Skip to content

For you · B2C

You chat with AI, and hope it is right. Now build something you can prove.

Every day you wonder how much of what AI tells you is just confident nonsense. In one afternoon, in a small room, you stop pulling the slot-machine lever and learn the craft, then walk out holding work you would hand a client without flinching.

A brief you would send a client Slides & a one-page site A workflow that does the legwork
Tell us what you want to build See the three sessions
Four to eight people, one room in Ho Chi Minh City. Not a video you watch alone and abandon.
hope
proof

You leave with

Something real

Why it works

Hope is not a method. This is.

Guessing has a cost: the confident wrong answer you shipped, the deck you cannot reproduce next month, the hours lost re-prompting in circles. A free video shows you a trick. It cannot sit beside you while you build the real thing. The fix is a handful of habits, swapped in for the reflexes most people never question.

  1. Trust the confident answer and ship it.

    Source and verify

    Every claim traced to something checkable, or cut. You learn to catch the confident wrong answers most people ship.

  2. Re-prompt in circles when it misses.

    Edit the prompt, not the output

    You fix the instruction, not the result. The fix sticks, and it works again tomorrow.

  3. Hand it the keys and hope.

    Automate the reversible, gate the rest

    It runs the safe, repeatable work. Anything you cannot undo (send, pay, post, delete) waits for your sign-off.

sourced · re-runnable Habits that outlive whatever app is hot next year.
What you walk out with

Proof you can hold.

Not slideware about AI. Real work you make yourself: sourced, re-runnable, and good enough to ship the next morning. Yours to keep using long after the room empties.

  • A sourced document

    Every claim checkable, plus the re-runnable prompt that regenerates it. The kind of brief you would put your name on.

  • Slides, PDF, a one-page site

    Write the substance once, render it many ways, all in your own mini design system, so everything looks and reads of a piece.

  • A workflow that does the legwork

    A small automation you can rely on. It drafts and prepares, but never sends, pays or deletes without your sign-off.

Optional annex

Images, done responsibly

An optional creative annex: generate visuals for real work, with rights and provenance kept straight and AI use disclosed.

The three sessions

Three sessions, one climb.

Take one on its own, or follow all three. Each session sends you home with a real deliverable, and your stack grows as you climb.

  1. 01
    Session 1 · Reliability

    The reliability method

    Stop taking the first answer on faith. Learn to source, verify and edit the prompt, so what you make actually holds up.

    You leave with: A sourced document on your own subject, every claim checkable or cut, plus the re-runnable prompt that regenerates it.

    • How AI really works, and where it fails
    • Source and verify every claim
    • Edit the prompt, not the output
    • Keep your prompts, reuse them

    Your growing stack

  2. 02
    Session 2 · Craft

    Deliverables that hold up

    Write the substance once, render it many ways. Then drive a real assistant workspace with one source connected and supervised.

    You leave with: A polished, sourced deliverable, regenerable from your own project and prompt, with one connector you wired on purpose and keep under your control.

    • Write once, render many
    • Your own mini design system
    • Polished docs, slides, PDF and one-page sites
    • Connect a source, supervised

    Your growing stack

  3. 03
    Session 3 · Autonomy

    Projects that run, safely

    Give your AI a workspace it can find its way around, automate one task that repeats with you signing off on anything you cannot undo, and lock in one habit that lasts.

    You leave with: A structured personal workspace, one recurring automation with a human checkpoint on every irreversible step, and a habit plan you'll actually keep.

    • A workspace your assistant can navigate
    • Automate the reversible, gate the rest
    • Make one habit actually stick

    Going further: scope a small piece of real work, run it in a sandbox, then review the result, with agentic tools like Codex as one example. The discipline, not the tool, and no lock-in (optional).

    Your growing stack

Sessions 2 and 3 each open with a quick recap of the method, so a standalone afternoon still stands on its own.

A small hands-on workshop, people gathered around laptops in a warm room
In the room

Think, produce, critique.

No lecture. Four to eight people, about three and a half hours, and a real subject of yours. We show the gesture, you make it your own, the room sharpens it. You leave having done the thing, not just heard about it.

  1. 01 Think

    Decide what good looks like before you prompt, and what is worth delegating.

  2. 02 Produce

    Make the real deliverable on your own subject, hands on the keyboard.

  3. 03 Critique

    Honest peer and self checks: sourced, verified, reproducible. Then you keep it.

Partner, not professor. Prefer to keep your work private? Run the whole session on a neutral shared case and still leave with a complete deliverable.

How to start

One afternoon, or the whole climb.

1 of 3
Standalone

Take one session

One afternoon, hands on the keyboard, and you walk out with something finished. Start at session 2 or 3 if you like, each one opens by catching you up.

All 3
Full programme

Or follow the path

Stacked together, the three sessions build: the reliability method, deliverables that hold up, then projects that run on their own under your watch.

Tell us what you want to build

Your move

Stop hoping. Start building.

Tell us where you are and what you want to make. No price, no pressure: we run small in-person groups in Ho Chi Minh City and reply personally to help you pick your starting point.

Tell us what you want to build See the sessions