Software Engineer — zero to hero

From "AI autocomplete is neat" to shipping features with an AI pair — and knowing when not to trust it.

You're a hero when…

AI drafts your boilerplate, reviews your diffs, and explains legacy code — while you architect, verify, and ship an agent-assisted side project.

13 steps · 📖 read a guide · 🛠️ try a tool · 💪 do a real mission (with a copyable prompt)

0 of 13 done

1 Foundations

Step 1 📖 Read

What Is AI, Actually? →

Knowing it's a next-token predictor explains every weird thing it does with your code — including inventing APIs.
Step 2 📖 Read

Prompting Basics →

The difference between "fix this" and a great coding prompt is the difference between garbage and a mergeable diff.
Step 3 📖 Read

Why AI Makes Things Up →

Models invent package names, flags, and function signatures with total confidence. Verifying is a core engineering skill now.
Step 4 🛠️ Try

Prompt Grader →

Run your last real coding prompt through it — most engineers fail "context" on the first try.

Step 5 💪 Do

Explain a legacy file

The fastest daily win: stop reverse-engineering old code by eye.
Show the mission prompt
```
Here's a file from our codebase. Explain what it does top-down, list any hidden gotchas or side effects, point out the riskiest part to change, and suggest 3 unit tests that would catch regressions. [paste file]
```
Step 6 💪 Do

Debug out loud

Paste the error AND the context — the model is a rubber duck that talks back.
Show the mission prompt
```
I'm getting this error: [paste full error + stack trace]. Here's the relevant code: [paste]. Here's what I've already tried: [list]. Walk through the 3 most likely causes in order of probability, and give me one cheap check to confirm or rule out each before we change any code.
```
Step 7 💪 Do

Pre-review your own PR

Catch what the reviewer would — before the reviewer does.
Show the mission prompt
```
Review this diff like a senior engineer who cares about correctness first, style second. Report every issue you find, including ones you're unsure about, each with severity and confidence. Then list anything that needs a test it doesn't have. [paste diff]
```
Step 8 🛠️ Try

Prompt A/B Lab →

Take your debug prompt and a lazier version, and see exactly which parts of the good one earn their keep.

Step 9 🛠️ Try

System Prompt Architect →

Build a standing assistant that knows your stack, style guide, and testing conventions — stop re-explaining your codebase every chat.
Step 10 📖 Read

Vibecoding →

The full workflow for building features by describing intent — including how to review AI code without drowning.
Step 11 🛠️ Try

API Cost Calculator →

The moment you build ON the API instead of chatting, cost per token becomes your problem. Model it before your first invoice does.

Step 12 📖 Read

Loop Engineering →

Agents — plan, act, observe, verify — are the biggest change to software work since CI. Learn to build loops that stay safe.
Step 13 🛠️ Try

App Architect →

Capstone: blueprint a side project end to end, then hand the master prompt to your AI and actually ship it.