Skip to content

September 1 workshop · Ryan Wanner

How do I turn my working week into useful content?

Collect material you own, explain what it means, and use four context cards to draft one useful answer. Review the facts and voice, save a specific correction, and improve the next run. The workshop includes real setup failures and troubleshooting.

All classes

Watch the real build, including what failed

2:04:44 workshop, including setup and troubleshooting · Watch on YouTube

Every class on this page is free to read and watch. No account needed.

The useful-answer test

Would this help someone who never contacts you?

At 8:53, the class critiques a generated video that asks a question but never answers it. Try that review on one draft before making a week of content.

A small loop you can finish

  1. Collect one answer's ingredients

    Choose a real client question with identifying details removed. Put one permitted source, an owned photo if useful, and a note explaining the intended use in one intake folder. Open each file before relying on it.

  2. Write the four cards

    In self.md, record real voice samples. In guardrails.md, record factual and permission limits. In channel.md, name the intended format. In machine.md, describe the draft, review and revision steps. Read the summary back and correct it.

  3. Make an asset ledger

    List each input, its source, what it establishes, usage permission and unknowns. A photo does not establish every property fact. Keep unverified claims out of the draft.

  4. Answer before adapting

    Draft one complete answer to the question. Underline the sentence that answers it. The recording shows an attractive video failing this exact test. Repair the answer before making more formats.

  5. Review and record one correction

    Check facts, voice, readability and usefulness. Keep the draft in Waiting until the correction is checked. Approved describes a reviewed version; actual publishing remains a separate decision.

  6. Improve the next run

    Put the reusable correction into the relevant card. Rerun the same input, compare both outputs and explain what changed. Adapt a checked answer for two channels with distinct captions.

Four files, four different jobs

self.md

Actual writing samples and what your business does.

guardrails.md

Facts, permissions, unknowns and review limits.

channel.md

The reader, account and useful format.

machine.md

The intake, ledger, draft and correction process.

Keep these as owned files you can reopen. A project, installed skill and scheduled automation are different things.

Listen for the review decisions

It asked what the question was, but it never actually answered that question.
8:53 · An answer must deliver value, not merely state the hook.
the goal is no longer just to get a good output. The goal is to fix the system
12:34 · A recurring error belongs in the reusable instructions.
You would not want to use the same captions for each.
2:02:10 · Keep facts consistent while adapting the format to its destination.
Believing an install message

Reopen the actual project, files and folder. A success sentence does not establish saved instructions, connector access or a working automation.

Generating a batch before an answer

Review one source-based answer first. Multiply a useful explanation after its facts and voice are checked.

Treating silence as approval

Keep an explicit review status for the exact version and destination. A folder name does not authorize posting.

Start with the practice sheet

Create a free student portal account to download this class's files.

  • The intake-to-answer practice sheet

    Current practice companion with fictional inputs, a working request and explicit checks.

  • Guided setup and verification

    Current companion: read, practice and verify one draft before external use.

  • The Four Context Cards

    self, guardrails, channel, and machine, written far enough that you edit rather than start from a blank page. Banned phrases and fair housing rules already filled in.

  • Content Machine Prompt Pack

    The seven station prompts: asset ledger, brand-compare briefs, the still, the faceless video script, the schedule, and the weekly learning pass.

  • Content Machine Agent Skill

    Plain instructions for the intake, draft and review loop. Load the skill and only permitted context into a tool that can read them, verify access, and review each result; file, connector and agent support vary by account.

  • How the Content Machine Folders Work

    The one-page explanation of the four-folder system. Short enough to hand to an assistant without explaining it yourself.

Read the full working sheet here
# Intake to a useful answer

Current replay companion · November 8, 2026. Practice is editorial; no completed attendee outcome is claimed.

## One source packet
Client question, with identifying details removed:
Source file and owner:
Permission and intended use:
Verified facts:
Unknown facts:

## Four cards
self.md — two writing samples and phrases you use/avoid:
guardrails.md — facts, rights and review limits:
channel.md — audience, format and intended account:
machine.md — intake, ledger, draft, review and correction:

## Ledger
File | What it establishes | Permission | Unknowns

## Answer check
Question:
Sentence that answers it:
Source for each material claim:
Useful next step:

## Review
Exact version:
Waiting / Revision Requested / Approved:
One correction:
Card updated:
Same-input comparison:

Approval is not publication. Reopen saved files to verify persistence. A manual upload is a valid first test.
Read the draft request
Using only the source packet and four cards I provide, prepare one useful draft answer to my client question. First list the sources you can actually read, the verified facts and unknowns. Ask for a missing fact instead of inventing it. Draft the answer in my voice, identify the sentence that answers the question, and show a claim-to-source check. Propose one specific improvement to a reusable card. Stop at a draft; do not post, schedule, send or connect an account.

When the setup or draft gets stuck

Editorial questions for applying the workshop.

Do I have to appear on camera?

No. A useful written answer, permitted photograph or faceless explanation can work. Choose the format after deciding what the reader should learn.

Will one prompt permanently install everything?

No. The live workshop shows installation and project-creation limits. Check current app capabilities, use supported setup steps and reopen the saved files. A manual upload still supports the exercise.

What belongs in the first folder?

One anonymized question, a small set of permitted sources and a note explaining what each source establishes. Keep rights, unknown facts and intended use visible.

Does Approved mean it is published?

No. It means a specific version passed review. Publishing needs the account, destination and approved action to be explicit, followed by an actual readback.

Why does the draft still sound generic?

Compare it with your writing samples. Replace a vague request such as make it better with an observable correction, then test the same input again.

How can I tell whether the loop works?

Open the input, ledger, draft and revision yourself. Check the actual answer and every material fact. Record an incomplete step as incomplete instead of calling the workflow installed.

Selected teaching passages

Selected caption excerpts from the historical recording. These are deliberately non-contiguous passages, not the full transcript. Current practice and corrections are written separately above; the embedded recording is unchanged.

3:14 · Turn raw material into an organized input

So, one of the biggest problems that I found with making content is like I have to categorize it and I have to understand it and I have to figure out a creative concept to do something with it, right?

My process is once the AI starts working, the first thing it does is it analyzes everything and categorizes it and then it tries to make connections.

6:43 · Explain the weekly input and its limits

so you can see do not post. this is kind of the message to the AI of what it's supposed to do with the weekly context.

So you can add whatever context via uh like document or note or whatever it is to this folder.

8:53 · Answer the question the content raises

It asked what the question was, but it never actually answered that question.

11:21 · Make the revision request specific

And then you can also move stuff into revision requested and then make that request, right? add it on a document.

Say, "Hey, I put the following things into revision requested. Here's why. Here's why it doesn't match my brand."

Make sure to update the system and redo generate these assets.

12:24 · Repair the process behind the defect

It's flickering. Please help. please resolve this and recreate the assets and figure out why it happened and fix the system, right?

the goal is no longer just to get a good output. The goal is to fix the system

2:02:07 · Adapt captions to the destination

And so they're two different setups that you want to use for that. You would not want to use the same captions for each.

Keep building the skill

Context Cards and NotebookLM for Useful Client Explanations
Inspect a workflow against observable evidence.

Make content sound like you
Use writing samples to review the voice of a draft.

Check connected app access
Verify what the tool can actually read before designing the loop.

Follow the next classes: YouTube · LinkedIn · Instagram

After the answer review

Give the content process a home in your business.

The workshop teaches the intake, draft and review method. Content Machine is the separate paid offer to review when you want help setting up that work in your own Claude environment.