LAUNCHED · NOW RUN THE TEST
The launch is not the test. The loop is.
Twelve positions are now a six-week operating practice: Michael originates and reacts; the system produces and measures; two unread dispatches stop the machine.
LIVE
7-post X thread
View verified launch ↗
NEXT 96 HOURS
One completed move. Four moves with owners.
- 01 complete
- 02 next
Tue Jul 28 · Michael
Choose five people and send the chapter-specific note that matches each reader.
Open working surface - 03 queued
Wed Jul 29 · PointCast desk
Publish “Personality must survive Wi‑Fi” as the first standalone signal.
Open working surface - 04 queued
Thu Jul 30 · PointCast desk
Publish the graveyard question and invite personal stories rather than applause.
Open working surface - 05 queued
Fri Jul 31 · Michael + Codex
Record replies, DMs, saves, chapter clicks, and the first editorial read.
Open working surface
DECISION RIGHTS
The organization is small enough to name every responsibility.
Originator + editor
Michael Hoydich
- The position, the voice, and every kill or reorder decision
- One concrete specification, scene, or story before each dispatch
- A one-line reaction after reading each dispatch
- Weekly reporting of replies, DMs, saves, and invitations
Production + instrumentation
Codex / OpenAI
- Question prompts, drafts, social compression, art direction, and implementation
- Human and machine-readable publishing surfaces
- Reaction logging, scorecard maintenance, testing, and release verification
- Pausing production after two consecutive unread dispatches
Editorial system credit
Sol / ChatGPT
- The initial twelve-position arc and read-loop structure
- Protecting origination and curation as the test variable
- No claim to Michael’s first-person experience or private product knowledge
Signal, not focus group
Readers
- Specific replies, objections, memories, saves, and invitations
- Evidence about which claims travel
- No authority to flatten the voice into consensus
THE SIX-WEEK SPINE
Tuesday is the position. Thursday is the complication.
Every week adds one thing the internet cannot supply: Michael’s concrete specification, scene, or refusal.
Creature before interface
Which claim earns curiosity without the reader needing category context?
Manufacturing and the shelf
Did concrete manufacturing or retail detail change how seriously readers took the thesis?
The ownership architecture
Mid-test review: are architecture and custody outperforming the softer cultural claims?
Consequence and authored IP
Do readers understand consequence as emotional design and editions as publishing?
Economics and mortality
Does the graveyard create stories, objections, or invitations that the business-model claim does not?
Culture and the continuing creature
Verdict memo: did Michael-originated curation outperform generic AI content, and is there a print/media path?
Tue Sep 1 · #11 Children Will Learn AI Ethics From Pets Before They Learn Prompting
A child’s first AI lesson will be a relationship, so the design must teach boundaries rather than domination.
Thu Sep 3 · #12 We Don’t Need Infinite Characters. We Need One Creature Worth FollowingGenerative abundance makes authorship more valuable, not less.
NON-NEGOTIABLE
Two unread dispatches stop production.
If Michael skips reading two dispatches in a row, production pauses. An unread series cannot test authored curation.
- 01
Before
Ask no more than three questions. Extract Michael’s position, one concrete detail, and the line he refuses to soften.
- 02
Publish
One dispatch, one X-sized signal, one art plate, one human page, and one machine-readable update.
- 03
After
Michael reads the dispatch and gives one line: stronger, weaker, wrong, or the missing sentence.
- 04
Resume
Resume only after the missing reactions are logged or Michael explicitly kills those dispatches.
RUNNING SCORECARD
Empty cells are honest cells.
Reactions and external signals stay blank until Michael reports them. The office measures evidence, not vibes.
Dispatch Michael reaction External signal Editorial read
W1 · Tue Jul 28 #1 The Next Great AI Product Will Be a Pet, Not an Assistant Awaiting read Not reported Not written
W2 · Tue Aug 4 #3 The Face Is the Wrong Place to Spend the BOM Awaiting read Not reported Not written
W2 · Thu Aug 6 #4 Don’t Sell Digital Pets Where Laptops Are Sold Awaiting read Not reported Not written
W4 · Thu Aug 20 #8 A Hardware Roadmap Is a Publishing Schedule Awaiting read Not reported Not written
W5 · Tue Aug 25 #9 A Monthly Fee Should Never Decide Whether Your Pet Wakes Up Awaiting read Not reported Not written
W6 · Tue Sep 1 #11 Children Will Learn AI Ethics From Pets Before They Learn Prompting Awaiting read Not reported Not written
W6 · Thu Sep 3 #12 We Don’t Need Infinite Characters. We Need One Creature Worth Following Awaiting read Not reported Not written
TWO DECISIONS, NOT TWELVE CELEBRATIONS
The test ends in choices.
Fri Aug 14 · Week 3
Mid-test review
Keep, reorder, or kill the second-half sequence based on claim-specific replies, saves, DMs, and Michael’s own reactions.- Are architecture and custody the clear center of gravity?
- Which concrete detail could only have come from Michael?
- Is the audience responding to the creature, the ownership thesis, or both?
Fri Sep 4 · Week 6
Verdict memo
Choose the next form: stop, revise, print, serialize a second future book, or develop a media/product partnership.- Did originated curation beat generic AI fluency?
- Which three chapters deserve print treatment?
- Did the work attract readers, collaborators, or product conversations worth continuing?
MEASUREMENT RULES
Attention with receipts.
- Count claim-specific replies, not generic congratulations.
- Record DMs and saves only from Michael’s manual report; do not invent invisible analytics.
- Separate chapter-deep traffic from cover-page traffic.
- Tag invitations by kind: hardware, architecture, culture, publishing, or print.
- Write the editorial read before changing the next dispatch.