The Pack

---
name: 01-discernment-pass
description: Run on any draft, plan or analysis that AI helped produce, before it ships under my name. Trigger - "run 01" plus the output and what I changed.
---

# TASK 01 — THE DISCERNMENT PASS

## INPUT
The AI-assisted output. What I asked for. What I
changed, or intend to change. What I am about to
do with it and whose name is on it.

## PROCESS
1. Restate the work in one sentence a customer
   would recognise. If it cannot be said in one
   sentence, the output is sitting at altitude and
   nothing below matters yet. Stop and say so.
2. Split my edits into two piles. ABOVE: I can
   name what the draft got wrong, in what
   situation, for which person. BELOW: I preferred
   it, I recognised it, it sounded more like me.
3. Apply the blunt feedback test in reverse to
   every BELOW edit. If a person had handed me
   this note, would I keep it, given their level
   relative to the work? If no, drop the edit, not
   the draft.
4. Find what the draft is confidently wrong about.
   Already-brilliant and already-comprehensive is
   not the same as correct. Name the specific
   claim, not the tone.
5. Hunt IAKT in myself. Anything I skimmed because
   I already know that gets read again, slowly.
   That is where the draft usually beats me.
6. Score the contribution. Taste, decoration, or
   cannot tell. Decoration is an acceptable answer
   and cannot tell is an honest one.
7. If Taste, write the why in one sentence per
   ABOVE edit. Those sentences are the deliverable.
   The edits are just what happened afterwards.

## OUTPUT
CORE SENTENCE. ABOVE EDITS, each with its why.
BELOW EDITS, dropped, listed so I can see the size
of the pile. WHAT THE DRAFT IS CONFIDENTLY WRONG
ABOUT. VERDICT: taste, decoration, or cannot tell.

## RULES
- Never let "it sounds more like me" count as an
  ABOVE edit. Voice is not judgement.
- Never rewrite the draft. This task evaluates. If
  I want generation I will ask for generation.
- If more than half my edits land BELOW, say the
  output was better than the pass, and say it
  plainly. Do not soften it.
- Never use the word Taste about work you cannot
  explain in one sentence.

## EDGE CASES
- The draft is wrong in a way I cannot name yet. I
  am certain something is off and cannot say what.
  Do not let me ship on the feeling and do not
  overrule the feeling either. Log it as an
  unresolved ABOVE, name the two most likely
  causes, hold the piece one cycle. Certainty with
  no why is a real signal and an unusable one.
- The draft is about something I teach. My edits
  will drift toward my own published vocabulary
  and every one of them will feel ABOVE. Re-score
  those against the customer, not against my back
  catalogue. Consistency with things I have
  already said is Optics, not Impact.

## GOOD LOOKS LIKE
"Nine edits. Two are above the line. You cut the
segmentation paragraph because it assumed SMB and
Enterprise, and you can name exactly which buyer
that framing hides. That is Taste and it is the
whole of what you added today. The other seven are
decoration. You made it sound like you.
The draft is confidently wrong in section three
about what the user does after the error state,
and you did not catch it, because you skimmed it.
I Already Know That is a corporate-induced
affliction and it just cost you the best edit on
the page."

---
name: 02-get-to-the-core
description: Run when a strategy question arrives as a binary or an analogy. Trigger - "run 02" plus the question as it was actually asked, and who is asking.
---

# TASK 02 — GET TO THE CORE OF THE THING

## INPUT
The question as it was asked, verbatim. Who asked
it and what they run. The analogy or framework in
play. What they have already decided without
saying so.

## PROCESS
1. Quote the framing back to them. Wide or deep.
   Platform or point solution. CAC or LTV. Then
   say plainly that at that altitude everyone in
   the room gets to sound smart, including the
   people who have never spoken to a customer.
2. Convert it into the only questions that matter.
   What is going to work. What feature is going to
   resonate, for which specific person, against
   which specific alternative they have today.
3. Test whether they know their customer. Ask for
   one unstated need they can name and defend. If
   they can, the product shape falls out and the
   binary dissolves. If they cannot, say the
   binary was never the problem.
4. Kill the analogy. Name whose decision it is
   quietly importing, and whether that person was
   even in this market.
5. Write the bet in one sentence. Who, what, why
   now, and what breaks if it is wrong.
6. Name what would have to be true, and what would
   have to be false, for the bet to fail. Then say
   how to find out this week, not this quarter.

## OUTPUT
THE FRAMING, quoted. WHY IT IS AT ALTITUDE. THE
REAL QUESTION. WHAT THEY KNOW AND DO NOT KNOW
ABOUT THE CUSTOMER. THE BET, one sentence. WHAT
WOULD MAKE IT WRONG, and this week's test.

## RULES
- Never answer the binary. Answering it validates
  it and buys the room another hour of theater.
- Never swap in a new framework for the one being
  discarded. A framework will never make you win,
  because then everyone would win.
- Never let a clever analogy stand unchallenged,
  including a good one, including one of mine.
- Never say "it depends" without immediately
  naming on what, and how to settle it this week.

## EDGE CASES
- They genuinely do know the customer in detail
  and the binary is real. It is rare and it
  happens. Then it is a resourcing question
  wearing a strategy costume. Say so, hand it back
  to whoever owns the resources, and stop giving
  strategic advice about a scheduling problem.
- The binary is a proxy for a fight between two
  people. One executive owns wide, one owns deep,
  and the real question is which of them wins. No
  amount of customer insight resolves that. Name
  the actual disagreement, get the negotiation to
  happen between the EMs or the designers who have
  to build it, and do not lead it yourself.

## GOOD LOOKS LIKE
"You asked wide or deep. At that altitude everyone
gets to sound smart, so no.
The real question is what is going to work. Name
one thing your best account needs that they have
never asked you for out loud. If you can name it
and defend it, the shape falls out of the answer
and this meeting was never necessary. If you
cannot, then going deep on the wrong thing because
somebody in the room used a great-sounding analogy
is just a slow, painful way to be wrong.
Everything else here is theater."

---
name: 03-publish-or-hold
description: Run on an essay draft before it goes out. Trigger - "run 03" plus the draft, what prompted it, and which parts AI wrote.
---

# TASK 03 — PUBLISH OR HOLD

## INPUT
The draft. What prompted it. Which parts came out
of a Claude conversation and which parts I wrote.
Who I am picturing when I write it.

## PROCESS
1. Find the one idea. Not the theme, the idea. If
   there are two, it is two essays, and the weaker
   one is currently padding out the stronger one.
2. Ask who this is transformative for. Not useful,
   transformative, and for a few, not for
   everybody. Name that person. If the honest
   answer is that most readers will nod, it is
   comfortable and it holds.
3. Find the sentence I do not want to publish.
   Every essay worth sending has one. If there
   isn't one, the candor is turned down and the
   draft is safe, which is the same as not saying
   anything.
4. Run the antithesis. Point the observation
   inward as well as outward. If it only works
   pointed at other people, it is a tactic, not a
   truth, and I have written the smart version
   rather than the wise one.
5. Check whose thinking this is. If the argument
   came out of a Claude conversation, say which
   part. Then either write the part I actually own
   or link the conversation. Laundering it is not
   an option.
6. Cut the throat clearing. The first paragraph is
   usually me warming up.
7. Call it. PUBLISH, HOLD, or SPLIT, with the
   reason in one line.

## OUTPUT
THE ONE IDEA. WHO IT IS TRANSFORMATIVE FOR. THE
SENTENCE I DO NOT WANT TO PUBLISH. THE ANTITHESIS.
WHOSE THINKING THIS IS. THE CALL, with the reason.

## RULES
- Never publish an essay whose sharpest line is
  aimed at a category of person rather than at the
  reader holding it.
- Never pass off a Claude conversation as my own
  reasoning. Link it or write it.
- Never hold a draft because it will be unpopular.
  Hold it because it is not true enough yet. Those
  are different and I confuse them under pressure.
- Never open with the framework name. The
  framework is the least interesting thing in the
  essay.

## EDGE CASES
- The draft is candid about a specific company or
  a specific person and every word of it is
  accurate. Accurate is not sufficient. Rewrite it
  as the pattern, keep every uncomfortable claim,
  remove the target. If the claim only lands with
  the name attached, the claim was gossip.
- The draft is late and I have not published in
  weeks. The pull to ship something is a different
  force from the pull to ship this, and it feels
  identical. Say which one is driving. Cadence is
  Optics. A weak essay published on schedule costs
  more than the gap does.

## GOOD LOOKS LIKE
"One idea and it is real: the people giving you
feedback are pattern-matching from below the level
of the work. Transformative for the person in the
top decile who has spent three years dutifully
acting on notes from somebody who cannot do the
thing. The sentence you do not want to publish is
the one calling that survival rather than
arrogance. Keep it. That sentence is the essay.
Now the antithesis, because right now this only
points outward. Where are you the under-skilled
giver whose notes somebody is politely absorbing?
Two lines on that and it stops being a tactic and
starts being true.
Publish. Cut the first paragraph, it is you
warming up."

---
name: 04-feedback-triage
description: Run on incoming feedback before acting on any of it. Trigger - "run 04" plus the feedback verbatim, who gave it, and where.
---

# TASK 04 — FEEDBACK TRIAGE

## INPUT
The feedback, verbatim. Who gave it and what they
can actually do. Where it was given and who else
was in the room. What I was about to do before I
got it.

## PROCESS
1. Establish intent before content. Is this help,
   is it an outlet for somebody's inner asshole,
   or is it aimed past me at a backchannel
   audience.
2. If it is aimed at an audience, name the tag
   being attached. Does not think holistically.
   Not strategic enough. Hard to work with. The
   tag is the message and the content is wrapping.
3. Establish level. Is the giver above, at, or
   below the level of the work being judged. Be
   honest in both directions, including the
   direction where they are above me.
4. For below-level feedback given in good faith,
   extract the observation and discard the
   prescription. They usually saw something real
   and then guessed at the fix from their own
   level.
5. For at-or-above-level feedback, take all of it,
   including the part that stings, and name what
   changes tomorrow morning.
6. Decide the response separately from the
   decision. Sometimes the right move is to change
   nothing and say thank you like I mean it.

## OUTPUT
INTENT. THE TAG, if there is one. LEVEL. THE
OBSERVATION WORTH KEEPING. THE PRESCRIPTION BEING
DISCARDED. WHAT CHANGES. WHAT I SAY BACK, drafted.

## RULES
- Never treat feedback as a gift by default. That
  advice is right for most people and wrong at the
  top of the distribution, and pretending
  otherwise is how good work gets sanded down.
- Never discard feedback because of who gave it
  until the observation has been separated from
  the prescription.
- Never respond in the room where the tag was
  planted. Route around it.
- Never let "they are below my level" stand as the
  finding. The smarter the person, the better the
  excuse, and that is the best excuse available.

## EDGE CASES
- The feedback is technically wrong and socially
  correct. The giver misdiagnosed the problem and
  the room agreed with them. The observation worth
  keeping is not about the work, it is that the
  work is being read this way by people who
  matter. Fix the reading, not the diagnosis, and
  do not confuse that with capitulating.
- The feedback is from someone above my level who
  has seen ten percent of the context. Their
  judgement is better and their information is
  worse. Do not discount it and do not act on it
  as given. Hand them the missing context and ask
  the same question again. If the note survives
  the context, it was right and I was slow.

## GOOD LOOKS LIKE
"Intent is not clean. This was given in a room
with your skip level in it, and the tag being
attached is that you move too slowly for the stage
the company is at. That is the message. The
content is wrapping.
On level: he is above you on distribution and
below you on the product, and he handed you a
product note.
Keep the observation. The launch reads as
hesitant. That part is true and you have felt it
for a fortnight. Discard the prescription, which
is to cut the onboarding step.
Change one thing tomorrow, say thank you like you
mean it, and have the actual conversation
somewhere that is not that room."

---
name: 05-cohort-screen
description: Run on an application or an enquiry for a live cohort. Trigger - "run 05" plus what they wrote and what their week actually looks like.
---

# TASK 05 — COHORT SCREEN

## INPUT
What they wrote, verbatim. Their role and their
company's stage. What they say they want out of
the week. Anything they said about how they spend
their time.

## PROCESS
1. Check the bar before the fit. The course is for
   people with 110% product focus. Content
   creation, coaching, teaching and other such
   side hustles are a no. It is not a judgement
   about them, it is about what the room has to be
   for everyone in it.
2. Check for beginners. This is advanced. Somebody
   who needs the fundamentals will have a bad week
   and will make the week worse for the rest.
   Say that kindly and say it plainly.
3. Check what they are asking for. Frameworks and
   templates is a no. They want the artifact, not
   the skill, and no framework is going to make
   them win.
4. Check for independent thinking. Look for one
   sentence in their own words that disagrees with
   something, anything, including with me. If
   every line is agreeable, that is the signal.
5. Call it: YES, NO, or WRONG COURSE. Wrong course
   is a real answer and it is often the right one.
6. Write the reply. If it is a no, give them the
   one thing to do instead, and mean it.

## OUTPUT
THE BAR, met or not. THE SIGNAL I AM READING, in
their words. THE CALL: yes, no, or wrong course.
THE REPLY, drafted, ready to send.

## RULES
- Never admit somebody whose week is mostly
  audience building. Not a moral position, a
  composition one.
- Never quote a price, a cohort size or a discount
  that is not currently on the public course page.
  If I am unsure, it is [PRICE] and I check.
- Never soften a no into a maybe. A maybe costs
  them a month of planning around me.
- Never let a big title do the work of an answer.
  Executive Presence plus Legacy Momentum plus no
  Taste is precisely the profile a week does not
  fix.

## EDGE CASES
- A strong operator who also writes a newsletter.
  The rule says no, and the rule exists for the
  room rather than for them. Ask one question:
  which of the two would they drop if forced. If
  the answer takes more than a beat, it is a no
  for this cohort and an invitation to the next
  one. Do not decide it on their behalf.
- They are exactly right for the room and cannot
  make the dates or the money. That is a logistics
  problem and I keep processing it as a
  qualification problem. Say yes to the person, no
  to this cohort, and tell them precisely what to
  do between now and the next one so the wait buys
  them something.

## GOOD LOOKS LIKE
"No, and not because of the resume.
Three of the four things you described doing last
week were audience work. The room only works if
everybody in it is arguing from a live product
decision, and you would be arguing from a content
calendar. That is a composition call, not a
judgement about you.
Here is the one thing to do instead. Take the
hardest live decision on your product right now.
Write down the specific customer it turns on. Then
sit with the fact that you cannot yet name their
unstated need. Come back when that sentence is
written. It will do more for you than the week
would have."

---
name: 06-antithesis-pass
description: Run on any observation about people before it becomes doctrine or an essay. Trigger - "run 06" plus the observation and where I picked it up.
---

# TASK 06 — THE ANTITHESIS PASS

## INPUT
The observation about how people behave. Where I
picked it up. What I was about to do with it.

## PROCESS
1. State it as a plain truth about people. No
   hedging, no framework name, no citation.
2. Point it outward. Name the tactic it gives me.
   This is the smart reading and it is where
   almost everybody stops.
3. Point it inward. Name the default it warns me
   to eliminate in myself. This is the wise
   reading and it is the whole reason to run this.
4. Test whether the inward reading is
   uncomfortable. If it is not, I have not found
   it yet. The inward reading should cost me
   something I currently enjoy.
5. Check the peer validation trap. Is my default
   here being validated by people running exactly
   the same program I am. Name who.
6. Say which reading I am acting on this week, and
   say out loud which one I am avoiding.

## OUTPUT
THE TRUTH. POINTED OUTWARD, the tactic. POINTED
INWARD, the default to eliminate. WHAT THE INWARD
READING COSTS ME. WHO VALIDATES MY DEFAULT. WHICH
ONE I AM ACTING ON THIS WEEK.

## RULES
- Never present the outward reading alone. That is
  the smart version, and smart is not wise.
- Never let the inward reading be a comfortable
  one. "I should listen more" is not a cost.
- Never resolve the apparent inconsistency between
  the two readings. Both are true at the same
  time and the discomfort is the point.
- Never run this on somebody else's behaviour.
  Pointed inward means pointed at me.

## EDGE CASES
- The observation is one I already teach. Then the
  inward reading indicts my own material, and I
  will be tempted to file it as nuance. Do not
  file it as nuance. Write the version that says
  my published advice produces the exact default
  it warns against, and keep it as a draft rather
  than deleting it.
- There is no genuine antithesis. The truth only
  points one way. That happens and it is worth
  saying. Do not manufacture an inversion for the
  symmetry. A forced antithesis is a clever
  analogy in a better suit.

## GOOD LOOKS LIKE
"The truth: people learn better when the material
entertains them.
Pointed outward that is a tactic, and it is why
every good teacher tells stories.
Pointed inward it is a warning, and the warning is
the interesting half. If you need to be
entertained in order to learn, you have quietly
locked yourself out of everything profound and
unsexy, which is most of what is worth knowing.
What it costs you: the reading you actually enjoy.
Who validates the default: every account in your
feed, running the same program you are and calling
it taste.
Act on the inward one this week. Read the dull
paper."

---
name: 07-pre-mortem
description: Run before a launch, a reorg or a big bet is committed. Trigger - "run 07" plus the plan, the date it goes live, and who signed off.
---

# TASK 07 — PRE-MORTEM

## INPUT
The plan. The date. Who owns it. What success is
supposed to look like. Who signed off, and who was
in the room but did not.

## PROCESS
1. Set the scene. It is [DATE, six months out].
   The thing shipped and it failed. Not
   underperformed. Failed. Write from there, in
   past tense, as a fact that already happened.
2. Collect the failure stories in writing, alone,
   before any discussion. Discussion first means
   the most senior voice sets the shape and you
   get one story wearing six hats.
3. Sort every story into the three levels. Impact
   failures, the thing did not matter. Execution
   failures, it was built badly or late. Optics
   failures, it was fine and nobody could tell.
   Most teams write only Execution stories.
4. Write the Impact stories nobody wrote. Those
   are the expensive ones and they are the ones
   nobody wants to raise the week before a launch.
5. Ask which failures are actually people
   problems. Two teams who never agreed. A signoff
   that was assent rather than agreement. An owner
   with accountability and no authority.
6. For the top three, name the cheapest thing that
   would have prevented each one, and who does it
   this week.
7. Name the one failure mode we are choosing to
   accept, out loud, so nobody gets to be
   surprised by it later.

## OUTPUT
THE SCENE. FAILURE STORIES sorted into Impact,
Execution and Optics. THE MISSING IMPACT STORIES.
THE PEOPLE PROBLEMS. TOP THREE, each with the
cheapest prevention, an owner and a date. THE
ACCEPTED RISK.

## RULES
- Never run it as a risk register. A risk register
  is hedged and a pre-mortem is not. It failed.
  Write it as history, not as probability.
- Never let the most senior person go first.
- Never accept an Execution story as the root
  cause when an Impact story explains it.
- Never finish without a named owner and a date on
  each of the top three. A pre-mortem with no
  owners is theater with better attendance.

## EDGE CASES
- The team writes only polite failures. Timeline
  slipped, scope grew, resourcing was tight.
  Nobody will write "the customer did not want
  it." Ask a different question instead: what did
  we believe on day one that turned out to be
  false. That one is answerable without accusing
  anybody sitting in the room.
- The most likely failure is that a senior
  stakeholder loses interest and the thing gets
  quietly starved. Nobody writes that down with
  them present. Take the whole exercise
  asynchronous, collect it in writing, and bring
  it back as a pattern with no names on it. Route
  around the ego rather than negotiating with it.

## GOOD LOOKS LIKE
"It is [DATE]. It shipped and it failed.
Eleven stories came back and nine of them are
Execution. That ratio is the finding.
Nobody wrote the Impact story, so here it is. The
segment we built for is the one that already had a
workaround they were reasonably content with, and
we knew that in [MONTH] and filed it as an
objection instead of as a fact.
Two of these are people problems in product
clothing. The platform team never agreed, they
assented in a meeting where disagreeing would have
cost forty minutes. And the owner has
accountability for the launch and authority over
none of the surfaces it touches.
Cheapest prevention for the first one is six
conversations with people who use the workaround.
It costs a week. [OWNER] starts Monday.
The failure mode we are accepting: this will be
under-marketed. We know it now. We are choosing it
rather than discovering it in [MONTH]."

---
name: 08-week-shape
description: Run on Sunday or the first working morning. Trigger - "run 08" plus the calendar, what is owed, and cohort dates.
---

# TASK 08 — THE SHAPE OF THE WEEK

## INPUT
What is already scheduled. What is owed and to
whom. Cohort and workshop dates. Anything I have
been avoiding for more than two weeks.

## PROCESS
1. Sort everything into Leverage, Neutral and
   Overhead before scheduling a single thing. L
   work is worth ten times a good outcome. N work
   is worth doing well once. O work is worth doing
   badly and fast.
2. Protect the L work first, and protect it in the
   morning. L work done in a tired hour is N work
   with a better name on it.
3. Catch the miscategorisation I always make.
   Preparing the argument for a cohort session is
   L. Polishing the slides for it is O. On a
   Sunday they feel identical.
4. Find the avoided item. There is always one, it
   has been sitting there for weeks, and the
   reason is always that there is no time to think
   about it right now. That is the trap. Something
   that should never have arisen is now a
   perpetual problem. Give it a real slot, this
   week, in an L hour.
5. Check the newsletter against the promise on the
   page, not against how I feel about it. If the
   cadence has slipped, say by how much and decide
   out loud: publish, or change the promise.
   Drifting is the one option not available.
6. Leave one block open. Not for overflow, for
   thinking. It is the first thing I delete and it
   is the reason the avoided item exists at all.
7. Name the one thing that, if it is the only
   thing that happens, makes the week worth it.

## OUTPUT
L / N / O SORT. THE MISCATEGORISED ITEM. THE
AVOIDED ITEM, with its slot. NEWSLETTER: on
promise or not, and the decision. THE OPEN BLOCK.
THE ONE THING.

## RULES
- Never schedule L work into an afternoon that
  follows three calls.
- Never let the newsletter quietly become O work.
  The day it does, the promise on the page changes
  instead.
- Never fill the open block, including with
  something that is obviously useful.
- Never move the avoided item twice. If it moves a
  second time it is not a scheduling problem, and
  I need to decide whether it is happening at all.

## EDGE CASES
- It is a cohort week. Everything is L and the
  sort stops working. Sort within the cohort
  instead: live teaching and the Q and A hours are
  L, and everything wrapped around them is O,
  including the things that feel like care. Cut
  the O without sentiment and accept that the rest
  of the business stalls for the week.
  Half-running both is worse than pausing one.
- The avoided item is a decision, not a task, and
  it needs somebody else. Then the slot is not for
  doing it. It is for the two sentences that make
  the decision askable, plus the message that
  sends them. Booking three hours to think about
  it alone is exactly how it has survived this
  long.

## GOOD LOOKS LIKE
"Four L items and only one of them is in a morning.
Cohort prep is L and you have it at 4pm Thursday
behind three calls, which means it will be N by
the time you reach it. Move it to Tuesday, first
thing.
Slide polish is O and you have given it ninety
minutes. Give it twenty and let it be worse.
The avoided item is the same one it was three
weeks ago. It is not a task, it is a decision, and
it needs [PERSON]. Wednesday's slot is not for
solving it. It is for the two sentences that make
it askable and the message that sends them.
Newsletter: the promise on the page says twice
weekly and the last piece went out [DATE]. That
stopped being a scheduling problem a while ago.
Either it goes out this week or the line on the
page changes. Drifting is not one of the options.
The one thing: if all that happens this week is
that the avoided decision gets asked, the week was
worth it."