How I Learn in the AI Era

My personal framework for developing AI skills and staying ahead of the curve.

Feb 12, 2025

Not long ago, I was chatting with one of Asana’s solutions engineering leaders. We were discussing the rapid evolution of AI and how it’s reshaping work at a lightning pace. As I walked him through some of my current AI projects and experiments, he asked a simple question that I’ve been thinking about ever since: “How did you learn all of this?”

The more I thought about this question and my learning journey to this point, the more I realized his question represented a bigger challenge facing all of us today: How deeply do we want to learn about AI, and how do we structure that learning so we don’t get left behind?

In this post, I’ll share my personal framework for learning AI in a way that aligns with your interests and goals. Use it as a starting point to craft your own approach, whether you’re looking to master prompt engineering or dive all the way into building and deploying large-scale AI systems.

The Great AI Learning Paradox

We live in extraordinary times. AI can now break down complex concepts, help us acquire new skills, and even automate its own creation. Yet the speed of AI’s advancement means there’s more to master than ever.

This is the paradox: AI makes learning easier while also making continuous learning more critical.

It’s worth pausing to reflect on that. The very tools that can guide you through complex material are the same ones transforming industries, rewriting job requirements, and generating entire new types of work. So the question isn’t whether to learn AI—it’s how deeply you want to immerse yourself.

A Fundamental Shift in How We Process Information

To understand why AI literacy is so crucial, let’s zoom out. Humans are unique in our capacity to imagine and convey abstract ideas, whether they’re about planets light years away or intricate mathematical constructs. Civilization has always pushed for new ways to spread knowledge faster, more widely, and at lower cost.

Now, we’re on the cusp of a technological sea change. Our core idea of “basic computing” is evolving into something more powerful than anything we’ve seen in the past 50 years. Here’s the key point: these new capabilities don’t replace what came before; they’re layered on top of past breakthroughs, extending how we interact with technology.

The Cell Phone Analogy

A simple analogy clarifies why timing and attitude matter. Think of a grandparent who avoided smartphones for years, only to finally embrace them and discover they had a steep learning curve. Contrast that with a child who intuitively starts experimenting with a new phone from day one, quickly mastering features you never even knew existed.

The difference often isn’t age; it’s mindset and adoption timing. Those who started exploring smartphones early gained experience incrementally with each update.

Those who held off found themselves suddenly overwhelmed by a mountain of unfamiliar functionality.

Now project yourself 5 or 10 years into the future, when AI is dramatically more advanced and woven into every aspect of work and life. Do you want to be the late adopter, scrambling just to keep pace? Or the person others rely on for guidance as you all navigate this new landscape?

The Four Layers of AI Understanding

Over time, I’ve identified four layers of AI knowledge—like a pyramid, where each layer adds depth and functionality to the one beneath it. You can stop at any layer that suits your goals, but each additional step unlocks more insight.

Layer 1: Application of AI

This foundational layer is where everyone should begin. It’s about:

Working with different interfaces (Chat, Automation, Agents)
Mastering basic prompt engineering
Recognizing high-value use cases
Getting comfortable with data configuration basics

Think of it as learning to drive a car without worrying about how the engine works. You can become extremely effective at this level. In fact, many people will find that a solid grasp of AI applications is all they really need.

How to Build This Knowledge
The best approach is to use AI for everything—even tasks you suspect won’t benefit from it. Try drafting emails, analyzing documents, generating code, creating images, or summarizing complex research. Such hands-on experimentation is the only way to grasp the frontier of AI’s capabilities, which keeps shifting almost daily.

If you prefer more structured learning, here are resources I trust:

Asana's AI Certification (free) - An excellent starting point covering practical applications and prompt engineering fundamentals: AI for Work Skill Badge
Similar programs are offered by other companies, though I can personally vouch for Asana's quality.

Conclusion for Layer 1

Even if you never go beyond this layer, building comfort with AI interfaces and prompt engineering is likely to deliver significant returns for your productivity and professional growth.

Layer 2: Building Large Language Models

This layer helps you understand how the "engine" works. It covers:

How models are trained
Different types of post-training methods
Model capabilities and limitations

Why does this matter? Understanding how LLMs work helps you:

Better predict when they'll succeed or fail
Know which models to use for different tasks
Understand the limitations and potential risks

For those interested in diving deeper, I highly recommend these resources:

Andrej Karpathy's "Intro to Large Language Models" - A fantastic one-hour overview that makes complex concepts accessible
His "Building LLMs from Scratch" series - More technical but incredibly enlightening
His recent 3.5-hour comprehensive LLM overview
Stanford CS 229's LLM lecture - A slightly gentler introduction that still covers the key concepts
Heads up - some of these can get very technical. Try to use AI to explain the topics you don’t understand.

Conclusion for Layer 2

By grasping how large language models operate, you’ll not only boost your AI “intuition” but also make better decisions about which tools and techniques align with your goals.

Layer 3: Technical Infrastructure

Think of this layer as understanding the "factory" where AI is built and operated. Just like you don't need to understand how a car factory works to drive a car, you don't need this knowledge to use AI. But understanding it helps you grasp why some things are possible with AI while others aren't, and where the technology is heading.

This layer explores:

The Physical Foundation
- Specialized AI chips (like those from NVIDIA) that power AI
- Massive datacenters where models are trained and run
- Why training ChatGPT required more computing power than putting humans on the moon
The Economics
- Why training large models costs hundreds of millions of dollars and soon billions
- How cloud computing made AI quickly accessible to everyone
- Why some AI features are expensive while others are nearly free
The Limits and Possibilities
- What Moore's Law means for AI development
- Why bigger models aren't always better
- How infrastructure constraints shape what's possible

A favorite resource on this topic is The NVIDIA Way, a book that delves into how modern GPU technology catalyzed today’s AI revolution. This level of understanding clarifies why some feats are suddenly possible and where the hardware bottlenecks still lie.

Conclusion for Layer 3

You don’t need to master data center architecture to benefit from AI. But learning these fundamentals offers an invaluable macro perspective on how AI came to be and which breakthroughs might lie ahead.

Layer 4: Historical Context & Future Implications

The deepest layer covers:

Evolution of information processing technology from the creation of language to today
Key historical breakthroughs and their significance
Future trajectories

Understanding this historical context is crucial because it helps you:

See patterns in technological development
Better predict future trends
Understand why certain approaches succeed or fail

I highly recommend "The Information" (thanks BenGG for sharing this with me) for a comprehensive look at how information technology has evolved throughout human history.

Conclusion for Layer 4

While few roles require a historian’s view of AI, having this background can sharpen your strategic vision and help you predict the field’s trajectory more confidently.

Here's the crucial point: you don't need to master all layers to be effective with AI. Most people will find tremendous value just focusing on Layer 1. But understanding that these layers exist helps you make an intentional choice about how deep you want to go.

Each layer builds on the previous ones:

Layer 1 teaches you what AI can do
Layer 2 helps you understand how it does it
Layer 3 shows you what makes it possible
Layer 4 helps you understand where it's going

Choose your depth based on your interests and goals, but start with Layer 1 and build from there.

My Personal Learning Methods

Start with Real Problems

The fastest way to learn is by doing. When I helped manage large enterprise accounts, I faced a bottleneck in handling an overwhelming volume of communications. My solution? Experiment with prompt engineering and automation. I wasn’t learning AI in a vacuum; I was learning it to solve a pain point.

Look for Problems That Are:

Meaningful enough to spark motivation
Well-defined so you can address them in steps
Measurable, so you can confirm whether your AI solution is working

My first AI project involved researching more efficiently. I didn’t begin by memorizing the ins and outs of GPU architecture; I simply asked, “How can I make sense of this data faster?” That question anchored my learning, ensuring everything I picked up was immediately put into practice.

Build in Public (or Semi-Public)

You’re here reading my Substack because I believe in the power of openly sharing within communities. Writing and teaching force you to articulate new knowledge in a way that’s easy to grasp. But you don’t need your own newsletter to do this:

Document experiments in your internal company wiki
Post short findings on Slack channels dedicated to innovation
Offer to mentor a junior team member who’s curious about AI
Participate in online + in-person AI communities

At Asana, I quickly started sharing everything I was learning about building with AI. That documentation eventually snowballed into our company-wide AI education program. You never know where your notes and shared experiments might lead.

Create Feedback Loops

Another advantage of AI is the near-instant feedback. Traditional software development often requires lengthy compile times or complicated deployment steps. With AI, you can tweak a prompt or re-run an inference and see results in seconds.

My Iterative Process Generally Looks Like This:

Draft a prompt or prototype a small solution
Test it right away
Note what worked and what didn’t
Share it briefly for peer or user feedback
Adjust and repeat

This rapid cycle ensures your learning never stalls. Each tiny iteration makes you more fluent in AI’s capabilities and limitations.

Focus on First Principles

Though AI evolves rapidly (see how many new models launched in 2024 alone), the bedrock principles stay relatively stable. Whenever I try a new tool, I look for:

Why it works instead of just how
Core design patterns that appear in different AI systems
The fundamental problem this tool aims to solve

This perspective streamlines your learning across new AI releases. You won’t need to start from scratch when a next-generation model/agent/application arrives; you’ll already recognize shared concepts.

Making Your Learning Plan

Now that you’ve seen the layers and practical tips, here’s a simple plan to structure your own AI education:

Step 1: Choose Your Depth

Review the four layers and decide how far you want to go. There’s no “one-size-fits-all” answer. Reflect on:

Your career aspirations
Available time and energy
Personal curiosity about AI’s mechanics

Step 2: Start with Layer 1

Regardless of your ultimate goal, you should build a solid foundation in AI applications:

Learn how to prompt effectively
Explore different AI interfaces
Practice identifying promising use cases
Understand basic data configurations

Step 3: Find Your Learning Style

Pick a method—or blend of methods—that resonates with you:

Project-Based: Learn by creating real solutions
Structured Learning: Enroll in courses or follow tutorials
Exploration: Experiment, make mistakes, learn through spontaneous play

Don’t hesitate to mix and match. The key is staying motivated and consistent.

Staying Motivated and Avoiding Overwhelm

With AI’s unrelenting pace, it’s easy to feel like you’re perpetually lagging. Here’s how I keep personally keep my sanity:

Double down on fundamentals before chasing trends
Deep dive into only a handful of areas instead of dabbling in everything
Accept that nobody is across all AI developments
Celebrate incremental wins, whether that’s building a small chatbot or improving a single workflow

Remember: the objective isn’t to know it all—it’s to have a purposeful path that aligns with your needs and interests.

Looking Ahead

As we push deeper into 2025 and beyond, AI literacy will become an even more crucial asset in just about every field. The best part is you can control how involved you become. By steadily incorporating AI skills into your day-to-day, you’ll build knowledge incrementally, rather than suddenly feeling you have to catch up on five years of change overnight.

Which brings us back to that original question: How are you going to learn all of this? By choosing your depth, focusing on real-world problems, sharing what you discover, iterating quickly, and leaning on first principles. The earlier you start, the more natural your growth will feel—and the more prepared you’ll be when AI takes its next big leap.

So, What’s Next?

Decide how deep you want to go in your AI exploration.
Pick a real problem—big or small—and tackle it using AI.
Dive into the basics of prompt engineering if you haven’t already.
Document your journey in a place where others can learn from it.
Share and collaborate—even if it’s just with a small group, your future self will thank you.

You don’t have to become an AI expert this week or even this year. But taking an intentional step today will pay exponential dividends tomorrow. Wherever you are on your own path—whether just discovering AI’s possibilities or already knee-deep in advanced model training—I’d love to hear your story.

Drop a comment below on how deep you’re gonna take your AI learning journey. I’ll be back in two weeks with more insights on AI applications and practical tips you can use right away.

Information Processed

Discussion about this post