Free guide · no gatekeeping
The AI Field Guide.
The no-jargon, no-slop guide to actually building with AI — for total beginners. Get your first real win in 15 minutes, steal my copy-paste prompts, and finally understand what DNS, an API and MCP actually are. No tech-bro energy. Just the real thing, in plain English.
Want it in your inbox + a short letter most weeks?
Part 1
Your first win, in 15 minutes.
You don't need to understand AI to start using it. You need a setup and a first conversation. Here's the exact path — no code.
- 1Sign up for an AI coding tool — Claude Code ($20/mo) or OpenAI Codex (free tier). You talk, it builds.
- 2Install Superwhisper so you can speak to AI anywhere on your computer.
- 3Make a folder called "workspace" — this is where your projects live.
- 4Create a free GitHub account. When Claude asks to connect, just follow the prompts.
- 5Open Claude, point it at your workspace folder. That's it — you're connected.
- 6Say: "Interview me until we're both 100% clear on what I'm building." Answer honestly, then say "let's build it."
Want the fuller version with links + screenshots? The complete Start Here walkthrough →
Part 2
Talk to AI properly.
The gap isn't intelligence — it's how you ask. Here are six prompts I actually use. Copy them, swap the brackets for your stuff, paste them in.
Interview me before you build
Stops AI building the wrong thing. It drills you until the brief is clear, then builds.
Interview me until we're both 100% clear on what I'm building. Ask me one question at a time about the goal, who it's for, what's in and out of scope, and what done looks like. Don't start building until I say "yes, that's it."
Sound like me, not like a robot
The anti-slop move. Teach the AI your voice so it amplifies you instead of flattening you.
Here are 3 things I've written: [paste them]. Study my voice — sentence length, rhythm, the words I use and avoid, how formal I am. Now write [the thing] in MY voice. No clichés, no "in today's fast-paced world", no em-dash filler. If it sounds like generic AI, you've failed.
Turn a rambling voice note into tasks
Walk, talk, paste. It pulls the actions out of the mess.
This is a voice note I recorded while walking: [paste transcript]. Pull out every actual task or decision. Return a clean checklist, each item starting with a verb, grouped by project, with anything time-sensitive flagged. Ignore the waffle.
The 6-part prompt template
The anatomy of a prompt that actually works. Fill the blanks.
ROLE: You are [an expert in X]. TASK: [The one thing I want done.] CONTEXT: [What you need to know — audience, constraints, background.] REASONING: Think step by step before you answer. OUTPUT: [Exact format — a table / 5 bullets / an email.] STOP: If anything's unclear, ask me before you start.
Build me a simple page
Your first 'vibe coded' thing. Plain English in, real website out.
Build me a one-page website for [what it is]. It needs [sections]. The vibe is [warm / minimal / bold]. Use my words: [paste]. Walk me through it as you go, in plain English, and tell me exactly how to put it live.
Summarise and decide
Turn a pile of information into a recommendation, not just a summary.
Here's everything I know about [decision]: [paste]. Summarise the key trade-offs in a short table, then give me your actual recommendation and the one reason it wins. Tell me what would change your mind.
Part 3
Real workflows you can copy.
These aren't demos. They're the systems I run every day. Start with one.
Part 4
The jargon, demystified.
"What the f$%k is DNS? API keys? Webhooks?" — this was me a couple of years ago. Here's the whole vocabulary, in one breath each.
- Vibe coding
- Telling AI what you want in plain English and letting it build. You don't write code — you describe the outcome.
- LLM
- The AI model itself (ChatGPT, Claude). Think of it as a very well-read assistant that predicts what to say next.
- Prompt
- What you type or say to the AI. A better prompt = a better answer.
- Token
- How AI counts text — roughly ¾ of a word. Your plan limits are measured in tokens.
- Claude Code
- A tool that lets AI work inside your real files and projects, not just a chat box. Where the leverage lives.
- Terminal / CLI
- The "black screen" where you type commands. You don't need to learn it — Claude drives it for you.
- Markdown
- Plain text with light formatting (# for a heading, - for a list). The simplest way to keep AI-friendly notes.
- GitHub
- Where your work is saved and versioned online — a backup that AI can also read and build on.
- Repo
- One project's folder of files, tracked so you can see every change and roll back.
- API
- A way for two apps to talk to each other automatically — e.g. your site telling Stripe to take a payment.
- API key
- A password that lets one app use another's API. Keep it secret, never paste it in public.
- MCP
- A standard that lets AI plug into your real tools — calendar, database, Shopify — and actually do things.
- Webhook
- An automatic "ping" one app sends another the moment something happens (a sale, a signup).
- DNS
- The internet's address book — it turns a domain name (yoursite.com) into the server it lives on.
- Environment variable
- A setting or secret (like an API key) stored outside your code so it never leaks.
- Deploy
- To publish your site or app so it's live on the internet for everyone.
Part 5
The stack I actually use.
No affiliate links, no fluff. The tools that run a farm and five brands — most on free tiers.
What next
You've got everything you need to start.
Pick one prompt. Run it today. Then, if you want company on the journey: a free monthly call, a short letter most weeks, and — when you're ready — a guided build in a small group.
Get the practical stuff in your inbox, most weeks.
Free · No gatekeeping · Unsubscribe in one click