The Daily AI Homework

When people ask how I started learning to build with AI, they expect a course. The truth is smaller and more boring than that. I gave myself homework, and I did it almost every day.

This is the version of that homework I would hand to someone starting today.

Find space, not time

Here is the reframe that made it stick. Most people wait for a free afternoon to "finally sit down and learn AI." That afternoon never comes. So you keep not starting.

The move is to stop looking for time and start looking for space. The commute. The dishes. The walk. The ten minutes before a meeting starts. You stack the learning onto time you are already spending. Done that way, the homework does not compete with your life. It rides along with it.

How it actually started

Calling it homework makes it sound more disciplined than it actually was. I did not sit down one day and assign myself a syllabus. I just started consuming everything I could find about AI, on LinkedIn, on X, on YouTube. And a pattern showed up on its own: I was connecting dots.

Every time I understood a new concept, tried something, or checked out a new tool, another little thread connected inside what had felt like a nebulous web of "how do I even make sense of what AI is, and what it could be?" Those tiny connections started feeding into each other. And honestly, it brought me real joy. Getting to learn and digest all of these new ideas and terms was fun.

Part of what made it work was a mindset: it is okay to experiment. I would try a tool, maybe abandon it, and that was fine. I was not trying to master everything at once. I was slowly building a foundation of comfort with AI, and a vocabulary, so I could keep going. That foundation is the thing that lets you keep learning and evolving, long after any one tutorial fades.

The daily homework

Three small inputs, most days. That is the whole assignment.

1. Listen to one thing. A podcast episode while you do something else. This is the purest find-space-not-time move: you are not carving out a study hour, you are just changing what is in your ears on the way somewhere.
2. Watch one thing. One YouTube tutorial or build-along. Short is fine. The goal is to see how someone actually does it, not to finish a curriculum.
3. Collect five to ten things. Bookmark the posts on X that make you stop and think. You are not reading them all right now. You are building a library you will mine later.

When I started, my version was louder: two podcasts, a stack of YouTube, and at least four hours building. That was the aspirational version, and honestly it was not sustainable as a daily habit. The version above is the one that actually survives a busy week. Consistency beats intensity. A small assignment you keep doing teaches you more than a big one you abandon.

Who I learn from

A starting list of the creators whose work shaped my homework. Do not follow all of them. Pick a few whose voice fits how you think, and let the rest go.

• Andrej Karpathy — first-principles explainers on how LLMs actually work. (Already on my "worth your attention" shelf.)
• Boris Cherny — Claude Code and building with coding agents.
• Peter Steinberger — agentic engineering and "loop engineering." Sharp, hands-on takes on getting coding agents to do real work, and the shift from prompting to designing the loop.
• Greg Isenberg — building AI products and internet businesses in public.
• Alex Finn — AI tools and workflows for non-engineers.
• Nate Herk — agents and automation, hands-on.
• Lenny's Podcast — product, growth, and how building actually works.
• Aakash Gupta — product thinking in an AI world.
• David Ondrej — practical AI building and agents.
• Deborah Liu — product and leadership wisdom from a 20-year tech exec (former CEO of Ancestry, built Facebook Marketplace). Her "Perspectives" newsletter is clear-eyed on going from AI-curious to AI-native without the FOMO.
• Liza Adams — applied AI for go-to-market and marketing teams. Her "Practical AI in Go-to-Market" newsletter is full of real human-and-AI playbooks. People-first, AI-forward.

The point of a list like this is not to follow gurus. It is to surround yourself with people who are doing the work out loud, so the noise starts sorting itself into signal. Build your own list, and keep an eye on whose voices are in it.

Make it stick

A few things that kept the habit alive for me:

• Stack it onto something you already do. Coffee, commute, the dog walk. The habit borrows an existing trigger instead of needing a new one.
• Keep one bookmarks folder. Everything you save goes in the same place. Revisit it once a week. That weekly pass is where the patterns show up: the same idea from three different people, the tool everyone keeps mentioning.
• Aim for most days, not every day. A streak you can actually keep beats a perfect week you quit. Missing a day is not failure. Missing a week is just a signal to make the assignment smaller.

The point of the homework

You are not doing this to become an engineer. You are building enough fluency to start building from your own context: your work, your problems, your judgment. The homework is the on-ramp. The building is the destination, and it comes sooner than you think. Once you start making things, the learning loop gets fast and the homework starts paying for itself.

That is the whole secret. Not a course. Not a free afternoon. Just a small assignment, most days, fit into the space you already have.

You are not behind

If you feel behind, I want to gently put that down for you. Everyone in this space is figuring it out in real time, including the people who look like they already have it all worked out. There is no line you were supposed to have crossed by now.

This is a journey, not a race. Some days you will take one tiny baby step. Other days you will feel like you are running. Both are the work. Both count.

I am here to support you. Welcome to the fun.