The Pack

---
name: 01-experiment-brief
description: Run before a hunch eats a week. Trigger - "run 01" plus the hunch in your own words, the channel it touches, and how long you will give it.
---

# TASK 01 — EXPERIMENT BRIEF

## INPUT
The hunch, written the way you said it out loud.
The channel, format or offer it touches. What you
are doing instead of it right now. How long you
are willing to give it before you stop.

## PROCESS
1. Turn the hunch into a claim that can be wrong.
   "Short form deserves more effort" is not a
   claim. "Three researched Reels a week move more
   email signups than five repurposed ones" is.
   If nobody could disagree with it, send it back.
2. Name the golden metric before designing
   anything. Email, almost always. If the honest
   metric here is followers, say so in one line
   and say what those followers are for, because
   that is a different experiment with a different
   risk.
3. Write down the number expected. Pre-register
   it. A result with no prior expectation is a
   story told after the fact.
4. Set the sample and the kill date in the same
   sentence. The kill date is a date, not a
   feeling, and it goes in the calendar today.
5. Choose the barbell end on purpose. Low effort
   and genuinely honest, or unreasonable effort
   and genuinely researched. If the design lands
   in the middle, redesign it or do not run it.
6. Name the one thing that would make the result a
   lie. A cart open week underneath it. A guest
   with a bigger audience. Seasonality. Say it
   before the data, not after.
7. Decide now where the write up goes: Sunday
   essay, a Lab call, or nowhere. "Nowhere" is a
   legitimate answer and it changes the design.

## OUTPUT
CLAIM. GOLDEN METRIC. EXPECTED NUMBER. DESIGN,
meaning what ships, how many, over what window.
KILL DATE. THE CONFOUND. WHERE THE WRITE UP GOES.
One line at the end: what I will do differently if
this comes back false.

## RULES
- Never run an experiment across a launch week and
  call the result clean. Say the launch polluted
  it or move the window.
- No experiment that cannot fail. If every outcome
  gets filed as "learning", it is a content plan.
- Never report a result without the expected
  number printed next to it.
- Do not draft the essay before the data. Deciding
  the conclusion first is the exact thing I say
  other people do.

## EDGE CASES
- The hunch is really a fear. Somebody else grew
  fast on a platform and the honest sentence is "I
  am behind." That is not a claim, that is envy
  with a spreadsheet. Rewrite it as what their
  result would have to prove about my audience,
  and if it proves nothing about my audience, kill
  the idea here and say why in one line.
- The thing can only happen once. A launch, a
  summit, a single collaboration. It is a case
  study, not an experiment, and the sample is one.
  Pre-register the expected number anyway, then
  label the write up a case study so nobody reads
  a sample of one as a rule.

## GOOD LOOKS LIKE
"Claim: three researched Reels a week move more
email signups than five repurposed ones. Golden
metric is email, not views, not followers.
Expected: 40 signups over four weeks against a
baseline of [BASELINE]. Kill date [DATE], in the
calendar today. The confound is the Lab doors
opening in week two, which will lift email on its
own, so week two comes out of the count and the
window stretches to five weeks. If it comes back
false I stop paying for the research hours and go
back to the low effort end of the barbell, which
is where the honest stuff has been winning
anyway."

---
name: 02-copycat-line
description: Run before publishing anything built on somebody else's proven format. Trigger - "run 02" plus the piece, the format it borrows, and where you first saw it working.
---

# TASK 02 — THE COPYCAT LINE

## INPUT
The piece, drafted or described. The proven format
it is built on and where it was first seen
working. Whose data, story or argument is inside
it. The metric it is being made for.

## PROCESS
1. Separate the FORM from the WORK. Form is hook
   shape, pacing, structure, thumbnail grammar,
   the order the beats land in. Work is the
   research, the number, the story, the take.
   Borrowing form is craft. Borrowing work is the
   copycat problem, and the line runs exactly
   there.
2. Ask the ownership question. If the original had
   never been published, could this piece still
   exist? If the answer is no, it is a repost with
   my face on it, whatever the caption says.
3. Count what is being added. My own data. My own
   failure. A read that cuts against the original.
   Zero additions is midslop. One is thin. Two or
   more and it is mine.
4. Check which signal it serves. If the only thing
   it can move is follower count, say the awkward
   part out loud before publishing: follower count
   is the signal I posted on 2 September 2026 that
   brands overvalue, and farming it is the
   behaviour I called the copycat problem. Decide
   knowing that, do not decide around it.
5. Barbell test. Low effort and honest, or high
   effort and researched. A borrowed format run at
   medium effort is the definition of the middle.
6. Attribution call. If a viewer would recognise
   the format, credit the person who built it, by
   name, in the post and not in a comment.
7. Publish, rework or bin. Say which, and give the
   one sentence reason.

## OUTPUT
FORM VS WORK, two short lists. THE OWNERSHIP
ANSWER, yes or no. WHAT IS BEING ADDED, counted.
WHICH SIGNAL IT SERVES. BARBELL END. ATTRIBUTION
LINE, written out ready to paste. THE CALL, with
the reason.

## RULES
- Never publish a piece whose whole argument is
  somebody else's argument in my voice.
- A credit in the caption does not buy the work.
  Attribution fixes form. It does not launder
  substance.
- "It performs" is not a reason. It performing is
  what created the copycat problem in the first
  place.
- If the piece only moves follower count, it does
  not go out under Creator Science. That is the
  rule I did not have on 2 September and it is the
  rule now.

## EDGE CASES
- The format is mine and somebody else scaled it
  past me. The instinct is to post about it. Do
  not litigate it publicly, it reads as small and
  it markets them. Republish the idea with the one
  thing they cannot copy, which is my own data and
  the cost line underneath it.
- The proven piece is my own, from two years ago,
  and it did numbers. Reposting yourself is still
  repackaging if nothing has been learned since.
  It only ships if the reissue carries what
  changed, what I got wrong, or a number that has
  since moved. Otherwise it is midslop with
  seniority.

## GOOD LOOKS LIKE
"Form borrowed: the thirty day test structure and
the receipts on screen. That is craft, take it.
Work borrowed: none, the numbers are the summit
launch numbers I already published, $101,666 on
$23,300 of cost. Ownership answer: yes, this
exists without the original, because the original
does not have my costs. Adding two things, the
cost line and the part where the affiliate
programme did almost nothing. Signal it serves:
email, because the CTA is the teardown episode,
not the follow button. Credit [CREATOR] by name in
the caption for the structure. Call: publish."

---
name: 03-brand-deal-screen
description: Run on every inbound sponsorship, affiliate offer or paid post. Trigger - "run 03" plus the pitch, the ask, and whether you have actually used the product.
---

# TASK 03 — BRAND DEAL SCREEN

## INPUT
The pitch as it arrived. What they are asking for
and where it would run. Whether the product has
been used or tested, honestly. Anything they said
about why they picked me.

## PROCESS
1. Run the free test first. Would I recommend this
   without the deal? If no, stop. Everything below
   is wasted work and the answer is already
   written.
2. Used or tested. The standard on my own sponsor
   page is companies and products we used or have
   tested and love. Hold to it. "It looks good" is
   not tested.
3. Name who it is for, specifically. Not "my
   audience". The creator who is [SPECIFIC
   SITUATION]. If the honest answer is "some of
   them", that is a mismatch, not a discount.
4. Read what they are actually buying. If the
   pitch quotes a follower number back at me
   rather than a fit, they are buying the signal,
   not the audience, and that is the same
   misplaced signal I posted about on 2 September
   2026. Either reprice against email and
   downloads, or decline and tell them why.
5. Long term or nothing. One off placements cost
   the same trust and build none of the
   relationship. Ask for the multi channel version
   or pass.
6. Price the failure. If this product disappoints
   [X] readers, what does that cost the thing all
   of it runs on. Write the number of readers, not
   a feeling.
7. Write the decision as a yes with conditions, or
   a no in one sentence I would happily say to
   their face.

## OUTPUT
FREE TEST, pass or fail. USED OR TESTED. WHO IT IS
FOR, one sentence. WHAT THEY ARE BUYING. TERMS
ASKED FOR. FAILURE COST. THE DECISION, with the
reply drafted and ready to send.

## RULES
- Never write ad copy for a product that has not
  been used. No exceptions for good rates.
- The rate is never the reason. [RATE] stays out
  of the reasoning and out of the copy.
- Do not soften a no into a maybe to keep the door
  open. The audience knows when you are only
  saying it for the money, and so does the brand.
- No exclusivity or category lockout without an
  end date on it.

## EDGE CASES
- The product is genuinely good and the founder is
  a friend or a portfolio company. The free test
  passes and that is exactly the problem, because
  I would recommend it for free, so the audience
  cannot tell which one this is. Disclose the
  relationship in the read itself, not in a
  footnote, or take it as a free recommendation
  and no money.
- The pitch is strong, the fit is real, but the
  brief demands a claim I cannot verify. Do not
  negotiate the claim down to something vague.
  Offer to say what I actually observed, with the
  number, and let them decline. A softened claim
  is still a claim in my voice.

## GOOD LOOKS LIKE
"Free test: fail. I have had the trial open for
three weeks and I have not opened it once, which
is the answer. Reply going out today: 'Thanks for
the detailed pitch, and genuinely, the positioning
is sharp. I only run partners I would recommend
without the deal, and I am not there with this
one yet. If that changes after I have actually
used it in a real project, I will come back to
you first.' No counter, no rate, no maybe."

---
name: 04-sunday-essay
description: Run when Sunday's newsletter needs to become an argument. Trigger - "run 04" plus the idea, the evidence you actually have, and what you are still unsure about.
---

# TASK 04 — SUNDAY ESSAY

## INPUT
The idea in one messy paragraph. The evidence
actually in hand: an experiment I ran, a number I
published, a conversation I am allowed to repeat.
The part I am still unsure about.

## PROCESS
1. Find the incremental revelation, not the
   epiphany. The reader should finish one notch
   further along, not converted. If the draft
   promises transformation, cut the promise.
2. Write the spine as one sentence somebody could
   disagree with. If it takes two sentences, the
   idea is not ready and Sunday can take a
   different one.
3. Pick the barbell end and commit. Either this is
   the honest emotional one, written fast and
   true, or it is the researched one with the
   receipts. Say which at the top of the outline
   so the effort matches.
4. Put my own evidence in the first third. Not a
   study, not a quote from someone famous. What I
   ran, what it cost, what it returned.
5. Include the part that cuts against me. The
   experiment that failed, the thing I got wrong,
   the place the advice does not apply. That
   paragraph is the trust, the rest is the
   content.
6. Ten minute test. Read it against the promise on
   the subscribe page: a smarter creator in ten
   minutes. Anything that does not serve the spine
   comes out, however good the sentence is.
7. Land on one thing to do this week, specific
   enough to do on Tuesday morning without asking
   a follow up question.

## OUTPUT
SPINE, one sentence. BARBELL END. OUTLINE with
section headers. MY EVIDENCE, with the number and
its cost. THE PART THAT CUTS AGAINST ME. THIS
WEEK, one action. SUBJECT LINE options, three, no
curiosity gap tricks.

## RULES
- No essay that only works because someone famous
  said something. If the only evidence is borrowed
  authority, it is not ready.
- Never a number without its cost in the same
  paragraph. Revenue without cost is a brag.
- Do not publish a framework with more than four
  parts. If it needs five, two of them are the
  same part.
- No subject line that promises a thing the essay
  does not deliver in the first screen.

## EDGE CASES
- The strongest evidence came from a Lab call or a
  private conversation with a guest. It is the
  best material in the draft and it is not mine to
  publish. Ask the person directly, in writing,
  before it goes in. If the answer is no or the
  answer is slow, write the mechanism without the
  identifying detail and say a version has been
  held back.
- The honest conclusion is that I do not know. The
  temptation is to pick a side by Sunday. Publish
  the open question instead, with the two
  positions stated fairly and the experiment that
  would settle it. That is a real essay. A forced
  conclusion is the thin one.

## GOOD LOOKS LIKE
"Spine: the signal brands pay for is the signal
that says least about whether anyone trusts you.
Barbell end: researched, this one needs receipts.
My evidence goes in the first third, the summit
launch at $101,666 on $23,300 of cost, where every
dollar came from email and none of it came from a
follower count. The part that cuts against me: I
spent August trying to get better at short form,
which means I was chasing the same number I am
telling you not to chase. This week: open your
last three brand deals and check which number was
in the first line of their pitch."

---
name: 05-guest-question-design
description: Run before Tuesday's interview. Trigger - "run 05" plus the guest, what they are known for, and the two or three things you have already read or watched.
---

# TASK 05 — GUEST QUESTION DESIGN

## INPUT
The guest and what they are known for. Everything
already read, watched or listened to about them,
listed. The number I want them to say out loud.
What my listener is stuck on that this guest can
actually move.

## PROCESS
1. Find the question only this guest can answer.
   If another operator in the same lane could
   answer it just as well, it is a filler
   question. Cut it.
2. List the three answers already given on five
   other shows. Those are banned openings. Start
   where their other interviews stopped.
3. Pick the number to ask for and pick it before
   the call. Revenue, conversion rate, cost, churn,
   hours. Ask for it plainly, once, early, and
   accept a no without negotiating on air.
4. Design the earn order. The uncomfortable
   question goes in the back third, after they
   have watched me handle two easier ones fairly.
5. Mark exactly one place to push. One. Write the
   push as a sentence, not an intention, so it
   does not turn into an argument or evaporate.
6. Write the mechanism questions: how did you
   decide, what did it cost, what did you stop
   doing. Not what is your advice.
7. Name the listener takeaway before the call. If
   I cannot say what somebody does differently on
   Wednesday, the booking is wrong, not the
   questions.

## OUTPUT
THE ONLY QUESTION THEY CAN ANSWER. BANNED
OPENINGS, three. THE NUMBER TO ASK FOR, with the
exact wording. QUESTION ORDER, in three acts. THE
ONE PUSH, written out. LISTENER TAKEAWAY, one
sentence. Two backup questions if the room goes
cold.

## RULES
- Never open with the origin story. It is on five
  other shows and the listener has heard it.
- Do not ask for a number I would refuse to give
  about my own business. Ask for the ones I have
  published.
- Never push twice. One push is an interview, two
  is an ambush and the guest stops being useful.
- No question that starts "talk to me about your
  journey".

## EDGE CASES
- The guest is a friend and the episode will be
  soft. Friendship makes the easy questions easier
  and the real one impossible. Write the push
  before the call, send it to them in advance, and
  tell them it is coming. Warned is not softened.
- The guest's headline number cannot be verified
  and the whole booking rests on it. Do not
  platform it unchallenged and do not accuse them.
  Ask how it is measured, what is included, and
  over what period. If the answer is vague, the
  episode reframes around the method and the
  number stays out of the title.

## GOOD LOOKS LIKE
"The only question they can answer: you built a
system for finding proven formats, and I have just
argued in public that the same behaviour is what
is making Instagram feel like the same six videos.
Where is your line between studying a pattern and
reposting somebody's work? Banned openings: how
you got started, why you left your job, what your
first video was. Number to ask for, in these
words: what percentage of the accounts you work
with convert views into email, and what is the
median. The push, act three, once: if the format
is the value, what stops your client's competitor
from running the same system next month?"

---
name: 06-offer-architecture
description: Run before a launch is built. Trigger - "run 06" plus the product, who it is for, and what they are currently doing instead of buying it.
---

# TASK 06 — OFFER ARCHITECTURE

## INPUT
The product as it stands today. Who it is for, in
one sentence, and what they do instead of buying
it right now. The price being considered. How much
of my time each sale consumes.

## PROCESS
1. Write the promise in their words and make it
   specific. "Double your paid members" beats
   "grow your membership". Vague promises are why
   people do not buy, not price.
2. Build the stack: core, upsell, downsell. Every
   time the summit launch switched somebody to
   another rung, it made more money. One rung is
   a missed sale waiting to happen.
3. Find the fulfilment ceiling before the price.
   If the core offer caps at [N] buyers because I
   personally deliver it, that is the real revenue
   number and everything above it is a refund.
4. Reduce risk in the order that costs least:
   evidence first, guarantee second, more of my
   time last. My time is the expensive lever and
   it is the one I reach for first, so stop me.
5. Set the trust threshold. A [PRICE] promise needs
   more proof than a newsletter promise. List the
   micro promises already kept that earn this one,
   and if the list is thin, the price is early.
6. Write the urgency that is true. A cart date, a
   cohort start, a capacity cap. If there is no
   true reason to buy now, the launch is a
   discount pretending to be a deadline.
7. Lock the stack and the landing page before
   registration opens. Changing the offer mid
   launch is where the last one bled hours.

## OUTPUT
PROMISE, one sentence in their words. THE STACK,
core, upsell, downsell, with prices. FULFILMENT
CEILING, a number. RISK REVERSAL, in cost order.
TRUST THRESHOLD, with the micro promises listed.
TRUE URGENCY. LOCK DATE. Then a go or hold call.

## RULES
- Never launch a core offer whose fulfilment has a
  ceiling I would be embarrassed to hit.
- No guarantee I have not decided in advance I
  will honour without argument.
- Do not build the marketing before the content
  exists. The summit taught that one the expensive
  way.
- Never price against what other people charge. My
  costs and my ceiling set the price.

## EDGE CASES
- The offer is genuinely good and the audience has
  not been warmed for it at all. The instinct is a
  bigger discount. Discounts do not fix a trust
  gap, they widen it. Hold the launch, spend the
  window publishing the evidence, and put the same
  offer out at full price when the proof exists.
- The upsell is more profitable than the core and
  easier to deliver. That is not permission to
  make it the core. Ask why anyone buys the core
  at all, and if the honest answer is "nobody
  should", cut the core and rebuild the stack
  around the thing that works.

## GOOD LOOKS LIKE
"Promise, in their words: fill your membership
with people who stay past month three. The stack
is core at [PRICE], upsell at [PRICE], downsell
self paced, because on the summit every single
switch to another rung made more money, and the
$47 add on sold every day with zero refunds.
Fulfilment ceiling is [N] because I am in the
room, so that is the revenue number, not the
model. Risk reversal in cost order: three named
case studies first, money back second, more of my
calendar never. Hold. Two of the three case
studies do not exist yet, and that is a proof
problem, not a price problem."

---
name: 07-launch-teardown
description: Run after any launch, experiment or event closes. Trigger - "run 07" plus the numbers, the costs, and the moment you knew something was wrong.
---

# TASK 07 — LAUNCH TEARDOWN

## INPUT
Every revenue line and every cost line, including
the ones that are embarrassing. The timeline, from
first idea to cart close. The moment it was
obvious something was off. Who did the work.

## PROCESS
1. Print revenue and cost in the same table before
   writing a word of narrative. Margin is the
   headline, not revenue. A launch reported
   without its costs is marketing, not a teardown.
2. Break revenue down by rung. Core, upsell,
   downsell, add ons, with buyer counts, so the
   reader can see which rung actually carried it.
3. List what worked and why it worked, mechanism
   first. "The live pitch converted" is not a
   finding. "The live pitch with a bonus attached
   converted, the one without it did not" is.
4. List what to change next time as instructions
   to my future self, not regrets. Each one starts
   with a verb.
5. Name the expensive lesson plainly, with its
   number. A tool that failed, a domain that cost
   more than planned, a channel that returned
   almost nothing for real effort.
6. Credit the people by name and say what they
   actually did. Not a thank you list. What would
   have gone wrong without them.
7. Write the 80/20 version, the shortest sequence
   that would have produced most of the result,
   and mark what gets cut next time.

## OUTPUT
THE TABLE, revenue and cost, margin at the bottom.
BY RUNG, with buyer counts. WHAT WORKED, mechanism
first. WHAT CHANGES, verbs. THE EXPENSIVE LESSON,
with its number. CREDITS, with what they did. THE
80/20 PLAYBOOK, in order.

## RULES
- Never publish revenue without cost in the same
  screen.
- No lesson stated as a feeling. Every finding
  ties to a line in the table.
- Do not blame a vendor by name for a failure that
  was a planning failure. Name the tool, own the
  choice.
- Never publish a buyer's, member's or partner's
  numbers to make my own look better.

## EDGE CASES
- The launch missed its target and the honest
  teardown makes the product look weak to people
  who have not bought it yet. Publish it anyway,
  and put the mechanism in front. The audience
  that stays for the miss is the audience that
  buys the next one, and a teardown that only ever
  runs after wins is an advert with a table in it.
- The biggest cost was somebody's performance fee
  or a partner's cut, and printing it prices them
  publicly. Print it as a percentage of revenue
  and describe the role, not the invoice, and
  clear it with them before it goes out.

## GOOD LOOKS LIKE
"Revenue $101,666, costs $23,300, margin $78,400,
and the margin is the headline. By rung: live
cohort $63,739 from 74 buyers, cohort upsell
$9,000 from 9, self paced $17,752 from 97, VIP
tickets over $11,000 from 238 at about 7.5%. What
worked, mechanism first: the day two live pitch
carried a bonus and the day three one did not, and
that is the entire difference between them. What
changes: build the presentation before any
marketing asset. Lock the stack before
registration opens. Skip affiliates above $1,000,
they returned $1,800 of commission for real work.
The expensive lesson has a number on it, $7,800
for a domain, and I would spend it again."

---
name: 08-week-shape
description: Run on Friday or Monday morning to shape the week. Trigger - "run 08" plus what is already committed, what is behind, and whether a launch or Lab opening is live.
---

# TASK 08 — WEEK SHAPE

## INPUT
What is already in the calendar. What is behind
from last week. Whether a launch, a cohort or a
Lab opening is live. Which experiments have a kill
date inside this week. Anything owed to a guest, a
partner or a member.

## PROCESS
1. Protect the two fixed points first. Sunday
   newsletter and Tuesday episode ship. They are
   the micro promises everything else stands on,
   so nothing else gets scheduled on top of them.
2. Place the deep work against the barbell. One
   unreasonable effort block for the researched
   piece, and the honest short form in the gaps.
   Never the reverse.
3. Check the experiments with kill dates this
   week. Each one gets a decision, not a check in.
   Continue, kill, or extend once with a reason.
4. Put the person owed something before the
   channel owed something. A guest waiting on a
   reply outranks a scheduled post.
5. Book the refusals into the plan. Which pitches
   get run through task 03 and answered, and which
   inbound gets a no this week rather than
   drifting to next week.
6. Name the one thing that, if it slipped, would
   make the week a loss. Put it before Wednesday.
7. Schedule genuinely off, and make it off. Out of
   office is not the same as off. If the phone
   still has the apps on it, it is not scheduled.

## OUTPUT
FIXED POINTS. THE UNREASONABLE EFFORT BLOCK, with
its subject. SHORT FORM SLOTS. DECISIONS DUE, one
line each with the ruling. PEOPLE OWED. THE NOs
GOING OUT. THE ONE THING. OFF BLOCK, with what
gets removed from the phone.

## RULES
- Never move Sunday or Tuesday to make room for a
  launch task. The launch is the temporary thing.
- No week with more than one unreasonable effort
  block. Two means neither gets it.
- An experiment past its kill date does not get
  rolled over silently. Kill it or extend it out
  loud.
- Do not schedule off time and then keep the apps
  on the phone. That is out of office, not off.

## EDGE CASES
- A launch week collides with a newsletter and an
  episode. The temptation is to skip one and call
  it a launch tax. Skipping is what breaks the
  micro promises the launch is selling against.
  Ship both smaller, from work already in the
  bank, and cut the launch's nice to have assets
  instead.
- Something genuinely good and unplanned lands
  midweek, a guest, an investment, an offer. It
  will fit only by eating the unreasonable effort
  block, which is the one block that cannot be
  rebuilt later in the week. Move it to next week
  unless it expires, and if it expires, say out
  loud which researched piece is being sacrificed.

## GOOD LOOKS LIKE
"Fixed points hold: Sunday essay, Tuesday episode,
no negotiation. Unreasonable effort block is
Wednesday morning and its subject is the signal
piece, the one about brands buying the number
that says least about trust. Short form in the
gaps, honest end of the barbell, no borrowed
formats this week. Decisions due: the Reels
experiment hits its kill date Thursday, and it
gets a ruling, not a check in. People owed
outranks channels owed, so [GUEST] gets the reply
before anything is scheduled. Two NOs going out
today, both from task 03, both one sentence. The
one thing is the Wednesday block. Off is Saturday
and the apps come off the phone Friday night,
because out of office is not off."