Building a Personal LLM Wiki: The Andrej Karpathy Workflow in Recall
This video demonstrates the tool Recall as a less-hacky alternative to Andrej Karpathy's manual markdown-and-wiki LLM knowledge base, walking through saving content via browser extension, chatting across your saved sources with selectable models, and visualizing, quizzing, and listening to your knowledge.
Astro K JosephWatchTranscript found
Quick learning frame
Read this before watching.
AI strategy is choosing where agents create durable leverage, then managing scope, adoption, risk, and measurable outcomes.
New playlist item from Astro K Joseph; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Skill you build: The ability to set up and operate a personal AI knowledge base in Recall so saved videos, articles, and PDFs become a queryable, source-cited context layer for any LLM.
Watch for the shift from claim to mechanism. The learning value is the point where the transcript reveals a repeatable action, tool boundary, context move, review habit, or artifact.
Concept diagram
Where this video fits.
01Use Case
02Workflow
03Agent Role
04Metric
05Risk
06Adoption
Deep lesson
Turn this video into working knowledge.
3,140 cleaned transcript words reviewed across 876 timed caption segments.
Thesis
Building a Personal LLM Wiki: The Andrej Karpathy Workflow in Recall teaches a practical ai strategy move: This video demonstrates the tool Recall as a less-hacky alternative to Andrej Karpathy's manual markdown-and-wiki LLM knowledge base, walking through saving content via browser extension, chatting across your saved sources with selectable models, and visualizing, quizzing, and listening to your knowledge.
The goal is not to remember the video. The goal is to extract the operating principle, tie it to timestamped evidence, test how far the claim transfers, and make something reusable.
0:26
Knowledge as edge
“and even podcasts into one AI-powered knowledge base that actually works with you. Because right now, tools like Notebook LM lock you into one ecosystem, and even tools like ChatGPT, or let's say Claude, feels like a blank...”
With everyone using the same AI tools, intelligence is no longer the differentiator; your edge is your accumulated personal knowledge, which is normally scattered across bookmarks, notes, and saves that never connect to the AI you use daily. List every place your knowledge currently lives (bookmarks, notes app, YouTube saves, PDFs) and note that none of it is queryable by your AI tools today.
7:09
Extension capture
“question that says, "Summarize everything that I've saved about AI agents and tell me what I'm missing." And if I click on this auto button right here, I'll be able to select the model that I want to...”
The browser extension is the fastest capture path: one click on the Recall icon adds any page or YouTube video instantly and immediately returns a concise summary plus a follow-up Q&A box, so long videos become skimmable without watching end to end. Install the Recall Chrome extension and one-click save three sources you are actively learning from, then read the auto-summary instead of consuming the full content.
8:57
Model-scoped chat
“dopamine detox, and again, here I have deep work, and here we have the answer, right? And the interesting thing is that every time Recall gives you an answer, you can also find the source from where the...”
In the chat view you pick the AI model (e.g. GPT, Claude Opus, Grok, DeepSeek) and set its scope to recall-only, web-only, or both; keeping 'recall plus web' lets it answer from your saved content while filling gaps from the internet, and every answer exposes the exact source it pulled from. Run a 'summarize everything I've saved about X and tell me what I'm missing' query with the scope set to recall plus web, then click the source chips to verify each claim traces back to your own material.
01
Use Case
Start with this video's job: This video demonstrates the tool Recall as a less-hacky alternative to Andrej Karpathy's manual markdown-and-wiki LLM knowledge base, walking through saving content via browser extension, chatting across your saved sources with selectable models, and visualizing, quizzing, and listening to your knowledge. Treat "Use Case" as the outcome you are trying to make visible, not a topic label. Anchor it to 0:26, where the video says: “and even podcasts into one AI-powered knowledge base that actually works with you. Because right now, tools like Notebook LM lock you into one ecosystem, and even tools like ChatGPT, or let's say Claude, feels like a blank...”
02
Workflow
Use "Workflow" to locate the part of the ai strategy workflow the video is demonstrating. Ask what changes in your real setup if this claim is true. Anchor it to 7:09, where the video says: “question that says, "Summarize everything that I've saved about AI agents and tell me what I'm missing." And if I click on this auto button right here, I'll be able to select the model that I want to...”
03
Agent Role
Turn "Agent Role" into the reusable artifact for this lesson: A one-page business case for one agent workflow. This is where watching becomes something you can inspect and reuse.
04
Metric
Use "Metric" as the application surface. Decide whether the idea touches a browser flow, a local file, a model choice, a source document, a UI, or a review step.
05
Risk
Use "Risk" to prove the lesson. The evidence should connect back to the video title, transcript anchors, and a concrete output, not a generic best-practice claim.
06
Adoption
Use "Adoption" to carry the idea forward: save the prompt, checklist, diagram, or operating rule that would make the next agent run better.
Example
Source-backed work packet
Convert the video into a scoped task that includes the transcript claim, target workflow, acceptance criteria, and proof. The output should be a one-page business case for one agent workflow..
Example
Claim vs. demo brief
Separate what the speaker claims, what the demo actually proves, and what still needs outside verification before you adopt the workflow.
Example
Teach-back module
Transform the lesson into a definition, a mechanism diagram, one misconception, one practice exercise, and a check-for-understanding question.
Do not learn it wrong
Treating the title as the lesson without checking what the transcript actually says.
Letting the prompt drift into generic advice that could apply to any video in the playlist.
Copying the tool setup without identifying the operating principle that transfers to your own stack.
Skipping the artifact, which means the learning never becomes operational or inspectable.
Do not count this as learned until these are true.
01
State the transcript-backed claim in your own words: This video demonstrates the tool Recall as a less-hacky alternative to Andrej Karpathy's manual markdown-and-wiki LLM knowledge base, walking through saving content via browser extension, chatting across your saved sources with selectable models, and visualizing, quizzing, and listening to your knowledge.
02
Explain the practical stakes without hype: New playlist item from Astro K Joseph; queued for transcript-backed review, topic mapping, and a practical learning artifact.
03
Map the idea onto the Use Case -> Workflow -> Agent Role -> Metric -> Risk -> Adoption sequence and name the weakest link.
04
Produce the artifact and include the evidence that proves it: A one-page business case for one agent workflow.
Put it into practice
Give this grounded prompt to Codex or Claude after watching.
You are helping me turn one specific YouTube video into real, durable learning.
Source video:
- Title: Building a Personal LLM Wiki: The Andrej Karpathy Workflow in Recall
- URL: https://www.youtube.com/watch?v=uc7ijchbdK4
- Topic: AI Strategy
- My current learning frame: Save five sources on one topic you are studying via the Recall extension, then ask the chat to compare or surface gaps across them using recall-plus-web scope and audit every source chip to confirm the answer is grounded in your knowledge base rather than generic output.
- Why this matters: New playlist item from Astro K Joseph; queued for transcript-backed review, topic mapping, and a practical learning artifact.
Transcript anchors from this exact video:
- 0:26 / Evidence 1: "and even podcasts into one AI-powered knowledge base that actually works with you. Because right now, tools like Notebook LM lock you into one ecosystem, and even tools like ChatGPT, or let's say Claude, feels like a blank..."
- 2:54 / Evidence 2: "creating a personalized knowledge base, right? And at the end of the day, you can use various AI models of your choice from Anthropic, Open AI, or even Gemini and ask questions and as soon as you hit..."
- 4:53 / Evidence 3: "paste a URL in here, a wiki link, upload a PDF file, import content from browser bookmarks, Pocket, or even upload markdown files, or you can directly write content in here as well. So, these are some of..."
- 7:09 / Evidence 4: "question that says, "Summarize everything that I've saved about AI agents and tell me what I'm missing." And if I click on this auto button right here, I'll be able to select the model that I want to..."
- 8:57 / Evidence 5: "dopamine detox, and again, here I have deep work, and here we have the answer, right? And the interesting thing is that every time Recall gives you an answer, you can also find the source from where the..."
- 10:39 / Evidence 6: "visualize your knowledge base, you have the option for that in here. And next up, if you click on this review option right here, you'll be able to select a card, for example, let's just say React tutorial..."
- 12:15 / Evidence 7: "to record my own voice and create a custom voice. And after I add this voice, Recall will basically talk in my own voice. So that's also interesting feature right inside of Recall. So basically, this is how..."
Your task:
1. Use the transcript anchors above as the primary source packet. If you add outside context, label it clearly as outside context and keep it secondary.
2. Create a source-check table with columns: timestamp, claim, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable claims from the video. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A one-page business case for one agent workflow.
5. Include:
- a plain-English definition of the core idea
- a diagram or structured model using this sequence: Use Case -> Workflow -> Agent Role -> Metric -> Risk -> Adoption
- 3 concrete examples that apply the video idea to real agentic work
- 2 failure modes the video helps prevent
- a checklist I can use the next time I run Codex or Claude
- one practical exercise with a clear done signal
6. Add a "learning transfer" section: what changes in my workflow tomorrow if I actually learned this?
7. Add a "source check" section that cites which transcript anchor supports each major takeaway.
Quality bar:
- Make this specific to "Building a Personal LLM Wiki: The Andrej Karpathy Workflow in Recall", not a generic AI Strategy essay.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- If evidence is weak, say what transcript segment or timestamp needs review instead of guessing.
- Finish with a concise artifact I could paste into my learning app.
Misconceptions
What to stop believing.
Every new AI tool deserves a trial.
Every tool has integration cost. Start from workflow pain, not novelty.
If an agent can do it once, it is automated.
Automation means repeatable, monitored, recoverable, and reviewable.
Practice studio
Learning only counts when you make something.
01
Transcript evidence map
Separate what the video actually says from what you already believe about the topic.
3 source-backed takeaways with timestamps, confidence, and a transfer note.02
One useful artifact
Apply the video to a real workflow and produce a one-page business case for one agent workflow..
A reusable artifact with a done signal and one verification step.03
Teach-back card
Explain the lesson to someone who has not watched the video yet.
A 90-second explanation, one diagram, one example, and one misconception to avoid.
Recall check
Answer first, then reveal — without rewatching.
When you save a page or video with the Recall browser extension in one click, what does it immediately give you back, and how does this change how you handle a long (e.g. 1h20m) video?
In Recall's chat view, what two things can you configure before sending a query, and which scope setting does the presenter say is best to keep?
Every answer Recall gives shows source chips. What specifically happens when a cited source is a video versus a web page, and what is the broader point about trusting the answer?
Source shelf
Use the video as a doorway, then verify with primary sources.
