The fastest way to stop seeing AI as magic is to watch a genuinely great person just... use it. This is the watchlist Mariah promised: Karpathy's everyday-AI walkthrough plus a few more, with notes on what to steal from each one, and a prompt that turns watching into doing.
Andrej Karpathy was a founding member of OpenAI and ran AI at Tesla. He's one of the best AI researchers alive. And in this video, he does something almost nobody at his level does: he just shows his screen and talks to AI the way he does every day. No hype, no launch, no secret tools.
Don't watch it to learn the latest features (it's over a year old now, and some tools have moved on). Watch it for the one skill that hasn't aged a day: how a great practitioner talks to AI. Plain questions. Quick experiments. Zero fear of asking something "dumb." By the end you stop thinking "I could never do that" because you can see, on screen, that there's no trick to it.
It's 2 hours 11 minutes, and you don't need to watch it in one sitting. It's chaptered, so treat it like a buffet:
If you only have 20 minutes: watch the opening sections on how he picks which model to use and how he phrases everyday questions. That alone changes how you talk to AI.
The sections worth slowing down for: using AI with your own files and data, having it run code and analyze things for you, and the voice and image workflows most people never touch.
What to steal: notice how SHORT his prompts are. He doesn't write essays. He asks, looks at the answer, and asks again. The skill is the back-and-forth, not the opening prompt.
What to skip: anything about specific model versions. The models he demos have been replaced twice over. The behavior hasn't.
Why this works on your brain
Mariah's point in the video is the real lesson: in college she thought founders were a different breed, until she worked with them at Meta and saw they weren't. Watching Karpathy work is the same exposure therapy for AI. Seeing greatness up close shrinks it to learnable size.
2. Karpathy: Deep Dive into LLMs like ChatGPT → The companion video. Where "How I Use LLMs" shows the practice, this one builds the mental model: what's actually happening when the AI answers, why it hallucinates, why it's great at some things and weirdly bad at others. Watch it when you're ready to understand the machine instead of just driving it.
3. Karpathy: Intro to Large Language Models → The one-hour version for the person in your life who needs the foundation first (or for you, if the deep dive feels heavy). It's the most-recommended LLM explainer on the internet for a reason.
4. How Boris Uses Claude Code → Boris Cherny built Claude Code at Anthropic, and his publicly shared workflow is the same watch-the-master effect for agentic AI: loops, automations, and agents reviewed by other agents. If Karpathy demystifies talking to AI, this demystifies putting AI to work.
Watch order if you're newer: 3, then 1, then 4, then 2. Watch order if you already use AI daily: 1, then 4, and keep 2 for a weekend.
Passive watching fades in a week. Twenty minutes of copying what you watched makes it permanent. This prompt is the bridge.
Copy this prompt into Claude right after you finish any video on the watchlist. It turns the 3 techniques you just saw into personalized practice reps on YOUR actual work, so the video becomes a skill instead of entertainment.
Copy this prompt
I just watched a video of an expert using AI, and I want to turn what I saw into my own skill, today. WHAT I WATCHED [Name the video. Example: "Karpathy's How I Use LLMs, the sections on file uploads and data analysis."] THE 3 TECHNIQUES THAT STOOD OUT TO ME 1. [Describe it in your own words, even roughly. Example: "He asks short questions and iterates instead of writing one big prompt."] 2. [Second technique] 3. [Third technique] MY ACTUAL WORK [2-3 sentences on what you do all day. Example: "I manage social media for a dental clinic: content calendars, replying to reviews, monthly reports."] NOW DO THIS: 1. For each of the 3 techniques, design one practice rep I can do in under 10 minutes USING MY REAL WORK as the material. Not a toy example. My actual tasks. 2. For each rep, give me the exact first message to send you, written out, so I can start immediately. 3. Run rep 1 with me right now. Coach me as we go: if my follow-up messages are weaker than the technique I'm copying, tell me what the expert would have asked instead. 4. At the end, tell me which ONE technique gave me the most leverage for my specific job, and what tomorrow's 10-minute rep should be.
The belief part
Steve Jobs said everything around you was built by people no smarter than you. Watch one of these videos this week, run the practice prompt after, and you'll have proof from your own hands. You're a thousand percent capable of this.
The Only AI Masterclass You Need
If this guide helped, but you’re looking to go deeper, I got you!! My 30-Day Challenge takes you from saving AI tips you never use to actually building with AI, step-by-step.
I show you exactly how I automated two e-commerce brands, my social media, and most of my personal life, then hand you the agents, workflows & systems to do the same. I’m teaching you every single thing I know with one lesson and one build a day.
Join the AI Masterclass →© 2026 Mariah Brunner. All rights reserved.