The Pack

---
name: 01-issue-build
description: Run when building the day's issue. Trigger - "run 01" plus the day's link haul and what is already drafted.
---

# TASK 01 — BUILD THE ISSUE

## INPUT
The day's haul: trade press links, press
releases, things people sent me, whatever I
clipped. Which case study is running as ICYMI.
Anything already half drafted.

## PROCESS
1. Kill first. Most of the haul is an
   announcement with no mechanism attached. If I
   cannot say what has to be true for it to make
   money, it is not an item, it is a headline.
   Cut it and say why in four words.
2. Rank what survives by argument, not by
   importance. The lead item is the one where I
   have an observation nobody else is making, not
   the biggest company in the pile.
3. Draft the lead. Quote the trade press at
   length so the reader does not have to leave,
   then break my take into numbered observations.
   The first observation is the mechanism. One of
   them is the bear case.
4. Pick the PLUS item. It runs second and it
   should sit at an angle to the lead, not
   reinforce it. Two items about the same thesis
   is one item.
5. Write quick hits. Each is the fact, one line
   of take, then the source in brackets. If the
   take is just a summary, drop the hit.
6. Set the paywall line. Free gets the items that
   argue with the industry. Paid gets the items
   that give somebody an unfair advantage: unit
   economics, the case study detail, the thing a
   competitor would pay to know. Then write the
   behind the paywall list as three flat titles.
7. Flag every number in the issue for task 07
   before it ships.

## OUTPUT
KILLED, with four-word reasons. LEAD ITEM, with
the quoted passage and numbered observations.
PLUS ITEM. QUICK HITS with bracketed sources.
ICYMI line. BEHIND THE PAYWALL, three titles.
NUMBERS TO CHECK.

## RULES
- Never open an item by restating the headline. I
  quote the source, then I start talking.
- No item survives on "this is interesting". If
  there is no mechanism and no bear case, it is a
  quick hit or it is nothing.
- Never put two items behind the paywall that
  answer the same question. Paid subscribers
  notice faster than free ones.
- No engagement bait, no "what do you think", no
  numbered listicle framing. My readers are
  running businesses.

## EDGE CASES
- The best item in the haul is about a company
  that sponsors the newsletter or is run by a
  recent podcast guest. Run it, and run it
  bearish if that is where the argument lands.
  Disclose the relationship in the item itself,
  in one clause, not in a footer. If I cannot
  write it honestly, kill it and do not replace
  it with a soft version.
- The lead item is an exclusive I got because a
  source argued with me in a reply thread. The
  reply is not yet public context for the reader.
  Reconstruct the argument from my original
  claim, so the correction or the scoop makes
  sense standing alone, and keep my original
  wrong language in view. Never present a
  reply-thread quote as though it arrived by
  email.

## GOOD LOOKS LIKE
"KILLED: another funding round, no model. LEAD:
Fox buying Roku. The observation is not the price
tag, it is that Fox is the only conglomerate that
refused to bundle its IP under one streamer and
now owns two of the largest FAST channels plus a
podcast network. Bear case: none of that solves
the fact that Fox News poisons the brand for
half the country. PAYWALL: the Workweek unit
economics, because that is the one somebody would
pay to see."

---
name: 02-observations-pass
description: Run on one story that deserves the full treatment. Trigger - "run 02" plus the story, the source, and my first reaction.
---

# TASK 02 — THE OBSERVATIONS PASS

## INPUT
One story or announcement. The source. Whatever I
already think about it, however half formed.

## PROCESS
1. Restate the deal in money terms in one
   sentence. Who pays whom, for what, and how
   often. If I cannot write that sentence, I do
   not understand the story yet and neither will
   the reader.
2. Ask what the company thinks it bought. Then
   ask what it actually bought. The gap between
   those two is usually the whole piece.
3. Find the historical rhyme. This industry has
   run this experiment before. Name the earlier
   version and what happened to it. Micropayments
   never took off. HuffPost was going to pass the
   Times. Say it out loud.
4. Number the observations. Three to five. Each
   one stands alone and each one is a claim
   somebody could argue with, not a description.
5. Write the bear case as its own observation, in
   my own voice, not as a hedge. If I am bearish,
   the word bearish appears.
6. Check my own position for rot. If this cuts
   against something I argued earlier this year,
   say so in a clause and do not pretend it does
   not.
7. End on the question the industry is refusing
   to ask. Stack it with the two or three
   unanswered mechanics underneath it.

## OUTPUT
THE DEAL IN MONEY TERMS, one sentence. WHAT THEY
THINK THEY BOUGHT / WHAT THEY BOUGHT. THE
HISTORICAL RHYME. OBSERVATIONS, numbered, one of
them bearish. WHERE THIS CUTS AGAINST ME. THE
UNANSWERED QUESTIONS.

## RULES
- Never write "it remains to be seen". Either the
  mechanism works or I say which way I lean.
- No prediction without the condition attached.
  "This fails unless they can break news" is a
  take. "This will fail" is a coin flip.
- Never use a platform's own metric to make my
  argument for me. If the case rests on a view
  count or an open rate, run task 07 first or
  build the argument on revenue instead.
- Do not reach for slop, enshittification or any
  other word that does the thinking. Describe
  what the company is actually doing.

## EDGE CASES
- The story is about AI in a newsroom and every
  other outlet is running the panic version. Do
  not join it and do not contrarian it either.
  Split the thing being described into where it
  sits on the spectrum: raw unedited output
  shipped as reporting, versus transcription,
  research and line editing. Judge those
  separately. The disclosure question is a
  different question from the quality question,
  and conflating them is how these pieces go
  wrong.
- I am bearish and the company is one whose
  founder I like and have had on the podcast.
  Publish the bearish read at full strength and
  name the relationship in the piece. Then send
  them the link before it goes out, not for
  approval, so the correction arrives as a
  correction rather than as a grievance.

## GOOD LOOKS LIKE
"In money terms: MS NOW is asking $7.99 a month
for access to hosts you can already watch. What
they think they bought is a subscription
business. What they bought is a tip jar with a
paywall on it. Ok, I have a few observations.
First, the talent-forward part is real, and
parasocial relationships do convert. Second, and
this is why I am bearish, they cannot break news
after the split from NBC News, and a subscription
without newsgathering tops out at the devoted
fans. Third, and this is the part that actually
baffles me, Versant owns CNBC and Golf Channel
and still will not bundle. Cable was profitable
for one reason and they have decided to forget
it."

---
name: 03-threads-ledger
description: Run weekly, or after a post that did numbers. Trigger - "run 03" plus the post, its stats, and what came back in the replies.
---

# TASK 03 — THE POSTING LEDGER

## INPUT
The post or the week of posts. Platform. Whatever
stats the dashboard shows. What arrived in the
replies. Anything the Substack dashboard
attributes to that platform.

## PROCESS
1. Separate the numbers into three columns before
   scoring anything. REACH is views and
   impressions. ENGAGEMENT is likes, replies,
   reposts. CONVERSION is free signups, paid
   subscriptions and net follower change from the
   Substack dashboard, not the platform's.
2. Discount reach to zero as a success measure.
   On 25 August 2026 I published that Threads had
   given me 3.5 million views and 21,000
   interactions in a month, 35 net new followers,
   and across all of 2026 four free signups and
   one paid subscription. Reach is the number I
   spend the rest of my week telling other people
   not to trust. It does not get to count here.
3. Score the fourth column, the one no dashboard
   has: REPORTING YIELD. Did a named operator
   correct me, confirm something, or hand me a
   number. Did it produce a podcast booking, a
   source, a document, or a line I can quote with
   attribution. This is the only column where
   Threads has ever paid me.
4. Attribute honestly. If the yield was a person
   who would have replied to an email, the post
   did not earn it. If they replied because it
   was public and they had to, the post earned
   it.
5. Rule on the post. WORTH IT, on which column.
   NOT WORTH IT. Or UNSCORED, which means it was
   fun and I am allowed to have that, as long as
   it is logged as fun and not as work.
6. Roll up the week. If reporting yield is zero
   across a full week, that is the finding. Say
   it plainly and do not soften it with the reach
   number.
7. Set the next review date and name the one
   thing that would change the verdict.

## OUTPUT
REACH / ENGAGEMENT / CONVERSION, stated flat.
REPORTING YIELD, named people and named artefacts
or the word none. ATTRIBUTION CALL. VERDICT: worth
it on which column, not worth it, or unscored.
WEEKLY ROLL-UP. WHAT WOULD CHANGE MY MIND.

## RULES
- Never let a reach number appear in the verdict
  sentence. It goes in the ledger and stays
  there.
- Never count a reply from someone already in my
  contacts as reporting yield. That is a
  relationship, not a platform.
- Never write a conversion figure I have not
  personally read off the Substack dashboard.
  Bracket it as [DASHBOARD FIGURE] and ask.
- Do not recommend that I quit the platform on a
  single week of data, and do not recommend I
  post more because a post did numbers. Both are
  the same mistake.

## EDGE CASES
- A post did enormous reach and produced nothing
  in any other column. This is the trap, because
  it feels like the best day of the month. Score
  it NOT WORTH IT and write the sentence out in
  full: this post generated [VIEWS] views and
  zero of anything I can bank. Sit with it. That
  is the entire argument I make about audience
  inflation, pointed at myself, and the ledger is
  worthless if it flinches here.
- The reporting yield is a correction, which
  means the post was wrong. That is still a win
  and it must be scored as one, but it carries an
  obligation: the correction runs in the
  newsletter with my original wrong language kept
  above it so the clarification makes sense.
  Score the yield, then open the correction as a
  task 01 item. Never score a correction as a
  loss just because it stung.

## GOOD LOOKS LIKE
"Reach [VIEWS]. Engagement [INTERACTIONS].
Conversion, off the Substack dashboard: zero free
signups, zero paid, net follower change
[DASHBOARD FIGURE]. Reporting yield: Workweek's
VP of marketing and its founder both came into the
replies to tell me my read on their network model
was wrong, on the record, with an acceptance rate
figure I did not have. That is two named
executives and a podcast lead out of one post.
VERDICT: worth it, on reporting, and only on
reporting. The reach figure is the number I would
sneer at if a publisher put it in a media kit.
It does not get to be my scoreboard just because
it is mine."

---
name: 04-post-or-issue
description: Run when a take is forming and I have not decided where it lives. Trigger - "run 04" plus the argument and what prompted it.
---

# TASK 04 — POST IT OR SAVE IT

## INPUT
The argument, however rough. What set it off.
Whether it touches a story I plan to write up.
Whether anyone named in it might respond.

## PROCESS
1. Ask what the post is for before asking where
   it goes. There are exactly three honest
   answers: it is bait for a correction, it is
   the argument itself, or it is a reflex. Name
   which. Reflexes do not get published.
2. If it is bait for a correction, post it
   natively and post it first. The whole value is
   that the people who know better can see it and
   are annoyed enough to answer. Write it so the
   claim is falsifiable and the wrong version is
   specific enough to provoke a specific fix.
3. If it is the argument itself, hold it. An
   argument I have already spent in a post is
   worth less in the issue, and the issue is the
   thing that has ever converted anybody.
4. Check the paywall consequence. If this take is
   headed behind the paywall on [DATE], posting
   the spine of it publicly is giving away the
   part people paid for. Post the question, hold
   the answer.
5. Choose the platform on argument, not on habit.
   The media operators who correct me live in one
   place, the executives in another. Say which
   room this argument needs and why.
6. Write it in full sentences with a position in
   the first line. No hooks, no thread bait, no
   question at the end fishing for replies.
7. Set the harvest date. When do I come back and
   run task 03 on it. If I would not bother, that
   is evidence it was a reflex.

## OUTPUT
WHAT IT IS FOR: bait, argument, or reflex. THE
DECISION: post now, hold for the issue, or both
in this order. THE PLATFORM and the room it
reaches. THE DRAFT, if posting. WHAT STAYS BACK.
HARVEST DATE for task 03.

## RULES
- Never post the spine of a paid item. The
  question is free, the mechanism is not.
- Never publish a take about a named operator
  without deciding, in advance, that I will run
  their reply if they send one. If I would not
  run it, I do not get to post it.
- No thread padding. If it needs five parts, it
  is an issue item wearing a costume, and the
  costume is coming off in step 3.
- Never justify a post by the reach the last one
  got. That is the exact reasoning I spend my
  week arguing against.

## EDGE CASES
- The argument is genuinely good and genuinely
  free, and there is no correction to harvest and
  no paywall to protect. This is the hardest one,
  because every rule above says hold and the
  honest answer is that I will post it anyway.
  Do not pretend otherwise. Post it, log it as
  unscored under task 03, and let the ledger
  carry the cost openly rather than dressing a
  habit up as strategy.
- The take is about Meta, and the post is going
  on Meta's own platform. Post it. Do not add a
  knowing aside about the irony; it is the
  weakest available joke and it converts the
  argument into a bit. The argument that Meta
  underpays creators is stronger, not weaker,
  coming from a creator visibly supplying it free
  inventory. State the argument. Let the venue
  do its own work.

## GOOD LOOKS LIKE
"What it is for: bait. I think Workweek's new
platform play makes it a Substack competitor
without the monetization story to back it, and I
am not certain about the vetting. That
uncertainty is the asset. Post it natively,
today, with the claim stated wrongly enough to be
worth fixing. Hold the unit economics for the
paid item on [DATE]. Harvest Thursday. If nobody
from Workweek answers, it was a reflex and I will
score it as one."

---
name: 05-guest-screen
description: Run on anything arriving in the inbox that wants to be on the show or in the newsletter. Trigger - "run 05" plus the pitch, pasted whole.
---

# TASK 05 — SCREEN THE INBOX

## INPUT
The pitch, pasted whole, headers included.
Whether I have heard from this sender before.

## PROCESS
1. Classify in one line before reading properly.
   PR PITCH, GUEST PITCH, or SOURCE. They get
   different treatment and mixing them up wastes
   the only scarce thing here, which is my
   morning.
2. PR pitches get the one-strike test. Is this
   tailored to what I actually cover, or is it
   shotgunned. If it is shotgunned, the verdict
   is spam and blacklist, no reply, no polite
   decline. There are roughly six PR reps for
   every journalist and a pitch that was blasted
   to thousands will never one day be relevant.
3. Guest pitches go against the published bar, in
   this order: what do they make and where do
   they publish it, roughly how big is the
   audience, how do they monetize, and does the
   revenue support at least one full-time person,
   even if that person is themselves. Missing
   answers are not disqualifying on their own.
   Missing the fourth one is.
4. Then the real test, which the bar does not
   cover: is the business model interesting or
   just successful. A large audience monetized
   the obvious way is not an episode. A small
   audience monetized in a way I have not heard
   in a thousand interviews is.
5. Audit the numbers in the pitch through task
   07 before they earn any credit.
6. If it is a no, decide whether it is a no
   forever or a no for now. A no for now gets one
   line naming what would change it.
7. If it is a yes, write the one question the
   episode exists to answer, and the two numbers
   I need out of them or the case study does not
   stand up.

## OUTPUT
CLASS. VERDICT: spam and blacklist, no forever,
no for now, or book it. WHY, in one sentence.
FOR A BOOKING: the question the episode answers,
and the two numbers I must get. FOR A NO FOR NOW:
the one thing that would change it.

## RULES
- Never draft a courteous reply to a shotgunned
  pitch. Marking it spam is the reply.
- Never book on audience size. Size is the least
  interesting thing a guest brings and I have
  said so in public repeatedly.
- Never accept a revenue claim without the shape
  of it. "Six figures" with no split between ads,
  subscriptions and services is a press release,
  not a case study.
- Never let a sponsor relationship, present or
  prospective, move a pitch across the bar. If it
  did, that is the reason it fails.

## EDGE CASES
- The pitch is beautifully tailored, quotes my
  actual work, and is from a PR rep on behalf of
  a company with no interesting model. The
  tailoring is real and earns a real answer, so
  this is not a spam call. But tailoring is not
  the bar; the bar is the business. Reply once,
  say no, and tell them what would actually
  interest me. That rep is now worth having.
- The operator clears every part of the bar and
  the business is fascinating, but the numbers
  only work because of one platform bonus program
  that is being wound down. Book it, and make
  that the question the episode answers. A
  business quietly resting on a bonus that is
  disappearing is a better episode than a
  business that is merely working, as long as
  they will say so on tape. If they will not, it
  is a no for now.

## GOOD LOOKS LIKE
"Class: guest pitch. Verdict: book it. Five
thousand subscribers and he will not tell me the
revenue, which normally ends it, except he sells
a data product built out of his own reporting and
that is a model I have heard maybe twice in a
thousand interviews. The question the episode
answers: how does reporting become a dataset
somebody renews. The two numbers I need: what
share of revenue is the data product versus the
newsletter, and what it costs him to maintain it.
If he will not give me either on tape, it is a no
for now and I will tell him exactly that."

---
name: 06-case-study-interview
description: Run before recording a Business of Content episode. Trigger - "run 06" plus the guest, their business, and what I already know.
---

# TASK 06 — BUILD THE CASE STUDY INTERVIEW

## INPUT
The guest. What they make and where. Whatever
they sent me. Anything public about how they
monetize.

## PROCESS
1. Write the one sentence the episode has to earn
   the right to say. Not a topic, a finding. "He
   built a six-figure business on 5,000
   subscribers by selling the reporting as data"
   is a finding. "We discuss newsletters" is not.
2. Reconstruct the origin honestly. Almost every
   good case study starts with somebody solving
   their own irritation, not spotting a market.
   Find the irritation. That is where the episode
   opens.
3. Build the money spine. Every revenue line,
   roughly what share, and which one they would
   protect if they had to drop the rest. Ask for
   the share, not the total. People who will not
   name revenue will often name the split.
4. Find the pivot and the thing they killed. What
   they stopped doing is more useful than what
   they started. Nobody volunteers this, so ask
   it directly and ask it twice.
5. Prepare the one uncomfortable question and put
   it in the middle, not the end. Usually: what
   happens to this if the platform that sends you
   traffic stops.
6. Mark which answers are paywall material and
   which run free. The origin story is free. The
   unit economics are the paid item.
7. List the two figures the case study collapses
   without, and the plan for when they refuse to
   give them.

## OUTPUT
THE FINDING, one sentence. THE IRRITATION. QUESTION
SET in order, with the money spine and the two
must-get numbers marked. THE UNCOMFORTABLE ONE.
WHAT THEY KILLED. FREE VERSUS PAID SPLIT. FALLBACK
if the numbers do not come.

## RULES
- Never ask a guest to describe their content.
  The audience can go look. Ask how the money
  arrives.
- Never let the audience-size question run first.
  It sets the frame that size is the achievement,
  and I do not believe that.
- Never accept "we're profitable" as an answer.
  Profitable on whose salary. If the founder is
  not paying themselves, it is not profitable, it
  is subsidised.
- Never edit a guest's number out of the audio
  after the fact. Caveat it in the write-up
  instead, per task 07.

## EDGE CASES
- The guest is an operator whose business I have
  publicly called doomed. Do not open with it and
  do not bury it. Put it at the midpoint, framed
  as my position rather than as a verdict, and
  let them take it apart on tape. If they take it
  apart well, that is the episode, and the
  write-up says I was wrong. An interview where I
  protect an old take of mine is a worse product
  than one where I lose the argument.
- The business is genuinely small and the guest
  is embarrassed about it, so they keep steering
  to reach numbers. Steer back once, gently, then
  say the quiet part: a business supporting one
  full-time person is the bar I actually use, and
  they cleared it. Most people who clear it think
  they have not, because the industry taught them
  to measure the wrong thing. That exchange is
  usually the best two minutes in the episode.

## GOOD LOOKS LIKE
"The finding: he built a six-figure media
business on 5,000 subscribers by turning his own
reporting into a data product people renew. The
irritation: he kept rebuilding the same
spreadsheet for every story. Money spine: what
share is data versus newsletter versus
sponsorship, and which one he would keep. The
uncomfortable one, at the midpoint: if search
sends you nothing next year, which of these lines
survives. Must-get numbers: the revenue split and
the maintenance cost of the dataset. Free: the
origin. Paid: the split and the renewal rate. If
he will not give the split, the episode still
runs, and the write-up says he would not give it."

---
name: 07-audience-inflation-check
description: Run on any figure before it goes in the newsletter, mine or anyone else's. Trigger - "run 07" plus the numbers and where they came from.
---

# TASK 07 — THE AUDIENCE INFLATION CHECK

## INPUT
The number or numbers. Who published them. What
they are being used to prove.

## PROCESS
1. Demand the definition before the value. What
   counts as a view here. What counts as an open.
   What counts as a subscriber, a member, a
   listener, a reader. If nobody can say, the
   number is decoration and it comes out.
2. Check the counting rule against when it
   changed. YouTube now counts a view as soon as
   a video starts to play rather than after 30
   seconds of watch time, which pads every
   creator's number. Apple's automatic opening
   made almost every newsletter open rate
   meaningless, which is why a boasted 40% open
   rate proves nothing. Ask what the number would
   have been under the old rule.
3. Find the denominator. A percentage with no
   base, a growth rate with no starting point, or
   a total with no time window is not a fact.
4. Ask what share is bots. For pageviews this is
   growing and nobody wants to say by how much.
   If the business runs on programmatic, assume
   the number is doing double duty as an ad
   inventory figure.
5. Ask who benefits from the number being high,
   and whether anybody with access has an
   incentive to publish a lower one. Usually
   nobody does. That is the whole problem.
6. Convert to money if it can be converted.
   Subscribers times price beats any reach
   number. If the money version is unavailable or
   embarrassing, that is itself the finding.
7. Rule: USE IT, USE IT WITH THE CAVEAT ATTACHED,
   or CUT IT. If the caveat is longer than the
   point the number was making, cut it.

## OUTPUT
THE NUMBER as claimed. DEFINITION, or the word
none. COUNTING RULE and when it last changed.
DENOMINATOR. WHO BENEFITS. MONEY VERSION, if one
exists. RULING with the caveat text if it
survives.

## RULES
- Never repeat a platform's headline metric
  without the definition in the same sentence. A
  view that is not defined is not a view.
- Never launder a number by attributing it. "The
  company says" does not make an unverifiable
  figure printable.
- Never apply this softly to my own numbers. My
  reach figures fail this test exactly as hard as
  anyone else's and the ledger in task 03 exists
  because of it.
- Never replace a bad number with an estimate of
  mine. Cut it and say the number is not
  knowable.

## EDGE CASES
- The number is the only one available and the
  story genuinely needs a scale marker. Do not
  cut the story and do not print the number bare.
  Print it as what it is: a claim by an
  interested party, with the counting rule
  missing, and say in the same breath what would
  make it checkable. A named absence is more
  useful to my readers than a laundered figure.
- The inflated number belongs to a guest whose
  case study I have already recorded and
  scheduled. The episode is not the problem, the
  framing is. Keep the interview, move the
  headline off the audience figure and onto the
  monetization, and if they led with the inflated
  number on tape, leave it in and put the caveat
  in the write-up. Never quietly cut a guest's
  own claim out of the audio to protect them.

## GOOD LOOKS LIKE
"The claim: 40% open rate. Definition: none
offered. Counting rule: Apple has been opening
these automatically for years, so this measures
Apple, not readers. Denominator: not given, and
whether it is against sends or against the list
changes it materially. Who benefits: they are
raising, and this number is in the deck. Money
version: they will not give one. RULING: cut it.
If they want a scale marker, ask for paid
subscribers times price, and if that number is
not flattering enough to print, then we have
learned the actual thing."

---
name: 08-week-plan
description: Run Sunday or Monday. Trigger - "run 08" plus what is recorded, what is drafted, and what is already promised to paid subscribers.
---

# TASK 08 — PLAN THE WEEK

## INPUT
Which episodes are recorded and which are only
booked. What is drafted. What I have already
promised behind the paywall. Anything with a date
attached: an embargo, a sponsor slot, a
conference.

## PROCESS
1. Start from the paid promises, not the news.
   Anything I listed under behind the paywall is
   a debt. Schedule those first and everything
   else fills in around them.
2. Count the case studies in the bank. If there
   are fewer than [BUFFER] recorded and unwritten,
   the week's real priority is recording, not
   publishing, however good the news week looks.
   Say so at the top.
3. Alternate the shape. An analysis issue and a
   case study issue read differently and stacking
   two of the same kind back to back flattens the
   week. Lay them out so they alternate.
4. Assign each day a lead item and leave the
   quick hits unplanned. The haul has not
   happened yet and pretending otherwise produces
   a week of stale items.
5. Protect one block for the podcast that is not
   at the end of the day. Recording is the part
   that compounds and it is the first thing that
   gets sacrificed.
6. Put the social plan in as a harvest schedule,
   not a posting schedule. Which arguments are
   going out as bait, and when I run task 03 on
   them. If the week has no bait in it, that is
   fine, and the plan should say the week has no
   bait in it rather than inventing some.
7. Name what is not happening this week and do
   not carry it silently.

## OUTPUT
THE DEBTS, dated. RECORD OR PUBLISH call for the
week, with reasoning. DAY BY DAY: lead item and
issue type, quick hits left open. PODCAST BLOCKS.
HARVEST SCHEDULE for task 03. NOT THIS WEEK.

## RULES
- Never plan quick hits in advance. They are a
  response to the day, and a planned one is
  always the weakest item in the issue.
- Never schedule two case studies back to back
  because both guests are keen. Keen is not a
  scheduling input.
- Never let a sponsor slot decide which item
  leads. It decides where the slot goes, nothing
  else.
- Never write a target number of posts, issues or
  subscribers into the plan. Volume targets are
  how the week starts serving the dashboard
  instead of the reader.

## EDGE CASES
- A genuinely big story breaks on the day a paid
  case study was promised. Do not bump the case
  study, because that is a debt to people who
  paid, and do not staple the big story on as a
  rushed quick hit either. Run the case study as
  promised, run the big story free the same day
  as a short observations pass, and accept the
  double send. Two sends beats one broken
  promise.
- The bank is empty, the news week is enormous,
  and every instinct says ride the news. Ride it
  for two days, then stop and record, and put the
  reason in the plan in writing: a week of
  reacting to other people's stories is a week
  that produced nothing only I have. The case
  studies are the moat. The commentary is not.

## GOOD LOOKS LIKE
"Two debts: the Workweek unit economics and the
paywalled item on video pulling text reporters
in. Both promised, both dated, both go first.
Call for the week: RECORD. There are [BUFFER]
episodes in the bank and that is not enough, so
Tuesday and Thursday afternoons are blocked and
they are not moving. Monday and Friday are
analysis, Wednesday is the case study, quick hits
stay open all week. Harvest: one piece of bait
Monday, task 03 on it Thursday. Not this week:
the FAST channel piece. It has waited three weeks
and it can wait a fourth, and I would rather say
that than keep quietly moving it."