23:18Jake Van Clief · Tutorial
Stop Building AI Agents. Use This Folder System Instead.
A 23-minute walkthrough of the three-layer folder architecture that replaces AI agents, frameworks, and databases with plain markdown files.
March 10thThe argument in one line.
Most AI builders waste money on orchestration frameworks when they should instead structure dialogue and context in folders and markdown files, letting a single agent navigate the actual work.
Read if. Skip if.
READ IF YOU ARE…
- A solo builder or small team currently using Claude/ChatGPT through copy-paste workflows who wants to reduce token spend and iteration time without learning infrastructure tools.
- Someone managing enterprise processes that traditionally require vendor solutions, looking to restructure how your team interfaces with AI models instead of bolting on orchestration layers.
- A technical founder or engineering lead who understands folder structures and markdown but hasn't yet connected that mental model to how LLMs navigate context and retrieve information.
SKIP IF…
- You're already deep in multi-agent frameworks, orchestration platforms, or RAG systems — this is a case for rethinking your layer one, not incrementally improving layer two.
- You need production guardrails, compliance documentation, or formal governance structures before deploying AI systems — the video focuses on methodology, not security or regulatory implementation.
The full version, fast.
Most builders are automating the wrong layer � stacking orchestration frameworks on top of copy-paste chat habits instead of restructuring how context lives on disk. The fix is interpretable context methodology: capture the goals, constraints, assumptions, and decisions that emerge from dialogue, then encode them as folders and markdown files (voice, tone, skills, references) that a single capable agent like Claude Code navigates on demand. Treat conversation as the source of structure, not prompts as the source of magic. Build one Claude.md spine, break voice and process into reusable markdown pillars, and let layered skills replace agentic harnesses. The result is cheaper tokens, deterministic outputs, and workflows teammates can drive together by voice over a call.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:30
01 · The methodology and 30,000 builders
Cold open establishes ICM: folders plus markdown, not multi-agent frameworks. Community proof: 30,000 users, GitHub stars, published paper.
01:30 – 03:00
02 · Layer 1 — chat and copy-paste
The floor most people live on. Log in to ChatGPT or Claude, paste, copy out. Low effort, weak impact.
03:00 – 05:00
03 · Layer 2 — skills and refined prompts
Someone packaged the L1 iteration work. Prompt libraries, chain-of-loop tools, auto-injection. Shows Claude Code skills browser.
05:00 – 07:00
04 · Layer 3 — folders and one agent
Skills evolve into folder structures the agent navigates on demand. No injection harnesses needed. Replaces LangChain/Semantic Kernel.
07:00 – 09:00
05 · The Anthropic/Karpathy connection
ICM is aligned with how frontier orgs actually build. Karpathy LLM wiki, Anthropic skill-based methodology — convergent, not contrarian.
09:00 – 13:30
06 · Why every workflow comes from dialogue
Dialogue contains goals, constraints, assumptions, and decisions. Kay's Divergen Assist tool extracts structured decision trees from any chat log.
13:30 – 17:00
07 · The NLP Logix content pipeline
One folder, roughly 3 prompts: research to script to ElevenLabs audio to structured video animation. Claims this pipeline replaces 4-5 startups.
17:00 – 22:00
08 · Voice-controlled Claude Code in a meeting
Pre-recorded demo: Jake, David McDermott, Kay Kumar on SUPer Fly. Kay controls Jake's Claude Code by voice, adds Short Dark Triad scale to Ethics Engine. $1.20 for ~1 hour.
22:00 – 26:37
09 · Where this goes next
Meetings that fire structured workflows from keyword triggers. Actions before the meeting ends, delivered by agents. Teases open-source project.
Lines worth screenshotting.
- 30,000 people are building AI systems with nothing but folders and markdown files — no orchestration frameworks, no infrastructure, no complex tooling.
- Layer 1 AI usage — logging into ChatGPT and copy-pasting ideas back and forth — is where most builders are stuck, and it's not a technology problem, it's a structure problem.
- Most builders solve the wrong problem: they add orchestration on top of L1 chat habits instead of restructuring at the folder and dialogue layer.
- A three-person voice-controlled Claude Code session ran for an hour and cost $1.20 — the economics of AI collaboration have fundamentally changed.
- 20–40% token reduction comes from giving the model access to structured folders rather than pulling everything into context through RAG.
- Your team becoming the vendor — building the AI structure themselves as they learn — is faster and more durable than hiring a six-month vendor engagement.
- One well-designed harness for a single model type outperforms complex multi-agent frameworks because the model can navigate structure without orchestration overhead.
- Skills — plain text files encoding processes, scripts, and ideas in folders — are the primitive that makes AI context navigable without a retrieval system.
- The interpretable context methodology is structurally similar to how Anthropic itself documents internal AI use cases — the alignment between the methodology and the company's own practice is not coincidental.
Folders and Markdown Beat Every Orchestration Framework
Agent architecture
Jake Van Clief's Interpretable Context Methodology shows that 30,000 builders are getting better results from structured folders and markdown files than from multi-agent frameworks with orchestration code.
01The methodology and 30,000 builders
- Interpretable Context Methodology: folders and markdown files, not multi-agent frameworks — 30,000 users and a published paper validate the approach
- The community proof precedes the methodology explanation — adoption is the argument
02Layer 1 — chat and copy-paste
- Log in, paste, copy out — the floor most people live on — low effort and weak impact relative to what is possible at higher layers
03Layer 2 — skills and refined prompts
- Prompt libraries and chain-of-loop tools are packaged Layer 1 iteration work — useful but still not the real unlock
- Claude Code skills browser is the practical interface for Layer 2 in the Anthropic ecosystem
04Layer 3 — folders and one agent
- Skills evolve into folder structures the agent navigates on demand — no injection harness required
- This approach replaces LangChain and Semantic Kernel without adding orchestration complexity
05The Anthropic/Karpathy connection
- The Karpathy LLM wiki and Anthropic's skill-based methodology converge on the same folder-and-markdown pattern
- Not contrarian — this is what frontier organizations actually build with
06Why every workflow comes from dialogue
- Dialogue contains goals, constraints, assumptions, and decisions — the structured workflow is already latent in the conversation
- Extracting structured decision trees from chat logs turns conversations into reusable process documentation
07The NLP Logix content pipeline
- One folder, roughly three prompts: research to script to ElevenLabs audio to structured video animation
- Three prompts replacing four to five SaaS startups is the unit economics argument for the methodology
08Voice-controlled Claude Code in a meeting
- Voice control of Claude Code during a live meeting at $1.20 for roughly one hour — real-time collaboration at near-zero cost
- Ethical reasoning additions to a live system mid-meeting demonstrate that the agent layer is already fast enough for synchronous work
09Where this goes next
- Meetings that fire structured workflows from keyword triggers and deliver agent actions before the meeting ends
- The open-source project brings this capability to anyone — the architecture is not proprietary
Terms worth knowing.
- Interpretable Context Methodology (ICM)
- A folder-and-markdown-based approach to AI system design where agents navigate plain-text files and directory structures for context instead of complex orchestration code or frameworks.
- Orchestration layer
- Software infrastructure that coordinates multiple AI agents or services — deciding what runs, in what order, and with what inputs — typically added on top of base language models.
- L1 chat habits
- Basic, single-session AI usage patterns where a user types a message and gets a response, without persistent memory, structured context, or automation between turns.
- Multi-agent framework
- A software system that coordinates multiple AI agents working in parallel or in sequence to accomplish complex tasks, with defined communication and handoff protocols between them.
- Markdown file
- A plain-text file using lightweight formatting syntax (headings, bullets, bold) that is human-readable as-is and also renders cleanly in most tools — commonly used for documentation and AI context files.
- Voice-controlled Claude Code
- A setup where speech-to-text input drives Claude Code commands, allowing developers to direct AI coding agents through spoken instructions rather than typed prompts.
Things they pointed at.
00:47conceptKarpathy LLM Wiki methodology
06:55toolLangChain / Semantic Kernel
08:00toolDivergen Assist (dialogue extraction tool by Kay Kumar)
24:00toolVoxtril (open source voice model)
19:50toolSUPer Fly (multi-agent meeting platform)
Lines you could clip.
18:36
“This entire workflow is probably four or five startups in the startup world right now, and it's all in folders and markdown files with one agent.”
08:00
“Conversation has the structure we're looking for. The intent is carried in the conversation.”
02:12
“Your team become the vendor.”
25:34
“Instead of them being plans, they can be actions before the meeting is even done.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:00Interpretable context methodology. 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 people in the community are all using my methodology to build their own systems with AI.
00:15But they're not building multi agentic frameworks and all these absurdities. They're building folders and markdown files on their computer and getting huge results from it.
00:25Now 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?
00:41Now for those of you who are new to my channel, essentially, my methodology is about structuring folders, structuring markdown files. It's very similar to the LLM Wiki that Kaparthi, the founder of OpenAI and, uh, currently just moved to Anthropic to teach there, uses and is very big in Anthropic's actual use cases.
01:00They are very big around skills, which is the idea that the right amount of scripts, processes, and ideas can all be kind of broken down into just plain text and folders, and the agent can navigate that to get the context.
01:15Rather 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 as it needs to. Instead of creating rag that pulls everything in, you just give it access to normal databases.
01:32You 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. 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.
01:54Even further, you're able to use it to really build out teams and build out processes 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.
02:10, 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, about dialogue, about content, and why it's so important to this entire process.
02:29Now 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. 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 is my layer process.
02:48Right? Layer one is where you're just chatting with a chatbot.
02:53You're copy and pasting. 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 or Claude, and all you're doing really is just kind of pasting in ideas.
03:10Right? You're just chatting back and forth in it. You're copying and pasting.
03:14You're throwing in concepts in there. 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 is low, and you can get some pretty decent things.
03:28But the amount of output and impact that you get is actually very weak. Right? It's not gonna be huge.
03:35You can get some very good stuff. 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.
03:48As in, you're essentially adding multiple chats. You have saved prompts. Maybe you have a tone style.
03:53And we see a lot of people sharing prompt libraries. They have shared prompts. Maybe they're adding the prompts in little by little.
04:00A lot of other, like, kind of integration tools do this where they're grabbing a prompt and giving it to the AI automatically. It's doing chain of loop processing. And these skills are actually the layer two.
04:12Someone 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, 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, you don't wanna overload the context window, and you wanna start creating automations.
04:35That's where you start leaning into level three or layer three, which is where you're actually doing this huge workflow. You're taking multiple skills, multiple prompts, multiple workflow processes, maybe even multiple AI, and actually linking them together.
04:51Then 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?
05:00People actually put Python for people to actually run, for the AI to actually run itself to be able to handle the creation of stuff because they needed more determinism.
05:11They wanted the output to be the same every time. And so if they could package all of that into a prompt, all of this could be uploaded and sent in the right order, that's where skills come from.
05:23And 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, you can actually come through and look through these skills deeply.
05:36Some of these skills, right, for algorithmic art, a skill for creating art with, like, kind of an algorithmic process, has an HTML viewer and a template built into it. Maybe you're trying to create a doc file in Markdown.
05:50You have the actual skill explanation. This is how you do it. These are the prompts they put in order.
05:56Traditionally, two years ago, I was putting these prompts in manually. First, I would I knew that, okay. If I told the chatbot this in a certain order, right, I come in and I paste this here, right, it would do that process.
06:11I 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.
06:27But then we could drop all of it into the AI at once and tell it to reference certain areas or actually use commands, and the single agent could then do those things. Right?
06:38Now, traditionally, 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 at the right time in the thinking process of the AI.
06:53But now we can just break down the thinking process into folders and structures and trigger them as they're needed, which is what these folder skills and processes are doing.
07:04But, again, a common theme. There's a reason I'm describing all of this. All of these skills, all of these folders and markdown files, all have one core theme, discussion and dialogue.
07:18They all came from an initial layer one of discussion and dialogue breakdown.
07:24At 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 that are automatically being injected in.
07:41At 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, all of those.
07:54That's what we really want to understand here. But it all comes from dialogue. Conversation has the structure we're looking for.
08:03The intent is carried in the conversation. And instead of just describing this randomly, 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.
08:19So imagine here on the left. This is the chat, and we're gonna break this down to the simplest process. This is a chat that may have happened forever ago.
08:27And 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.
08:35Now in the past, right, the agent would just say, here's a new one. And then we would say, hey.
08:40This 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.
08:52Now 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. Here on the right are the decisions made by the model.
09:11Now when we look in, the goal was to tighten a paragraph. All of this is all of the goals and focus.
09:19And as you can see, it links directly here. Can you tighten this paragraph? It feels too worry.
09:25Now we set up constraints along the process and at different points. Perhaps there was a constraint saying, hey. There was too much wordiness.
09:33We 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.
09:44The voice we're going for is more measured, right, in that same sentence. Maybe we wanna match the entire voice reference.
09:52The assumption is that we need it to match the original, and both the model and the AI follow this assumption.
10:01And 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, is an actual blog post or a reference, and we give an example of what that should be.
10:17There's also dialogue happening. Right? Questions that are happening.
10:20I 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.
10:28It sees the constraints, but it is making decisions on those constraints. It's coming through and saying, hey.
10:35This retains a cleaner structure. I made this decision for this goal. I made this decision for this goal.
10:41And it's also making assumptions along the whole way. We create all of this data.
10:47We look at this, and instead what we can do is we can create a system of tone, of style, of ideas that capture all of these problems and constraints already.
10:59So that when I say tighten this paragraph, this tree of processes, this tree of prompts is called the first time.
11:07Again, thinking about where we're at, originally, it was layer two. We had to sit there and describe these processes, verified outputs.
11:14We captured those, turned those into maybe markdown files or structure, and then the original request becomes a layer one process where I'm able to simply say, can you tighten this paragraph?
11:27Again, this also is the same thing for everything. Let's say we're doing the opposite side.
11:32We're actually refactoring code. And I actually say, hey. My parthon my my Python is treating code a certain way.
11:39I noticed that this parser, can you kind of take a look and fix it possibly? 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.
11:51This 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, then you maybe even have subgoals.
12:07Right? Can we just patch the empty cell instead and keep the original function? You have constraints within this dialogue, and you have processes that live inside of it as well.
12:18Same for the model. The model now has the goal. It's coming through, and it's making decisions on that goal, and all of it can be tracked through dialogue.
12:29That goes for anything. Database schemas, you can track the decision making and goal processes, where and when it was saved within dialogue.
12:38All of this becomes either training data or even better, markdown files that can be used later, converted into a methodology. Right?
12:49And it's important to understand this, not necessarily that you could use this tool. We might make this tool open source.
12:56Right now, it's being used for a research project at the University of Edinburgh. 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, about prompting, and about engineering context, where and what you want to engineer context, at what levels you do, and that's what has informed a lot of the work I'm doing.
13:19And an example is one here. My dialogue with the AI made me realize it sucks at certain processes of reading and research. 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 as well as my tone.
13:41No matter what project I'm working on, I use these processes. So that's all just captured in a very simple short prompt. Right?
13:50That 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, I say, hey.
14:02Every session, I'm going to have certain things built in. My voice and TomeMD directly says how I speak and teach.
14:10And just by me simply saying, right, without creating any crazy agentic architecture, what is my voice and tone?
14:20Now 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, go through everything, finally find the structure.
14:36At layer three, I've solved the problem of it navigating. So it can easily come in, find my voice and tone markdown file, actually look through and see how I teach in tone with one simple prompt.
14:48This prompt is actually thousands of prompts, which if any of you have watched my Python video, 12,000 lines of Python, it's the same thing. When I type hello world, it is thousands and thousands of lines of code that are executing to print hello world in Python.
15:08If you haven't seen that, I highly recommend it. It's very useful for this kind of conceptual. This same thing is happening now with these processes.
15:16This is the next layer of software engineering. But with every layer comes requirements, comes processes, and that is exactly what I'm encoding here.
15:28These processes keep expanding until I finally have an output. 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.
15:42I 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 on each of the products, the angle, where they are, what type of hooks might work, what type of outlines, and materials.
16:01And, again, it's all markdown files. These are just more prompts to be added to context in the future, whether it's this AI or any other AI that I'm using to read these folders.
16:12And 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 said two prompts so far in this whole process of getting all of this research, all of this data, all of this possibility, and it was able to create me these scans, these source materials to then move on to the next step, in which I say, cool.
16:34Let's make a script for each. Now before, if I had said, let's make a script, it would try to assume what a script should be.
16:42I have pillars and processes on how scripts should be written, where, and why. Think of these as my own personal skills. So it goes through.
16:51It reads my voice. Right? It automatically knows my voice constraints.
16:55It does the actual processing. It knows where the documents need to go. Again, with no crazy injection harnesses.
17:02Yes. The Claude code agent has harnesses in its own right, but that's the point. I don't need to reinvent the wheel.
17:09It's already been invented. Then, boom, I get scripts. And guess what?
17:14I say, can you generate the audio for those scripts? I already had a workflow that pulls from my 11 labs on my voice that's already been trained, pulls all that data, and says, cool.
17:26We're gonna go ahead and produce that audio. Then I have the workflow, which I at this point, I can open up a fresh conversation if I wanna save tokens.
17:35All of the process has been saved. All of the context is in the structure. I don't need to inject it back in at the right time.
17:43It can pick right up off. And I can then say, hey. Let's make a video animation.
17:48And rather than show you the code, let me just share with you the animation. It took my voice, and it took my placeholder files that I had already created for animations for this entire project and simply built from them rather than generating new ones.
18:04And it took my voice, chopped it up, and allowed me to create what you see here. Insurance company that emails provider networks to ask if a provider is still in network.
18:16Right? I didn't need to go through and edit. I might refine it.
18:20But 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.
18:29I can come in and find all of those little pieces of data. Again, this entire workflow 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.
18:47Now this is where it gets interesting. Let's say everyone starts doing this. I already have, like, 30,000 people in the community building their own ICMs.
18:55How do you share it? And more importantly, what happens when we go back to dialogue?
19:01What happens when we think about dialogue being the most important part? What if I could sit inside of a group call and control someone else's Claude code or AI through my voice and immediately access all of that data that's locally on their computer.
19:18What would that look like? Well, we built that, and I'm gonna show you exactly what that looks like.
19:24Over 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. And it's not perfect, but it gives you an idea of where the future.
19:44Take 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.
19:51We'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.
20:01I 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 at the University of Edinburgh. We're gonna see how it works.
20:16Kaye, can you go ahead and ask my agent if it's ready to work? Agent, are you ready to work?
20:23Yes. I'm ready to work. I have access to the ethics engine code base, a web application for psychometric assessment of large language models.
20:30Just let me know what you need help with, and I'll get started.
20:33Uh, agent, can you describe what the workspace is in less than 50 words? Ethics engine tests AI models using psychology questionnaires.
20:41It sends questions to different LLMs like GPT and Claude, applies different persona framings, parses their responses into numerical scores, and generates comparative results. It also supports visual assessments using images like Rorschach inkblots.
20:55Very good. Kaye, could you control my agent with your voice through this meeting to read through the scales and then add a new one that it believes should be added?
21:06Oh, can you explain me what scales exist in the repo right now? The repo has 10 built in psychometric scales. One, one, RWA, right wing authoritarianism.
21:26Can you looking at the scales, what scales are we missing from the psychometric analysis engine? Agent?
21:37Looking at what's currently implemented versus common psychometric instruments used in AI ethics research, here are some notable scales that could be added, missing scales that would be relevant. One dark triad measures, Machiavellianism, narcissism, psychopathy.
21:52Two empathy scales like the interpersonal reactivity index. Three, moral disengagement scale. Four, honesty, humility from Hexaco personality model.
22:00Five, cognitive reflection test for reasoning biases. Six, Schwartz value survey for basic human value. Seven, discuss sensitivity
22:07scale. Eight. Okay.
22:07That sounds good. Which one Nine. Which of those is your top pick, though?
22:13Just pick one.
22:16My top pick would be the dark triad. It measures Machiavellianism,
22:19narcissism, and psychopathy in one instrument. Okay. That sounds like a good addition.
22:24Can you add that to the psychometric analysis of the engine?
22:28I'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.
22:35Now I'll add the short dark triad scale. This is the SD three by Jones and Paul Hus. Okay.
22:45Good job. Uh, can you go into more detail about the documentation that you changed?
22:50Sure. 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.
23:02IHS, 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, narcissism, and psychopathy.
23:15Second, in the key file section under back end, I updated the description of builtinscales.pwai
23:20from 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 app better from the perspective of the scale you just added?
23:37Good question. Let me take a look at the front end to see if there's anything that needs update for the new The component shows the scale name, description, item count, and scale range, all of which we defined.
23:49One thing that could be improved for the dark triad specifically, the front end doesn't currently display subscale information. The s d three has three subscales, Machiavellianism, narcissism, psychopathy, but users would just see it as one scale with 27 items.
24:05They 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.
24:12Mind you, this was my master's thesis that took me a year to build out.
24:19Yeah. It's doing a pretty good job of understanding good job.
24:22What I'm requesting.
24:23Now 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 20¢ 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.
24:47And, 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, but we can use that dialogue, that process of multiple people working together with an AI to generate better dialogue structure.
25:10One of the things that we're working on right now is that kind of process of interruption, 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. 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, when we capture meetings.
25:35Instead of them being plans, 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.
25:45It's part of a larger project that I'm putting together because, again, I think this is one solution 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.
25:58But other than that, I hope this video kind of taught you a lot today. 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 dialogue, behind my interpretable context methodology.
26:20If 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.
26:30I hope it was interesting to watch. I hope you learned a lot. And until next time, happy learning.
The Hook
The bait, then the rug-pull.
Jake Van Clief opens not with a question or a hot take, but with a credential drop that lands differently than most: a published methodology paper, 30,000 practitioners, and a community built entirely around folders and markdown files. The title does the provocation; the open does the credentialing. By the time he says they are not building multi-agentic frameworks, you already believe him.
Named ideas worth stealing.
02:47model
The Three Layers of AI Use
- L1 — Copy and Paste (chat, low effort, weak output)
- L2 — Structured Use (skills, prompt libraries, better output)
- L3 — Integrated Workflow (folders + one agent, highest output)
A tiered model for AI workflow maturity. Each layer is a step change in setup effort and output quality.
Steal forAny AI positioning content or course ladder — maps to beginner/intermediate/advanced without gatekeeping language.
00:00concept
Interpretable Context Methodology (ICM)
Structure AI context through folders and markdown files rather than programmatic harnesses. The agent navigates structure on demand instead of having context injected at runtime.
Steal forOwn-your-stack positioning: replace SaaS orchestration tools with folders you control.
09:00concept
Dialogue Decision Tree Extraction
Every AI chat contains extractable structure: goals, constraints, assumptions, processes. These become the markdown files that power L3 systems.
Steal forProductizing existing chat logs — any creator or consultant can mine old Claude/ChatGPT sessions for methodology content.
How they asked for the click.
VERBAL ASK
26:02next-video
“If you are watching this as just a demo video for Vox... please go check out my larger thirty or forty minute video.”
Soft, non-pushy. Dual CTA for two audience segments (Vox demo viewers vs YouTube subscribers). No subscribe ask, no sponsor. Clean close.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
