Anthropic to teach there, uses and is very big in Anthropic's actual use cases. They are very big around skills, which is the idea that the right amount of scripts, 00:01:05.630 --> 00:01:07.070 processes, and ideas 00:01:07.235 --> 00:01:15.155 can all be kind of broken down into just plain text and folders, and the agent can navigate that to get the context. 00:01:15.235 --> 00:01:26.660 Rather than using really complicated workflows or infrastructures and harnesses, you can create one really good harness for a single type of model, and then that model can then navigate structure 00:01:26.820 --> 00:01:43.655 as it needs to. Instead of creating rag that pulls everything in, you just give it access to normal databases. You give it access to folder and context structure. If you haven't read anything or dive into this process, I do a deep one with Claude code in my video right here that you should see tagged. 00:01:43.815 --> 00:01:54.510 Essentially, it just looks into how to use this, how I use it, and a lot of people are seeing twenty, thirty, 40% token reduction and really just getting outcomes a lot faster without a lot of infrastructure. 00:01:54.750 --> 00:01:59.390 Even further, you're able to use it to really build out teams and build out processes 00:01:59.470 --> 00:02:12.825 that traditionally would have taken, you know, a vendor six months and an entire infrastructure update versus just working with them. And as the people learn how to use the AI, they're building the structure for you. Your team become the vendor, 00:02:12.985 --> 00:02:23.420 and then you can hire technical talent or bring in your own technical talent to amplify it, look at the security and governance behind it. But all of that aside, let me chat about transcripts, 00:02:23.420 --> 00:02:24.540 about dialogue, 00:02:24.540 --> 00:02:25.580 about content, 00:02:25.580 --> 00:02:28.860 and why it's so important to this entire process. 00:02:29.020 --> 00:02:38.745 Now this is an actual slide that I gave to a real enterprise company that's bringing in, I think, 60,000,000 a year or something like that, and they have learned and used this process. 00:02:38.905 --> 00:02:47.225 One of the biggest things that I try to share with people when we're trying to understand how to use AI, how to use these tools in your workflow 00:02:47.225 --> 00:02:50.550 is my layer process. Right? Layer one 00:02:50.790 --> 00:02:54.870 is where you're just chatting with a chatbot. You're copy and pasting. 00:02:55.270 --> 00:03:05.765 The layer one is this concept where I started years ago when I first started messing around with language processing systems, and a lot of people are today. It's where you log in to ChatGPT 00:03:05.765 --> 00:03:07.045 or Claude, 00:03:07.125 --> 00:03:14.725 and all you're doing really is just kind of pasting in ideas. Right? You're just chatting back and forth in it. You're copying and pasting. 00:03:14.725 --> 00:03:16.485 You're throwing in concepts in there. 00:03:17.090 --> 00:03:25.410 You're just asking it basically questions and things like that. What you actually wanna really understand is the amount of effort to get something back 00:03:25.650 --> 00:03:26.690 is low, 00:03:26.770 --> 00:03:28.610 and you can get some pretty decent things. 00:03:28.850 --> 00:03:34.085 But the amount of output and impact that you get is actually very weak. 00:03:34.165 --> 00:03:37.445 Right? It's not gonna be huge. You can get some very good stuff. 00:03:37.685 --> 00:03:48.080 But earlier, three years ago, it was very challenging to get any sort of outputs. And in order to really get a better output out of it, you have to longer and longer conversations. 00:03:48.320 --> 00:03:52.240 As in, you're essentially adding multiple chats. You have saved prompts. 00:03:52.320 --> 00:03:56.560 Maybe you have a tone style. And we see a lot of people sharing prompt libraries. 00:03:56.640 --> 00:04:07.925 They have shared prompts. Maybe they're adding the prompts in little by little. A lot of other, like, kind of integration tools do this where they're grabbing a prompt and giving it to the AI automatically. 00:04:08.005 --> 00:04:10.245 It's doing chain of loop processing. 00:04:10.245 --> 00:04:21.930 And these skills are actually the layer two. Someone else went through that layer two workflow, found the right amount of prompts that you actually wanna be able to give, in what order you wanna give them, 00:04:22.330 --> 00:04:30.655 and kind of handled that as a skill. But then skills evolve even further, and you realize you only need certain parts of a skill at a certain time, 00:04:30.895 --> 00:04:35.455 you don't wanna overload the context window, and you wanna start creating automations. 00:04:35.615 --> 00:04:39.615 That's where you start leaning into level three or layer three, 00:04:39.775 --> 00:04:42.495 which is where you're actually doing this huge workflow. 00:04:42.910 --> 00:04:49.150 You're taking multiple skills, multiple prompts, multiple workflow processes, maybe even multiple AI, 00:04:49.630 --> 00:05:02.035 and actually linking them together. Then they realized that you want to do even more steps. You want these agents, And that's where you have these other markdown files or scripts in many areas. Right? People actually put Python 00:05:02.035 --> 00:05:06.595 for people to actually run, for the AI to actually run itself 00:05:06.595 --> 00:05:10.995 to be able to handle the creation of stuff because they needed more determinism. 00:05:11.370 --> 00:05:18.410 They wanted the output to be the same every time. And so if they could package all of that into a prompt, 00:05:18.410 --> 00:05:20.010 all of this could be uploaded 00:05:20.010 --> 00:05:21.770 and sent in the right order, 00:05:22.170 --> 00:05:32.705 that's where skills come from. And this allows them to then, in this case, create more skills, but you have hundreds of skills available. And if you have Claude installed on your computer, 00:05:32.945 --> 00:05:42.670 you can actually come through and look through these skills deeply. Some of these skills, right, for algorithmic art, a skill for creating art with, like, kind of an algorithmic process, 00:05:42.910 --> 00:05:50.430 has an HTML viewer and a template built into it. Maybe you're trying to create a doc file in Markdown. 00:05:50.590 --> 00:05:52.990 You have the actual skill explanation. 00:05:52.990 --> 00:06:04.325 This is how you do it. These are the prompts they put in order. Traditionally, two years ago, I was putting these prompts in manually. First, I would I knew that, okay. If I told the chatbot 00:06:04.405 --> 00:06:08.725 this in a certain order, right, I come in and I paste this here, 00:06:09.870 --> 00:06:34.185 right, it would do that process. I would wait for its output, verify that it's right, and then I would move on to the next step, which is, okay. Well, then we need to edit that and have these things. And over time, we realized we could drop that into the AI all at once, and it would recognize that. But then we could drop all of it into the AI at once and tell it to reference certain areas or actually use commands, 00:06:34.505 --> 00:06:39.410 and the single agent could then do those things. Right? Now, traditionally, 00:06:39.570 --> 00:06:48.450 before the way they solve that is using lang chain or semantic kernel or all of these things to do these actions separately or to inject these 00:06:48.690 --> 00:06:49.810 at the right time 00:06:50.495 --> 00:06:52.895 in the thinking process of the AI. 00:06:53.055 --> 00:06:58.175 But now we can just break down the thinking process into folders and structures 00:06:58.255 --> 00:07:00.575 and trigger them as they're needed, 00:07:00.895 --> 00:07:03.615 which is what these folder skills and processes 00:07:03.615 --> 00:07:10.450 are doing. But, again, a common theme. There's a reason I'm describing all of this. All of these skills, 00:07:10.450 --> 00:07:13.090 all of these folders and markdown files, 00:07:13.490 --> 00:07:15.410 all have one core theme, 00:07:16.130 --> 00:07:17.170 discussion 00:07:17.170 --> 00:07:18.130 and dialogue. 00:07:18.745 --> 00:07:20.025 They all came 00:07:20.265 --> 00:07:23.385 from an initial layer one of discussion 00:07:23.385 --> 00:07:39.010 and dialogue breakdown. At layer one, you're pasting a prompt into ChatGPT and then copying the output or going through iterations of an output. At level two, you already expect what the output is from level one, so you have refined prompts and processes 00:07:39.090 --> 00:07:53.975 that are automatically being injected in. At level three, you're doing all of the level two automated multiple times so that you can then come in and look at these processes. And that goes for code review, data reporting, writing and drafting, 00:07:54.215 --> 00:07:59.415 all of those. That's what we really want to understand here. But it all comes from dialogue. 00:07:59.730 --> 00:08:00.850 Conversation 00:08:00.850 --> 00:08:08.290 has the structure we're looking for. The intent is carried in the conversation. And instead of just describing this randomly, 00:08:08.370 --> 00:08:18.795 here's Atul Kay, who is one of my, uh, cofounders and a really brilliant person, certainly studying at the University of Edinburgh, created to really help exemplify this process. 00:08:19.035 --> 00:08:25.115 So imagine here on the left. This is the chat, and we're gonna break this down to the simplest process. 00:08:25.355 --> 00:08:36.530 This is a chat that may have happened forever ago. And the idea is, can you tighten this paragraph? It feels too worry. Climate change, blah blah blah, and you put in the paragraph in there. Now in the past, 00:08:36.770 --> 00:09:07.770 right, the agent would just say, here's a new one. And then we would say, hey. This feels weirdly formal now. The original had more of a conversational rhythm. And so the agent does something, and then it's this constant back and forth until we finally get the output we wanted out of it. Now everywhere along this conversation, even in this simple one, there is a chain of decisions being made. And if we actually extract those chain of decisions using Kaye's tool here, here on the left is the decisions made by the user. 00:09:08.010 --> 00:09:11.210 Here on the right are the decisions made by the model. 00:09:11.645 --> 00:09:13.885 Now when we look in, the goal 00:09:14.045 --> 00:09:16.365 was to tighten a paragraph. 00:09:16.605 --> 00:09:23.725 All of this is all of the goals and focus. And as you can see, it links directly here. Can you tighten this paragraph? 00:09:23.805 --> 00:09:24.765 It feels too worry. 00:09:25.340 --> 00:09:33.260 Now we set up constraints along the process and at different points. Perhaps there was a constraint saying, hey. There was too much wordiness. 00:09:33.340 --> 00:09:47.355 We want to perceive the rhythm or conserve what the rhythm in the original paragraph was. Maybe we wanna have it more styled or evolved. Right here, we want it scarier, way behind. The voice we're going for is more measured, 00:09:47.515 --> 00:09:49.275 right, in that same sentence. 00:09:49.355 --> 00:09:53.540 Maybe we wanna match the entire voice reference. The assumption 00:09:54.020 --> 00:09:57.220 is that we need it to match the original, 00:09:57.220 --> 00:10:09.985 and both the model and the AI follow this assumption. And don't worry. We're gonna go over to this right one in a second, but I really want you to understand this. Finally, the goal of what we're making, the process, 00:10:10.225 --> 00:10:14.945 is an actual blog post or a reference, and we give an example 00:10:15.185 --> 00:10:31.790 of what that should be. There's also dialogue happening. Right? Questions that are happening. I think there's a way in which to do these things. When we go and look at the model over here, the same thing. The model sees the goal. It sees the constraints, but it is making decisions 00:10:31.950 --> 00:10:43.565 on those constraints. It's coming through and saying, hey. This retains a cleaner structure. I made this decision for this goal. I made this decision for this goal. And it's also making assumptions 00:10:43.565 --> 00:10:45.325 along the whole way. 00:10:45.565 --> 00:10:50.205 We create all of this data. We look at this, and instead what we can do 00:10:50.550 --> 00:10:55.110 is we can create a system of tone, of style, of ideas 00:10:55.430 --> 00:10:58.950 that capture all of these problems and constraints already. 00:10:59.030 --> 00:11:01.350 So that when I say tighten this paragraph, 00:11:01.750 --> 00:11:04.065 this tree of processes, 00:11:04.065 --> 00:11:10.225 this tree of prompts is called the first time. Again, thinking about where we're at, originally, 00:11:10.225 --> 00:11:18.705 it was layer two. We had to sit there and describe these processes, verified outputs. We captured those, turned those into maybe markdown files or structure, 00:11:19.240 --> 00:11:23.880 and then the original request becomes a layer one process 00:11:23.960 --> 00:11:27.400 where I'm able to simply say, can you tighten this paragraph? 00:11:27.560 --> 00:11:28.280 Again, 00:11:28.440 --> 00:11:31.000 this also is the same thing for everything. 00:11:31.160 --> 00:11:43.905 Let's say we're doing the opposite side. We're actually refactoring code. And I actually say, hey. My parthon my my Python is treating code a certain way. I noticed that this parser, can you kind of take a look and fix it possibly? 00:11:43.985 --> 00:12:05.215 And, of course, the instant assistant comes in, says what it thinks its fix is. Maybe we say some sort of reference, because we're an expert in the field, maybe. This is how they did training early on is this kind of back and forth, and you eventually get to the conclusion, the goal. When we actually extract that data, right, you end up getting the same thing. You have goals, 00:12:05.455 --> 00:12:07.375 then you maybe even have subgoals. 00:12:07.375 --> 00:12:12.095 Right? Can we just patch the empty cell instead and keep the original function? 00:12:12.255 --> 00:12:13.535 You have constraints 00:12:13.535 --> 00:12:14.815 within this dialogue, 00:12:15.250 --> 00:12:16.850 and you have processes 00:12:16.850 --> 00:12:24.930 that live inside of it as well. Same for the model. The model now has the goal. It's coming through, and it's making decisions 00:12:25.250 --> 00:12:28.945 on that goal, and all of it can be tracked through 00:12:29.025 --> 00:12:35.585 dialogue. That goes for anything. Database schemas, you can track the decision making and goal processes, 00:12:35.745 --> 00:12:38.785 where and when it was saved within dialogue. 00:12:38.945 --> 00:12:43.120 All of this becomes either training data or even better, 00:12:43.760 --> 00:12:49.040 markdown files that can be used later, converted into a methodology. 00:12:49.200 --> 00:12:52.160 Right? And it's important to understand this, 00:12:52.400 --> 00:12:59.905 not necessarily that you could use this tool. We might make this tool open source. Right now, it's being used for a research project at the University of Edinburgh. 00:13:00.065 --> 00:13:07.345 Um, but the idea here isn't for you to all use this tool right now. The idea is what this tool shows about dialogue, 00:13:07.505 --> 00:13:08.785 about prompting, 00:13:08.945 --> 00:13:10.465 and about engineering 00:13:10.465 --> 00:13:11.105 context, 00:13:11.530 --> 00:13:15.690 where and what you want to engineer context, at what levels 00:13:15.770 --> 00:13:17.770 you do, and that's what has informed 00:13:17.770 --> 00:13:21.770 a lot of the work I'm doing. And an example is one here. 00:13:21.930 --> 00:13:29.245 My dialogue with the AI made me realize it sucks at certain processes of reading and research. 00:13:29.405 --> 00:13:40.070 So I created a kind of pillar of idea and my voice and my patterns and constraints that capture my process of writing and researching 00:13:40.070 --> 00:13:45.270 as well as my tone. No matter what project I'm working on, I use these processes. 00:13:45.430 --> 00:13:49.110 So that's all just captured in a very simple short prompt. 00:13:49.190 --> 00:13:49.430 Right? 00:13:50.225 --> 00:14:01.425 That is referenced whenever I need it through the initial prompt, which is my Claude MD. That gives folder structure of wherever I'm hiding. And wherever that voice is, 00:14:01.745 --> 00:14:03.585 I say, hey. Every session, 00:14:03.745 --> 00:14:15.960 I'm going to have certain things built in. My voice and TomeMD directly says how I speak and teach. And just by me simply saying, right, without creating any crazy agentic architecture, 00:14:16.280 --> 00:14:17.800 what is my voice 00:14:18.955 --> 00:14:20.235 and tone? 00:14:20.395 --> 00:14:33.195 Now if I had asked an AI model that was at layer one or two years ago, it would have no idea. I'd have to paste a whole bunch of prompts in there, set them all in. If I was one at layer two, it would have to read the whole database, 00:14:33.275 --> 00:14:44.130 go through everything, finally find the structure. At layer three, I've solved the problem of it navigating. So it can easily come in, find my voice and tone markdown file, 00:14:44.450 --> 00:14:56.715 actually look through and see how I teach in tone with one simple prompt. This prompt is actually thousands of prompts, which if any of you have watched my Python video, 12,000 lines of Python, 00:14:57.035 --> 00:15:06.550 it's the same thing. When I type hello world, it is thousands and thousands of lines of code that are executing to print 00:15:06.630 --> 00:15:08.790 hello world in Python. 00:15:08.870 --> 00:15:12.710 If you haven't seen that, I highly recommend it. It's very useful for this kind of conceptual. 00:15:13.515 --> 00:15:20.235 This same thing is happening now with these processes. This is the next layer of software engineering. 00:15:20.475 --> 00:15:24.075 But with every layer comes requirements, 00:15:24.075 --> 00:15:25.515 comes processes, 00:15:25.675 --> 00:15:32.610 and that is exactly what I'm encoding here. These processes keep expanding until I finally have an output. 00:15:32.770 --> 00:15:54.475 In this case, my output was the ability to not only write and research in my own area. So I had a company, NLP Logics, that wanted me to create videos for them. I spent time building ideas and my tone around it, and I said, I wanna make three videos based on three of their products. So it goes through. It has access to my research and creates a research folder 00:15:54.715 --> 00:15:57.190 on each of the products, the angle, 00:15:57.350 --> 00:16:07.270 where they are, what type of hooks might work, what type of outlines, and materials. And, again, it's all markdown files. These are just more prompts to be added to context in the future, 00:16:07.590 --> 00:16:20.045 whether it's this AI or any other AI that I'm using to read these folders. And then all I have to say is, cool. Now do the research based off of all of that. Then with a third prompt, again, I'm only technically 00:16:20.285 --> 00:16:24.285 said two prompts so far in this whole process of getting all of this research, 00:16:24.445 --> 00:16:27.020 all of this data, all of this possibility, 00:16:27.180 --> 00:16:30.780 and it was able to create me these scans, these source materials 00:16:30.860 --> 00:16:37.420 to then move on to the next step, in which I say, cool. Let's make a script for each. Now before, 00:16:37.500 --> 00:16:50.335 if I had said, let's make a script, it would try to assume what a script should be. I have pillars and processes on how scripts should be written, where, and why. Think of these as my own personal skills. 00:16:50.495 --> 00:16:52.815 So it goes through. It reads my voice. 00:16:53.420 --> 00:17:00.060 Right? It automatically knows my voice constraints. It does the actual processing. It knows where the documents need to go. Again, 00:17:00.380 --> 00:17:10.855 with no crazy injection harnesses. Yes. The Claude code agent has harnesses in its own right, but that's the point. I don't need to reinvent the wheel. It's already been invented. 00:17:11.095 --> 00:17:17.735 Then, boom, I get scripts. And guess what? I say, can you generate the audio for those scripts? 00:17:17.975 --> 00:17:23.650 I already had a workflow that pulls from my 11 labs on my voice that's already been trained, 00:17:24.050 --> 00:17:31.330 pulls all that data, and says, cool. We're gonna go ahead and produce that audio. Then I have the workflow, 00:17:31.410 --> 00:17:40.885 which I at this point, I can open up a fresh conversation if I wanna save tokens. All of the process has been saved. All of the context is in the structure. 00:17:41.125 --> 00:17:45.685 I don't need to inject it back in at the right time. It can pick right up off. 00:17:46.005 --> 00:17:52.420 And I can then say, hey. Let's make a video animation. And rather than show you the code, let me just share with you the animation. 00:17:52.740 --> 00:17:54.500 It took my voice, 00:17:55.060 --> 00:17:56.820 and it took my placeholder 00:17:56.820 --> 00:18:12.805 files that I had already created for animations for this entire project and simply built from them rather than generating new ones. And it took my voice, chopped it up, and allowed me to create what you see here. Insurance company that emails provider networks 00:18:12.885 --> 00:18:16.085 to ask if a provider is still in network. 00:18:16.485 --> 00:18:36.225 Right? I didn't need to go through and edit. I might refine it. But the best part is this isn't random video generation. This is structured output that I can then edit at scale. I can edit with my personal touches. I can come in and find all of those little pieces of data. Again, this entire workflow 00:18:36.465 --> 00:18:46.705 is probably four or five startups in the startup world right now, and it's all in folders and markdown files with one agent, one singular agent. 00:18:47.290 --> 00:18:49.610 Now this is where it gets interesting. 00:18:50.010 --> 00:18:55.450 Let's say everyone starts doing this. I already have, like, 30,000 people in the community building their own ICMs. 00:18:55.610 --> 00:18:57.770 How do you share it? And more importantly, 00:18:58.090 --> 00:19:00.570 what happens when we go back to dialogue? 00:19:01.005 --> 00:19:09.885 What happens when we think about dialogue being the most important part? What if I could sit inside of a group call and control 00:19:09.965 --> 00:19:15.420 someone else's Claude code or AI through my voice and immediately 00:19:15.420 --> 00:19:20.700 access all of that data that's locally on their computer. What would that look like? Well, 00:19:20.940 --> 00:19:40.570 we built that, and I'm gonna show you exactly what that looks like. Over the next few minutes, you're gonna see an actual recording from a live call that me and my team did built around this idea. This is going to be myself, David McDermott, and Kay Kumar working on a project that Kay specifically has been building for a little while now around this concept. 00:19:40.810 --> 00:19:42.330 And it's not perfect, 00:19:42.730 --> 00:20:11.790 but it gives you an idea of where the future. Take it with a grain of salt, and when the clips are done, I'll come back and I have more to explain. Alright. This is test one of one. We've totally never tested this before. We've hooked up my local Claude code. It is currently running locally on my computer, obviously, through my subscription account. I opened it up inside of, um, it's running as a CLI inside of audit engine folder setup here, which is a big old database I used for my psychometric research 00:20:11.870 --> 00:20:15.870 at the University of Edinburgh. We're gonna see how it works. 00:20:16.110 --> 00:20:19.070 Kaye, can you go ahead and ask my agent 00:20:19.395 --> 00:20:21.075 if it's ready to work? 00:20:21.795 --> 00:20:32.115 Agent, are you ready to work? Yes. I'm ready to work. I have access to the ethics engine code base, a web application for psychometric assessment of large language models. Just let me know what you need help with, and I'll get started. 00:20:33.320 --> 00:20:41.560 Uh, agent, can you describe what the workspace is in less than 50 words? Ethics engine tests AI models using psychology questionnaires. 00:20:41.640 --> 00:20:49.815 It sends questions to different LLMs like GPT and Claude, applies different persona framings, parses their responses into numerical scores, 00:20:49.895 --> 00:20:55.655 and generates comparative results. It also supports visual assessments using images like Rorschach inkblots. 00:20:55.655 --> 00:21:00.695 Very good. Kaye, could you control my agent with your voice through this meeting 00:21:01.050 --> 00:21:03.130 to read through the scales 00:21:03.290 --> 00:21:15.195 and then add a new one that it believes should be added? Oh, can you explain me what scales exist in the repo right now? The repo has 10 built in psychometric scales. One, one, 00:21:16.635 --> 00:21:17.355 RWA, 00:21:17.515 --> 00:21:18.955 right wing authoritarianism. 00:21:26.090 --> 00:21:26.970 Can you 00:21:27.850 --> 00:21:32.890 looking at the scales, what scales are we missing from the psychometric analysis engine? 00:21:34.090 --> 00:21:34.810 Agent? 00:21:37.850 --> 00:21:43.755 Looking at what's currently implemented versus common psychometric instruments used in AI ethics research, 00:21:43.915 --> 00:21:50.875 here are some notable scales that could be added, missing scales that would be relevant. One dark triad measures, Machiavellianism, 00:21:50.875 --> 00:21:52.555 narcissism, psychopathy. 00:21:52.555 --> 00:22:07.120 Two empathy scales like the interpersonal reactivity index. Three, moral disengagement scale. Four, honesty, humility from Hexaco personality model. Five, cognitive reflection test for reasoning biases. Six, Schwartz value survey for basic human value. Seven, discuss sensitivity 00:22:07.125 --> 00:22:14.565 scale. Eight. Okay. That sounds good. Which one Nine. Which of those is your top pick, though? Just pick one. 00:22:16.005 --> 00:22:19.765 My top pick would be the dark triad. It measures Machiavellianism, 00:22:19.765 --> 00:22:23.890 narcissism, and psychopathy in one instrument. Okay. That sounds like a good addition. 00:22:24.610 --> 00:22:28.530 Can you add that to the psychometric analysis of the engine? 00:22:28.850 --> 00:22:40.685 I'll add the short dark triad scale to the engine now. Let me first check the existing scale structure to match the format. I see the scale structure. Now I'll add the short dark triad scale. This is the SD three by Jones and Paul Hus. 00:22:41.965 --> 00:22:42.605 Okay. 00:22:45.965 --> 00:22:49.965 Good job. Uh, can you go into more detail about the documentation that you changed? 00:22:50.560 --> 00:23:03.040 Sure. I updated the Claude MD file in two places. First, in the built in data section, I changed the scales table from 10 validated instruments to 12 validated instruments and added two new rows. IHS, 00:23:03.040 --> 00:23:13.655 the intellectual humility scale with six items on a one to five range, and SD three, the short dark triad with 27 items on a one to five range covering Machiavellianism, 00:23:13.655 --> 00:23:15.415 narcissism, and psychopathy. 00:23:15.655 --> 00:23:20.890 Second, in the key file section under back end, I updated the description of builtinscales.pwai 00:23:20.890 --> 00:23:30.730 from 10 built in validated scales to 12 built in validated scales to reflect the additions. Looking at the front end, is there anything we're missing in the front end that would make this 00:23:31.450 --> 00:23:35.765 app better from the perspective of the scale you just added? 00:23:37.685 --> 00:23:42.085 Good question. Let me take a look at the front end to see if there's anything that needs update for the new 00:23:46.530 --> 00:23:53.730 The component shows the scale name, description, item count, and scale range, all of which we defined. One thing that could be improved for the dark triad specifically, 00:23:53.810 --> 00:23:56.930 the front end doesn't currently display subscale information. 00:23:57.090 --> 00:24:00.290 The s d three has three subscales, Machiavellianism, 00:24:00.290 --> 00:24:01.170 narcissism, 00:24:01.170 --> 00:24:01.810 psychopathy, 00:24:02.135 --> 00:24:15.815 but users would just see it as one scale with 27 items. They wouldn't know which items match which trait. Would you like me to subscale display support so users can see the breakdown against my trait? That would be useful, not just for s e three, but also for other scales like RWA, LDA, and FQ to have subscales too. Mind you, this was my master's thesis that took me a year to build out. 00:24:19.050 --> 00:24:21.370 Yeah. It's doing a pretty good job of understanding 00:24:21.690 --> 00:24:23.130 good job. What I'm requesting. 00:24:23.530 --> 00:25:01.735 Now we proceeded to play with it for about another hour or two. And I think in total, with the tokens we were using, my custom AI voice that's routing in through API, we're using 11 labs at the time, but I think we're gonna switch over to Voxtril, which is a more open source voice model. Uh, we ended up spending a dollar and 20ยข for about an hour of our time and effort, and I think we could drop that number down exponentially if we used more open source models and whatnot. And, realistically, the voice one is something that can be amplified by a lot of NVIDIA works and things like that. The point is not only are we able to control someone else's computer with their agent and their context, or someone else can control mine, 00:25:01.895 --> 00:25:04.055 but we can use that dialogue, 00:25:04.215 --> 00:25:14.470 that process of multiple people working together with an AI to generate better dialogue structure. One of the things that we're working on right now is that kind of process of interruption, 00:25:14.710 --> 00:25:23.190 because I see an opportunity in the future where this dialogue structure, these structured workflows, all of these things are being triggered by keywords in conversations. 00:25:23.715 --> 00:25:34.115 When we have a meeting, we sit there and dive in and talk together, and I think that would be absolutely amazing to see these come and flow while we're in a meeting, when we capture transcripts, 00:25:34.115 --> 00:25:37.410 when we capture meetings. Instead of them being plans, 00:25:37.410 --> 00:25:50.610 they can be actions before the meeting is even done, and we can even hear about those actions directly from these agents. Again, this is more of an open source tool. It's part of a larger project that I'm putting together because, again, I think this is one solution 00:25:51.065 --> 00:26:02.505 to a larger problem that needs more solutions added in. And you'll see as I'm posting videos the kind of structure I'm building for it. But other than that, I hope this video kind of taught you a lot today. 00:26:02.745 --> 00:26:16.130 And if you are watching this as just a demo video for Vox, which I'm cutting this video up, then please go check out my larger I think it's gonna probably be thirty or forty minute video where I describe the content and the structure behind 00:26:16.210 --> 00:26:17.170 dialogue, 00:26:17.330 --> 00:26:18.850 behind my interpretable 00:26:18.850 --> 00:26:19.810 context methodology. 00:26:20.075 --> 00:26:35.675 If you are just watching this because this is the Vox video and this is the cut I did, well, then I hope you enjoyed it. I hope you check out more things. Other than that, thank you everyone for staying to the end. I hope it was interesting to watch. I hope you learned a lot. And until next time, happy learning.