AI Newsletter #5: How to actually learn AI
Hello everyone,
Right now, 50% of the global workforce has yet to use AI meaningfully. That means the window to become an industry leader is wide open, but it won't stay that way for long.
It’s incredibly easy to fall into the trap of comfortable patterns and old workflows. But at Oak Drive, we don't just want to talk about AI; we want to build with it. To stay ahead of the curve, we have to get out of our comfort zone and experiment with tools and workflows were not currently using.
To support this, I’m increasing our newsletter frequency. The focus is shifting entirely to instructional content. Every issue will feature an AI Practitioner who is building repeatable systems. I encourage you to watch their videos, study their workflows, and take the initiative to apply these skills to your daily tasks.I'm also encouraging everyone to move beyond standard AI web chats and learn more powerful AI dev or terminal tools. Later in this email, I’ll give you instructions to install Claude Code which is the typical power user setup. But before we get into that, let’s look at someone already setting the standard.
This Week's AI Spotlight: Tina Huang
This week, we’re looking at Tina Huang, an ex-Meta Data Scientist with a background in pharmacology who has pivoted to teaching professionals how to stop "blindly prompting".
The Philosophy: Systems > Tool Lists
Tina treats AI as an operational system grounded in data analysis and coding fundamentals. Her core principles are built for the structured professional who wants career leverage:
- Agentic Workflows > Direct Prompting: Don't ask AI to do an entire job in one prompt. Break tasks into discrete steps—like research, drafting, and revising—as if you were managing an intern.
- Context is the Engine: AI only knows what you feed it. She relies on the T.C.R.E.I. framework (Task, Context, References, Evaluate, Iterate).
- Breadth-First Execution: Build a messy, working prototype first. She advocates for "vibe coding" simple automations immediately rather than waiting to finish a 40-hour technical course.
Golden Workflows
- Research Synthesis: Dumping unstructured resources (manuals, voice notes, PDFs) into NotebookLM to generate structured outlines and audio overviews for passive learning.
- The Skeptic Data Analyst: Uploading a strategy document or presentation and using a "Skeptic Prompt" to find flaws in assumptions and generate the 10 hardest questions you’ll face in a meeting.
- Vibe Coding: Using AI coding agents like Claude Code to build local Python scripts that automate repetitive administrative tasks that require zero creative thought.
My Take on Tina Huang
What I appreciate about Tina is that she is practical and data-centric. She focuses on tools applied to daily tasks—emails, reports, and learning—rather than "get rich quick" hype. Her method for synthesizing unstructured data into clean formats is exactly what we need as we build out our own organizational context engines.
- Research Synthesis: 100% anytime you need to learn something NotebookLM should be the first tool you're reaching for. I use it all the time and even used it to distill Tina's YouTube channel into this summary deck. I scraped her YouTube channel for the latest 100 video links and pasted in NotebookLM then I asked Gemini for a prompt to generate the deck.
- The Skeptic Data Analyst: This one is also in my toolkit. Anytime you're brainstorming a proposal, or preparing a presentation to your boss. Ask the AI to find weaknesses or gaps in your logic, then revise based on the feedback (do it multiple times until you're bulletproof)
- Vibe Coding: The strongest skill by far! Because you can develop custom apps with AI.
Personalized AI Consultations
I am opening up my Calendly for small groups, teams or departments that prefer more personalized lessons. The lessons will involve using Claude ecosystem especially Claude Code.