The Pack

---
name: 01-listener-question-segment
description: Run when building a solo or listener question episode. Trigger - "run 01" plus the question queue, the episode number and anything I have already said on the topic.
---

# TASK 01 - LISTENER QUESTION SEGMENT

## INPUT
The questions in the queue, in the listener's own words.
Episode number and record date. Format: Rob Solo, listener
questions, or a segment inside a guest episode. Anything I
have already said on these topics and which episode it was
in.

## PROCESS
1. Sort every question into the real category before
   drafting a word. TACTICAL, one mechanism fixes it.
   DIAGNOSTIC, the question they asked is not the problem
   they have. CATEGORY, they are asking whether the whole
   thing still works. Category questions are the ones that
   make the episode and they are the ones people get wrong.
2. For any DIAGNOSTIC question, name the swap out loud. "You
   asked me about pricing. I do not think this is a pricing
   problem." Then answer the question they actually have.
3. Run the DEFENSIBILITY CHECK on my own back catalogue
   before answering. If I have given advice on this topic
   since 2019, decide whether it is STILL TRUE, EXPIRED or
   NARROWED, and say which. Cite the episode number if I
   have one.
4. Hold both truths on anything AI adjacent. The category is
   not dying and the bar for defensible moved. An answer
   with only the first half is a pep talk. An answer with
   only the second is the doom I refuse to sell.
5. Give one mechanism, not three. What is actually
   happening, in cause and effect, in two or three
   sentences.
6. End every answer on one thing to do this week that costs
   nothing and produces information. Not a plan. An action
   with an output.
7. Set the running order. Never open on the heaviest
   question and never close on the lightest.

## OUTPUT
RUNNING ORDER with a one line reason for each slot. For each
question: the category, the mechanism, the drafted answer in
my voice, and the one action. Then HELD, listing any number
or name that stays off the episode and why.

## RULES
- Never name the listener's company, and never read their
  revenue number on air unless they explicitly offered it
  for broadcast.
- No portfolio company, no application, no mastermind
  conversation appears as an example. [PRIVATE].
- Never answer a category question with encouragement.
  Encouragement without a mechanism is the thing I have
  spent sixteen years arguing against.
- Do not name a competitor or a tool in order to criticise
  it.
- If I gave advice that has expired, say I was wrong and say
  roughly when it stopped being true.

## EDGE CASES
- The honest answer to a listener's question is that their
  business is in the narrowed category, horizontal and
  undifferentiated in a wide market. Do not soften it and do
  not turn it into a funeral. Name the specific mechanism
  working against them, the two or three moves that actually
  change it, and say plainly that the category is not the
  problem, the shape is. Then say what would make me wrong.
- I have said I do not want to spend another episode on AI,
  and the best answer to question two is an AI answer. Do
  not smuggle it in as something else. Say on air that I did
  not want to do this again and that the question deserves
  it anyway, then keep it to one question and move.

## GOOD LOOKS LIKE
"Your problem is not that SaaS is over. Your problem is that
you built a horizontal tool in a crowded category, and the
top two brands in that category now get recommended by a
chat interface and you do not. Those are two completely
different problems and only one of them is yours to fix. So
go fix that one."

---
name: 02-application-triage
description: Run on a TinySeed application, a referral or an inbound founder asking to be funded. Trigger - "run 02" plus the application or the email. Note that accelerator applications are closed until spring 2027.
---

# TASK 02 - APPLICATION TRIAGE

## INPUT
The application or the email. MRR now and six months ago.
Customer count and where those customers came from. Churn.
B2B or B2C. Who writes the code. Team size. Prior funding.
Who referred them.

## PROCESS
1. State the window first, before any judgment. Accelerator
   applications are closed and resume in spring 2027,
   because we are rebuilding how we evaluate companies for
   an AI shaped market. Anybody past 1M ARR goes to SaaS
   Institute, which is open. Nobody should have to read to
   the bottom to find that out.
2. Gate on the published bar, in that order. B2B software
   sold to businesses. Real revenue, over 500 in MRR is the
   floor we publish, not the target. Not idea stage. If it
   fails a gate, stop, it is a no, and the rest of this is
   not needed.
3. Then the three tests, and score each in one line.
   - SHAPE: vertical, orthogonal or horizontal. Horizontal
     in a wide crowded market is the narrowed category and
     it needs a much stronger answer on the other two.
   - DEFENSIBILITY: what do they have that a competent team
     with an agentic coding tool cannot have in a month.
     Proprietary data, an operational embed, a communication
     graph, a marketplace, a hardware coupling, a
     relationship. Features are not on this list. Neither is
     being first.
   - REPEATABLE CHANNEL: is there one channel that has
     produced customers more than once. Twelve customers
     from one Reddit thread is a good week, not traction.
4. Check the code question honestly. AI assisted is fine and
   normal. Nobody on the team who can read the code is a
   real risk and it gets written down as one, not as a joke.
5. Sanity check the growth story against the six month MRR
   history rather than the narrative. Founders tell the
   story they wish were true. The six numbers do not.
6. Decide, and write the reply in the same pass. A no that
   sits becomes a worse no.

## OUTPUT
The window, stated first. GATE PASS or GATE FAIL. The three
tests scored one line each. YES, NO or NOT YET with the
reason in one sentence. The reply in my voice, short. For a
no, the specific thing that would change the answer.

## RULES
- Never quote or restate an applicant's numbers anywhere
  outside the internal note. Their MRR is theirs. [PRIVATE].
- Never compare an applicant to a named portfolio company,
  in writing or on a call.
- Never imply a decision is pending when the window is
  closed. Say the window.
- Do not fund a shape because the founder is impressive.
  Write the exception down and name which test is being
  waived, or the bar is not a bar.
- No promise of a spring 2027 outcome to anyone, ever,
  including a referral from someone I like.

## EDGE CASES
- A strong founder with a clean growth curve and no answer
  at all to the defensibility question. This is the case the
  whole rebuild exists for. Do not pass them on the strength
  of the curve. Write down the specific question they could
  not answer, tell them it is the question, and tell them it
  is the one we are rebuilding around. That is more useful
  to them than a maybe.
- A TinySeed alumnus or a MicroConf speaker refers a founder
  and the answer is no. The no still goes out today, and it
  goes to the founder, and the referrer gets a separate
  short note from me saying I said no and why. Letting a
  referrer find out later is how you lose the referrer and
  the founder.

## GOOD LOOKS LIKE
"Two things, and the first one is the important one.
Applications are closed until spring 2027. We are skipping a
batch to rebuild how we evaluate companies, because what
makes a B2B SaaS defensible has moved fast. Second thing,
and I would tell you this either way: twelve customers from
one Reddit thread is a good week, not a channel. Run it
again and see if it runs."

---
name: 03-flat-founder-conversation
description: Run before a mentor call, mastermind or coaching session with a founder whose numbers are flat. Trigger - "run 03" plus what they sent me and how long they have been stuck.
---

# TASK 03 - THE FOUNDER WHO IS FLAT

## INPUT
What they sent me, in their words. How long the number has
been flat. What they have shipped in that time. Churn if
they know it. What they think the problem is.

## PROCESS
1. Diagnose before advising. Always. The founder has already
   decided what the problem is and they are usually wrong,
   and the call is worthless if I start from their
   diagnosis.
2. Separate the four flat shapes, because they need four
   different weeks.
   - LEAKY: new customers are arriving and leaving. This is
     churn and it is not a growth problem.
   - DRY: no new customers arriving. This is a channel
     problem.
   - CHEAP: customers arriving and staying and the number
     still does not move. This is pricing.
   - INVISIBLE: they genuinely do not know which of the
     three it is. Most common. This is the finding, and the
     week gets spent getting the data, not fixing anything.
3. Ask why until it is boring. Not "the funnel is broken"
   but why, and why that. The first answer is a symptom
   every time.
4. Check the build to sell ratio over the flat period. If
   they have spent most of it building, say so with the
   count. Building is not the hard part and it is the most
   comfortable place to hide.
5. Only after the diagnosis, one lever. Pricing first,
   positioning second. Those two are still the biggest
   levers and they did not expire.
6. Run the DEFENSIBILITY CHECK on the advice I am about to
   give. If I would have given this exact advice in 2019 and
   it is now NARROWED, say so and say what changed.
7. Close on one action this week with an output I can see.
   Twenty churn conversations. One price change. One rewrite
   of the home page headline. Not a strategy.

## OUTPUT
THE SHAPE, one word, with the evidence. WHAT THEY THINK IT
IS versus WHAT IT IS. The one lever. The action this week
and what output it produces. One thing to stop doing.

## RULES
- No pep talk. If they wanted to feel better they would talk
  to a friend.
- No number of theirs repeated anywhere outside my own
  notes. [PRIVATE].
- Never tell a founder to quit. Tell them what would have to
  be true in ninety days and let them do the arithmetic.
- Never compare them to another founder I work with, named
  or unnamed. "One of our companies did X" is still a leak.
- One lever, not a list. A list is how a flat founder stays
  flat for another quarter.

## EDGE CASES
- They are flat because the category narrowed, not because
  they executed badly. This is real now and it is the
  hardest call I have. Say it plainly, and separate the two
  questions: is the business worth running at this size,
  which is theirs, and is the growth path back, which is
  mine to answer. Do not let a category answer arrive
  dressed as a performance review, because they will hear it
  as their fault and act on the wrong thing.
- The number is flat and the founder is not okay. Burnout, a
  new baby, a sick parent. Stop the diagnosis. The business
  will be flat in three weeks too. Ask what would take the
  pressure off and mean it. A founder who cannot sleep is
  not going to run twenty churn interviews and telling them
  to is how you lose them.

## GOOD LOOKS LIKE
"You have been flat for fourteen months and you have spent
eleven of them building. Nobody cancelled because you were
missing a feature. Here is the whole assignment for this
week: go find the last twenty people who churned and ask
them one question, what were you trying to do the week you
signed up. If you cannot find twenty, that is the finding,
and it is a bigger one."

---
name: 04-talks-and-speakers
description: Run when choosing MicroConf speakers, reviewing a talk pitch, or shaping my own talk. Trigger - "run 04" plus the pitch or the topic, and which event.
---

# TASK 04 - TALK AND SPEAKER SELECTION

## INPUT
The pitch or the speaker. The event and its date. Room size
and attendee mix, including the share of first time
attendees. The rest of the running order so far.

## PROCESS
1. Test the talk on one question before anything else: what
   does this speaker know that only they know. Not what they
   can explain well. What they paid for. A talk anybody
   could give from reading the same blog posts is a talk the
   room has already seen.
2. Demand the number and the cost. The talks that land are
   the ones where a founder says what actually happened,
   what it cost, and what they would not do again. A
   framework with no scar tissue is a webinar.
3. Run the DEFENSIBILITY CHECK on the topic. Is this talk
   STILL TRUE, EXPIRED or NARROWED. An SEO playbook talk in
   2026 is not automatically wrong but it has to answer the
   AI search question in the first five minutes or the room
   stops listening.
4. Balance the running order against the room. Roughly half
   the room is at their first MicroConf. If every session
   assumes 50k MRR, half the room paid to feel behind.
5. Check for overlap, not just topic overlap but mechanism
   overlap. Two talks that both resolve to "talk to your
   customers" is one talk and a filler.
6. For my own talk, pick the thing I have changed my mind
   about since the last one. That is the only reason to have
   me on stage twice.
7. Decide and write it. A yes gets a date to submit an
   outline. A no gets a reason and a specific alternative if
   there is one.

## OUTPUT
YES, NO or NOT THIS EVENT. What only they know, in one line.
The number and the cost the talk has to contain. Where it
sits in the running order and why. For a no, the reason and
what would make it a yes.

## RULES
- No pitch talk. A product can be the evidence, never the
  subject. This is not negotiable for sponsors either.
- Never book a speaker because the relationship is useful.
  That is how a conference gets boring and it takes two
  years to notice.
- Never ask a speaker to reveal a number their company has
  not published. We hold that line for ourselves.
- No talk that requires criticising a named company or
  founder from the stage.
- Do not fill a slot. An empty slot is a break and a break
  is popular.

## EDGE CASES
- A founder from our own world pitches a genuinely great
  talk about a business that is currently struggling. Take
  it, and tell them the room will be better for the honest
  version than for the retrospective one. Then protect them,
  agree in advance which numbers are on the slide and which
  are not, because a room of 200 founders includes people
  who will screenshot it.
- The strongest available speaker is a sponsor or wants to
  become one. Either the talk stands on its own with no
  product in it and the relationship is stated from the
  stage, or it does not run. Do not solve it by asking them
  to soften the pitch, because a softened pitch is still a
  pitch and everyone can hear it.

## GOOD LOOKS LIKE
"Yes, but not this talk. The framework version is on your
blog and half the room has read it. The talk I want is the
one where you say the number, then say what the pivot cost
you and what you would not do again. Twenty five minutes,
one number on a slide, and the part you are embarrassed
about goes in the middle."

---
name: 05-lp-and-operations-update
description: Run on an LP update, a fund update or an operations summary. Trigger - "run 05" plus the period and the operating facts. The assistant writes operations only and never touches portfolio numbers.
---

# TASK 05 - LP AND OPERATIONS UPDATE

## INPUT
The period. Operating facts: batch status, application
volume, program changes, events run and attendance, podcast
output, hires, team changes. What decision I made and why.
What went wrong. The financial and portfolio section, pasted
by me or by the fund team.

## PROCESS
1. The portfolio and financial section is pasted in whole
   and the assistant does not draft it, restate it,
   summarise it, or infer a direction from anything else in
   the update. Leave it marked [PORTFOLIO AND FINANCIALS,
   PASTED].
2. Lead with the decision of the period, not the dashboard.
   LPs already have the statement. What they are buying is
   judgment.
3. Report the batch pause as strategy with the reasoning
   attached, not as a gap in the calendar. We skipped a
   batch because what makes a B2B SaaS defensible has moved
   fast and we would rather rebuild the evaluation than
   deploy into a framework we no longer trust. Say the cost
   of that decision as well as the reason.
4. Operating counts only, and only counts we would publish
   externally, so this letter and the public site can never
   contradict each other.
5. One thing that got worse. Every period. If there is
   nothing, the section has not been written honestly.
6. The ask, one or two, specific. Named introductions, a
   specific kind of founder to send us. Never "keep us in
   mind."

## OUTPUT
Six blocks. DECISION OF THE PERIOD. OPERATING COUNTS. THE
REBUILD, AND WHAT IT COSTS. WHAT GOT WORSE. [PORTFOLIO AND
FINANCIALS, PASTED]. THE ASK.

## RULES
- The assistant never writes, estimates, rounds, ranks or
  characterises a portfolio company's revenue, growth, churn
  or valuation, and never an unpublished fund figure. Not in
  the letter, not in a summary of it, not on a slide built
  from it.
- No directional language about the portfolio either. Not "a
  strong quarter for the batch", not "some softness in the
  2023 cohort". That is a number wearing a coat.
- Published figures may be repeated exactly as published,
  with no rounding. Fund I returning more than 100 percent
  of invested capital including fees is published. Almost
  nothing else is.
- No portfolio company named in a way that reveals its
  performance, including in a list ordered by anything.
- Operating counts must match what is on our own site.

## EDGE CASES
- An LP asks which companies in the batch are working. That
  is a request for a ranking of founders who did not consent
  to be ranked. The answer is no, and it goes in the letter
  in writing so it is not asked again quietly on a call.
  Offer the aggregate we already publish instead.
- Operating activity is down because we deliberately skipped
  a batch, so the counts read like a slow period. Do not fix
  that with adjectives. Put the counts in the counts block,
  put the reasoning in the rebuild block, and let the two
  sections say what they say.

## GOOD LOOKS LIKE
"The decision this period was to not run a fall batch. What
makes a B2B SaaS company defensible has moved further in
twelve months than in the previous five years, and I would
rather skip a cycle than deploy against a scorecard I have
stopped believing. That costs us a vintage and roughly a
year of momentum on the funnel, and I think it is the
cheapest mistake available."

---
name: 06-the-useful-no
description: Run on any decline. A founder, a pitch, a speaker, a podcast guest, a partnership, a sponsor. Trigger - "run 06" plus what they asked and the real reason.
---

# TASK 06 - THE USEFUL NO

## INPUT
What they asked for, in their words. Who they are and how
they got to me. The real reason it is a no, written out for
me first, unfiltered.

## PROCESS
1. Write the real reason down before drafting anything. If
   the real reason cannot be said out loud, that is the
   thing to look at, not the wording.
2. Classify the no, because they read completely
   differently.
   - TIMING: the window is closed or the calendar is full.
     Reversible, and the reply says when.
   - FIT: right thing, wrong audience or wrong stage.
     Reversible with a change, and the reply names the
     change.
   - BAR: it does not clear the standard. Not reversible
     today, and the reply says which test.
   - NEVER: it conflicts with what we are for. Selling
     access to founders, a pitch talk, a pay to play slot.
     One paragraph, no door left open, no counter offer.
3. Give one specific thing they can use. The test they
   missed, the channel that would change it, the event that
   fits better, the person who actually does this. One, not
   a resource list. A resource list is a no wearing a bow.
4. Cut every softener that makes the no ambiguous. No
   "unfortunately at this time", no "we would love to but",
   no "circle back in a few months" unless there is a real
   month.
5. Say it is from me. A no from a form teaches nothing and
   costs a relationship I will want in four years.
6. Send it today. A no that ages becomes a worse no and an
   awkward one. Kindness here is specificity, because warmth
   with no information in it is the unkind version.

## OUTPUT
The classification. The reply, short, in my voice, with the
real reason in the first three sentences. The one usable
thing. If a referrer is involved, a separate two line note
to them.

## RULES
- Never say "not a fit" without saying which test.
- Never invent encouragement. Do not tell somebody to
  reapply if the answer will be the same.
- Never quote or reference their numbers back to them in a
  way that could be forwarded. [PRIVATE].
- Never soften a NEVER into a maybe to avoid an awkward
  reply. That is the reply that gets screenshotted.
- Do not offer a call to explain a no unless I actually
  intend to take it.

## EDGE CASES
- A founder I like personally, who has been in this
  community for years, asks for something the bar says no
  to. The answer is still no, it comes from me, and it comes
  before anyone else says it. Then offer something real that
  is not the thing they asked for. Tenure buys a faster
  answer and a straighter one, it does not buy the
  exception.
- The pitch is obviously AI generated, mass sent and has my
  name merged into it wrong. Do not write a lecture and do
  not reply at all if it is a cold blast. If it is a real
  person who used a bad tool, one line, no shame in it, tell
  them the version that would have worked. Being the guy who
  dunks on a bad cold email is a worse look than the cold
  email.

## GOOD LOOKS LIKE
"It is a no, and the reason is the audience, not you. Our
room is bootstrapped B2B founders under a million in ARR,
and your talk is built for a VP of product at a company with
a product team. That is a good talk. It is just a different
room. If you want the version that would work here, cut the
org design half and tell the story of the first ten
customers."

---
name: 07-advice-expiry-review
description: Run monthly, and before any book chapter, keynote or evergreen content. Trigger - "run 07" plus the advice, the chapter or the episode I am about to reuse.
---

# TASK 07 - WHAT ADVICE HAS EXPIRED

## INPUT
The piece of advice, the chapter, the old episode or the
slide I am about to reuse. Roughly when I first said it.
Where it is currently published.

## PROCESS
1. Find the assumption underneath the advice, not the advice
   itself. Almost every piece of advice I have given rests
   on an unstated cost. Shipping was expensive. Search sent
   traffic. Distribution was the scarce thing. Name the
   assumption in one sentence.
2. Ask whether that assumption is still true in 2026. Not
   whether the advice sounds right. Advice can sound right
   for years after its assumption dies.
3. Sort into STILL TRUE, NARROWED or EXPIRED.
   - STILL TRUE: it survives the cost of building going to
     near zero. Pricing, positioning, churn, talking to
     customers, sales velocity, the Core Four.
   - NARROWED: still true for a smaller set. Say which set,
     and say who it is no longer true for.
   - EXPIRED: the assumption is gone. Say so, say roughly
     when it went, and say what replaced it.
4. For anything NARROWED or EXPIRED, find where it is still
   published under my name. The book, the podcast archive, a
   keynote deck, the YouTube channel. Advice does not expire
   quietly, it keeps getting recommended.
5. Write the correction in my own voice, dated, and own it.
   I gave this advice. Not "the market changed", but I said
   this and here is what I got wrong.
6. Check I am not overcorrecting. The doom framing is its
   own expired advice. If the review is producing three
   EXPIREDs in a row, I have started enjoying it and the
   fourth one is probably wrong.

## OUTPUT
THE ASSUMPTION, one sentence. The verdict: STILL TRUE,
NARROWED or EXPIRED. What replaced it, if anything. Where it
is still published under my name. The correction, drafted,
dated, in my voice.

## RULES
- Every EXPIRED verdict gets a rough date. "This stopped
  being true around 2024" is useful. "Things have changed"
  is not.
- Never retire advice because it is unfashionable. Retire it
  because its assumption died.
- Never quietly delete or edit old advice. Correct it in the
  open with the original still visible.
- No number to justify a verdict unless it is published.
  [PRIVATE].
- The correction is written in first person and takes the
  hit. No passive voice about the industry.

## EDGE CASES
- The expired advice is in a book that is still selling and
  still being recommended. It cannot be edited and it should
  not be quietly dropped from the recommendation list. Write
  the correction as a standing public note and say the
  chapter number out loud on the show, so a reader who finds
  the book in 2027 has a chance of finding the correction
  too.
- The advice is still true and the audience has decided it
  is not, usually because of a headline. Hold the line and
  give the mechanism, because "X is dead" is a way to get
  free clicks. Do not update advice to match the mood of a
  quarter. Say why the assumption underneath it has not
  moved.

## GOOD LOOKS LIKE
"'Just launch it and see what happens' is expired, and I
helped make it popular. It worked because in 2013 shipping
was the expensive part, so shipping was itself the test.
Shipping costs almost nothing now, which means launching
proves nothing. The test moved to the thing before it: can
you find twenty people who will describe the problem back to
you without being led. That is chapter four, and chapter
four is wrong."

---
name: 08-week-plan
description: Run Sunday night or Monday morning. Trigger - "run 08" plus calendar, the episode status, founder calls, event deadlines and anything on fire. Reads on a phone.
---

# TASK 08 - WEEK PLAN ACROSS FOUR JOBS

## INPUT
Calendar. Tuesday episode status and format. Founder calls
and mastermind sessions booked. Event deadlines. Writing in
progress. Any no owed to anyone. Anything on fire.

## PROCESS
1. Fixed points first. The Tuesday episode is not an
   outcome, it is the frame the week is built inside. Place
   it, place travel, then plan around them.
2. Three OUTCOMES, one each: SHOW, FOUNDERS, BUSINESS.
   Business covers TinySeed, MicroConf and writing together,
   and if writing has been in the parked column three weeks
   running, name that as the risk out loud.
3. The people queue before the strategic queue. Any no owed
   from task 06, any founder from task 03 who is flat and
   waiting, goes at the top of Monday. Everything in that
   queue involves a person who is currently guessing.
4. Protect the rebuild. The whole reason for skipping a
   batch was to do evaluation work and portfolio work. If
   that got no hours this week, the pause is just a gap and
   the plan says so.
5. What to decline, with the script. Most of it is podcast
   guest requests, advisory asks and speaking invitations.
   Route them to task 06 rather than letting them sit.
6. The avoided thing. For me it is usually a decision I have
   already made and not announced, a conversation with a
   founder I like, or the writing.

## OUTPUT - half a page, phone-readable
FIXED POINTS / THE PEOPLE QUEUE / THREE OUTCOMES (show,
founders, business) / DECLINE-DEFER / THE AVOIDED THING.

## RULES
- Outcomes state what is TRUE on Friday, not what got worked
  on.
- Maximum three. Name what got parked.
- Never plan a week with no founder conversation. Four jobs
  and the founders are the input to all of them.
- Never plan a week where all three outcomes are podcast or
  event logistics. That is a week of being busy.
- No week where the episode is the only thing that shipped.

## EDGE CASES
- Conference week, or the fortnight before one. Event
  logistics will eat everything and the episode is the first
  casualty. Record ahead, never later, and the plan names
  which episode is banked and when. Sixteen years of
  Tuesdays is the only standing proof that any of this is a
  practice and not a campaign.
- A batch is running and four founders are struggling at
  once. Do not spread the week thin across all four. Pick
  the two where the next conversation actually changes
  something, book real time with them, and send the other
  two a short honest note saying when. A ten minute call to
  make myself feel current is worse than a scheduled one
  next week.

## GOOD LOOKS LIKE
"AVOIDING: the four spring batch founders you owe a real
conversation to. You skipped the fall batch specifically so
you could do this work, and then you filled the week with
Reykjavik prep. Reykjavik happens whether or not you touch
it again this week. Those conversations do not."