Services / Vibe Coding Training

Vibe coding training for founders and non-developers.

A working session on your own project. You learn five habits that change what your AI coding tool hands back, from someone who builds with Claude Code and Codex every day and runs ten products in production. Take 90 minutes on your own, or two hours with your small team.

$250 for a 90-minute solo session. $700 for two hours with up to five people. Online, or in person in the SF East Bay.

For people building real apps without a developer background

You built the first version yourself with an AI coding tool, and it mostly works. Then a new feature breaks one that worked yesterday. The tool forgets a decision you made an hour ago. It tells you a bug is fixed, in the same confident tone it uses when it's wrong. You start every session explaining your app from scratch.

None of that means you picked the wrong tool or need to learn to code first. A lot of it comes from how these tools work. The session covers what's going on underneath, then five habits for giving the tool context, planning before it writes code, and checking what it tells you, practiced on your app rather than a toy example.

How the tool actually behaves

We start with a short, plain-English explanation of what's happening when the AI writes code, because it explains a lot of the frustration.

It predicts; it doesn't calculate

A calculator gives the same answer every time. An AI model gives the most likely answer, so the same request can produce different code on a different day. That's normal, and it changes how you ask.

Its memory is limited

Some tools carry notes between sessions, but none of them reliably remember every decision you've made, and very long sessions get less reliable. Decisions you want kept should be written down where the tool will read them, not left in an old chat.

It's confident whether it's right or not

A 1%-likely answer comes out in the same tone as a 99% one. "It works now" from the AI isn't the same as working. You'll learn to check before you believe it.

It doesn't know your users

It can write a lot of code quickly, but it doesn't know who your app is for or which shortcuts you can afford. Those decisions stay with you; the habits below are how you tell the tool what you've decided.

Five habits, practiced on your own project

1

Tell the AI who you are and what you're building

Without context it writes generic code for a generic app. Who the users are, what the app does, and how much code you can read yourself change what it produces.

2

Write your project's context down once

Save it in whatever form your tool supports, such as a project instructions file, so you stop re-explaining your app and the decisions you've already made. We write a first version of it together.

3

Ask for a plan before any code

When a plan of mine said "persist the counter to the database" and I wanted it in memory, the fix was one sentence: "in-memory is fine, single instance, do not persist." Without the plan, the same mistake comes back as a repository class, a migration, and tests to throw away. More in Plan First, Code Second.

4

Ask for options, then push back

The first answer is the safest guess. Ask for three approaches and what's wrong with each one; the useful answer usually turns up after a follow-up question or two.

5

Tell the AI how to work with you

These tools are agreeable by default. One standing instruction to be direct, say when it's unsure, and push back on a bad idea changes the conversation.

Tools, cost, and your users' data

Free plan or paid plan

The free tiers are useful. We cover when it's worth paying, how to set usage limits so a long session doesn't surprise you on the bill, and why to keep a copy of anything tied to a free account.

What not to paste into an AI tool

Your users probably didn't agree to have their data sent to an AI company, and depending on your tool and settings, what you paste may be used to train future models. You leave knowing what stays out of a prompt and which settings to check.

The price

Online anywhere, or in person within the SF East Bay at no extra charge. Outside that area, travel is billed at cost and confirmed before booking.

$250

Solo Session

90 minutes, just you, online or in person.

  • The five habits, worked through on your own project
  • A first version of your project's context file, which you keep
  • A one-page prompt sheet
  • A 15-minute check-in a couple of weeks later
  • Questions come to me directly for 30 days
  • A full refund if you don't leave with habits you can use that day
$700

Small Team Session

Two hours, online or on-site, for up to five people.

  • The five habits, worked through on your team's real project
  • A shared project context file in your repository, which you keep
  • Plain-English notes on what's safe to put into an AI tool
  • Questions come to me directly for 30 days
  • A full refund if your team doesn't leave with habits they can use that day

Questions

Do I need to know how to code?

No. The session is built for people vibe coding with no coding experience, and for technical founders who want better results. I'll pitch it to where you are.

Is this Claude Code training or Cursor training?

Either. The habits apply to Claude Code, Cursor, Codex, and the browser-based builders, and we practice them in the tool you already use. Tell me which one when you book so I can confirm the session fits.

Will this make my app secure?

No. Training changes how you work from here on; it doesn't check the code you already have. For a specific question about existing code, a Focused Code Review gives you written findings at $150 an hour, capped at $600. Neither one certifies a whole app as secure.

What do I need to bring?

Your project, your AI coding tool, and something you're trying to build or fix this week. Don't send passwords or access keys.

Tell me what you're building.

A short call about your project and where the AI keeps letting you down. I'll tell you whether a session will help before we hang up.

Book a free 30-minute call