Before any code

How to Think About AI.

Most conversations about AI start with the technology. This one starts before that: when do you actually need AI, what a working AI system is really made of, what to build yourself versus take from the ecosystem — and why so much of that ecosystem is free. A short, plain-English series on AI solution discovery, pulled from a real discussion.

The mental model

The whole idea, in three moves.

Every episode comes back to these three. Hold them, and the rest — models, weights, build-vs-buy, open source — is just detail on top.

Start with the problem

Real problems never arrive clean — they’re messy. AI solution discovery starts with the problem in front of you, never with the technology you’d like to use.

Interpret, decide, act

A working AI system is three roles: something that interprets the messy input, something that decides, and something that acts. A model is only ever one part of it.

Use the ecosystem

An AI model is a shortcut for a function you can’t write yourself. Someone has usually built most of it already — your value goes on top, not underneath.

Why it's worth understanding

The conversation that comes before any code.

AI is about to unlock a whole new class of real-world problems — problems everybody can see. The next generation of builders won't sit in a room solving problems they assumed; they'll look at the real world, see the problem themselves, and solve it. This series is how to think about AI, models and the ecosystem so you know what to build, what to reuse, and when you don't need AI at all.

The series

How to Think About AI.

Eleven short episodes, each on one idea, in plain English. The Big Picture sets up where AI came from and what a model really is; Solution Discovery walks from a messy real problem to knowing what to build. Open any episode for a one-line summary and the video.

The Big Picture

Parts 1–4

Start here. Where AI came from, why it exists, and what is really inside a model — the ecosystem view before any technology choices.

PartEpisodeWatch
1 Play
2 Play
3 Play
4 Play

Solution Discovery

Parts 5–11

From a messy real-world problem to an AI solution: when you actually need AI, what a working AI system is made of, build-vs-buy, and why the ecosystem gives its models away.

PartEpisodeWatch
5 Play
6 Play
7 Play
8 Play
9 Play
10 Play
11 Play

Watch the whole thing in order.

11 episodes · The Big Picture → Solution Discovery — the full series as one playlist.

Watch the full playlist

A note on the series: this is a simple, high-level take built for intuition, not precision — so some specifics may differ from more formal explanations. That's the point. Get the right picture first; the rigour is easier to add once you have it.

FAQ

Questions this series answers

When do you actually need AI?+

When you know the input and the output but can't write the function that connects them. An AI model is a shortcut for a function you can't derive yourself.

Should you build your own AI model or use existing ones?+

Use the ecosystem's models and add your value on top. Building a large model from zero is a multi-year effort — like building a compiler just to sell samosas.

What is a working AI system actually made of?+

Three roles: something that interprets the messy input, something that decides, and something that acts. A model is only ever one part of the whole system.

Why do AI companies open-source their models?+

No one can build for every use case or language alone. Opening the models grows an ecosystem that does the work for them — the same lesson as Linux and the Windows SDK.