Where this course goes (and where you fit) — step 1 of 9
The course is a ramp, not a journey
There are 30 chapters after this one. They are not all the same shape. Some are pure syntax. Some are pure judgment. A few are deep technical sections that will cost you a weekend each. The honest read is that this course gets steeper as it goes, and the steepness is the point — by chapter 30 you are doing something that, in 2026, almost nobody knows how to do.
Here is the whole map, in five levels.
Level 1 — Python that doesn't lie to you (ch01-11)
Variables. Functions. Lists, dicts, loops, conditionals. Tracebacks. Mutation. Modules. Errors. Files. Classes.
This is the layer most people quit on, because they expect it to be exciting and it isn't. It's plumbing. You're learning the language the rest of the course is written in. By the end of ch11 you can read a 30-line Python program and know what it does — including the parts that fail.
The win at this level: when an LLM ships you code and tells you it works, you can actually check.
Level 2 — Talking to models (ch12-15)
HTTP and APIs. LLM APIs. Structured output. MCP.
This is where Python stops being abstract. You make your first real API call. You see the request, the response, the cost, the latency. You learn the messages pattern that every agent is built on top of. By the end of ch15 you have written code that sent text to a model and got text back, on purpose, for a reason.
The win at this level: you stop using ChatGPT.com and start using the underlying machine.
Level 3 — Loops, prompting, evals (ch16-21)
Agent loops. Git. Secrets. Prompting. Agent traces. Evals.
This is the level where you stop using AI and start building WITH AI. You write a loop that asks a model, reads the answer, decides what to do next, asks again. You write prompts as code, version them, test them. You learn to grade the model's outputs — which is the actual job in 2026.
The win at this level: you become someone who can ship an agent and tell whether it works.
Level 4 — Production realities (ch22-25)
Retrieval. Production tradeoffs. Debugging. Capstone.
This is where the demo dies and the product begins. You learn what breaks when an agent leaves your laptop — cost spikes, hallucinations in production, traceback that doesn't reach you because the agent swallowed the error. The capstone is a real thing you ship.
The win at this level: you build something a friend could actually use, and you know its failure modes.
Level 5 — Harness engineering (ch26-30)
Agent harnesses. AI image gen. AI video gen. Programmatic design. Harness engineering.
This is where you stop competing with displaced juniors and start out-shipping them. A "harness" is the discipline of turning a model into an agent — the scaffolding around the model that makes it reliable, observable, and ownable. By ch30, you are someone who could be hired to build the agent that does the job you used to have.
The win at this level: you are the scarce person now.
The honest read
This is not a journey. It is a ramp. It is not equally steep the whole way — ch01 is gentler than ch16 is gentler than ch26. You will struggle in places. You will feel slow. You will hit a chapter and think "I cannot do this," and then you will do it.
The students who finish are not the smartest ones. They are the ones who kept showing up after the lessons stopped feeling fun.
This lesson is the map. The next four lessons in this lesson are about which path through the map you should take, what realistic timelines look like, and where to be careful.
Where this course goes (and where you fit) — step 1 of 9
The course is a ramp, not a journey
There are 30 chapters after this one. They are not all the same shape. Some are pure syntax. Some are pure judgment. A few are deep technical sections that will cost you a weekend each. The honest read is that this course gets steeper as it goes, and the steepness is the point — by chapter 30 you are doing something that, in 2026, almost nobody knows how to do.
Here is the whole map, in five levels.
Level 1 — Python that doesn't lie to you (ch01-11)
Variables. Functions. Lists, dicts, loops, conditionals. Tracebacks. Mutation. Modules. Errors. Files. Classes.
This is the layer most people quit on, because they expect it to be exciting and it isn't. It's plumbing. You're learning the language the rest of the course is written in. By the end of ch11 you can read a 30-line Python program and know what it does — including the parts that fail.
The win at this level: when an LLM ships you code and tells you it works, you can actually check.
Level 2 — Talking to models (ch12-15)
HTTP and APIs. LLM APIs. Structured output. MCP.
This is where Python stops being abstract. You make your first real API call. You see the request, the response, the cost, the latency. You learn the messages pattern that every agent is built on top of. By the end of ch15 you have written code that sent text to a model and got text back, on purpose, for a reason.
The win at this level: you stop using ChatGPT.com and start using the underlying machine.
Level 3 — Loops, prompting, evals (ch16-21)
Agent loops. Git. Secrets. Prompting. Agent traces. Evals.
This is the level where you stop using AI and start building WITH AI. You write a loop that asks a model, reads the answer, decides what to do next, asks again. You write prompts as code, version them, test them. You learn to grade the model's outputs — which is the actual job in 2026.
The win at this level: you become someone who can ship an agent and tell whether it works.
Level 4 — Production realities (ch22-25)
Retrieval. Production tradeoffs. Debugging. Capstone.
This is where the demo dies and the product begins. You learn what breaks when an agent leaves your laptop — cost spikes, hallucinations in production, traceback that doesn't reach you because the agent swallowed the error. The capstone is a real thing you ship.
The win at this level: you build something a friend could actually use, and you know its failure modes.
Level 5 — Harness engineering (ch26-30)
Agent harnesses. AI image gen. AI video gen. Programmatic design. Harness engineering.
This is where you stop competing with displaced juniors and start out-shipping them. A "harness" is the discipline of turning a model into an agent — the scaffolding around the model that makes it reliable, observable, and ownable. By ch30, you are someone who could be hired to build the agent that does the job you used to have.
The win at this level: you are the scarce person now.
The honest read
This is not a journey. It is a ramp. It is not equally steep the whole way — ch01 is gentler than ch16 is gentler than ch26. You will struggle in places. You will feel slow. You will hit a chapter and think "I cannot do this," and then you will do it.
The students who finish are not the smartest ones. They are the ones who kept showing up after the lessons stopped feeling fun.
This lesson is the map. The next four lessons in this lesson are about which path through the map you should take, what realistic timelines look like, and where to be careful.