Ship 30 for 30: topic bank v1
Format status: verified against the curriculum map (2026-10-05), which is Ray's read-only read of the classroom. Each row has been re-cut to the rule for that day in the evergreen 30-day Ship 30 curriculum, which Ray read in the AI Writing Skool classroom. Details: [[01-projects/personal-brand/ship30/2026-10-05-ship30-curriculum-map]].
Caveat: the October cohort's new 2026 day-by-day modules are not published yet. As of Oct 5, the October course holds only onboarding Day 1 and 2 masterclasses. What is public about 2026 is the format arc from the live sessions:
- short-form on Oct 12
- Atomic Essays on Oct 15
- LinkedIn Carousels on Oct 19
- X Articles on Oct 22
- niche on Oct 26
- comments-as-content on Oct 29
- AI agent distribution on Nov 2
- library on Nov 5
Those dates are marked in the table. If a 2026 module sets a different rule for a day, the 2026 rule wins and Ray re-cuts that row.
Day 1 is confirmed as Saturday Oct 10. The cohort's Day 1 module says you publish your first piece on October 10th. Day 30 is Sun Nov 8, and graduation is Mon Nov 9.
Rules for this bank:
- Topics and angles only, never drafts.
- No client names, no phData internal detail, no exam content (covered by the certification Non-Disclosure Agreement, NDA).
- The voice is Ben's: clear, practical, show the work.
- Unless a row says otherwise, the default piece is an Atomic Essay: 300 words or less, one idea.
| Day | Date | Ship 30 rule for the day | Working headline (cut to the rule) | Angle (1 line) | Source |
|---|---|---|---|---|---|
| 1 | Sat Oct 10 | Announcement essay: "I recently signed up for... Here are a few reasons why... Over the next 30 days, I plan on writing about..." | Why I'm writing about AI agents in production for 30 days | Three reasons: building one now, the data-platform lens, showing the work. Three topics: definitions, agent failures, agent-shaped businesses. | topic bank themes (Ray-selected, awaiting his read) |
| 2 | Sun Oct 11 | 2-Year Test: teach the version of you from 2 years ago | Agents don't need your data warehouse. They need your definitions. | Two years ago he'd have built pipelines first. Now definitions come first, because two agents asked about "gross retention" should not give two answers. | [[2026-10-02-agents-need-your-definitions-v1]] |
| 3 | Mon Oct 12 | For Who / So That at the top of the page (live: Short-Form) | Your metric definitions belong in git | TOPIC: definitions as code. FOR WHO: data platform leads. SO THAT: every agent reads one versioned, reviewed answer. | [[2026-10-04-kutay-gtm-in-git]] |
| 4 | Tue Oct 13 | Personal credibility: name the experience that qualifies you | My AI agent didn't crash. It waited. | His own always-on agent: the worst failures throw no error, like a hidden password prompt or a silent wait. | [[2026-10-02-post2-agent-didnt-crash-v1]] |
| 5 | Wed Oct 14 | 4A: run one topic through all four A's, ship one | 3 agent failures your data platform team already knows how to catch | Topic: agent failures. Ship the Actionable cut (stale status, missing tools and silent waits map to monitoring you already run). Keep the Analytical, Aspirational and Anthropological cuts as spares. | post2 draft (failure list) |
| 6 | Thu Oct 15 | Endless Idea Generator: generate 5+ ideas, ship one, using a "30 approaches" container (live: Atomic Essays) | 4 reasons four of five jobs are agent-shaped (and the fifth is the whole business) | Container: Reasons. The job agents can't carry is the reason customers pay. | [[agentify-3000-series]] spine #1; drafts/2026-10-04-fifth-job-01-four-of-five-v2.md |
| 7 | Fri Oct 16 | Lean Writing: one topic through the 30 approaches, ship one small | The 4 steps I used to score 2,500 small businesses on 5 jobs | Container: Steps. Public blurbs, a 0-3 score per job, and why a model grading a blurb is a weak signal. | [[agentify-3000-series]] ("Shared data", n = 2,500 derivation); agentify-3000/IMPLEMENTATION-NOTES.md (0-3 scoring) |
| 8 | Sat Oct 17 | Headline = specific topic + proven approach (clear, not clever) | 3 reasons admin is the first job to hand to an AI agent | Start where the score is high and mistakes are cheap. Keep a human on the money. | series spine #2; [[archetypes]] |
| 9 | Sun Oct 18 | Headline + explicit audience | How small business owners can hand outreach to agents without spending their trust | Outreach and search engine optimization (SEO) are agent-shaped. Being vouched for is not. | series spine #3; Kutay note |
| 10 | Mon Oct 19 | Headline + audience + outcome (live: LinkedIn Carousels) | How agency owners can stop selling hours and sell judgment once agents carry production | The agency margin is payroll. Marketing, design and film agencies only. | series spine #4 (respect the CHECK note) |
| 11 | Tue Oct 20 | Headline + "Twist the Knife": stack the extra outcomes | Why owners of physical-product businesses should give agents the back office and keep delivery, so they cut admin hours without risking quality or repeat customers | Agents run the business around the product, not the product. | series spine #5; [[foils]] |
| 12 | Wed Oct 21 | 5 headline styles: write List / Story / Opinion / Question / Framework, ship one | Opinion style: "Most people think a one-person business can't be agent-native. Some already are." | They score 5 of 5. The gap is know-how, not possibility. Keep the other 4 headline styles as spares. | series spine #6 |
| 13 | Thu Oct 22 | Write an X hook for the essay: a 3-second preview, not the headline repeated (live: X Articles) | A model grading itself is not a measurement | Why every score in the dataset carries a counterpoint, and how to check one by hand. The hook teases the hand check. | series spine #1 counterpoint; [[foils]] gate |
| 14 | Fri Oct 23 | LinkedIn hook in zero-to-hero form: "A few years ago, I {problem}. Today, I am {accomplishment}. How? By doing {something unexpected}." | What studying for two AI certs taught me about how I learn | Zero-to-hero: he used to cram, today he holds two passed AI certs, and the unexpected part is spaced quizzes with rotating answers. Study method only, no exam content. | [[2026-10-02-post3-snowpro-genai-cortex-v1]]; 01-projects/certifications/ |
| 15 | Sat Oct 24 | Prep the Page: headline, credibility line, main points and takeaway decided before drafting | The 4-step check I run before citing any example | Outline is already 4 steps: re-read the full story, confirm the founder still runs it, check that revenue agrees, check team size in the header. | series plan, "gate for every foil" |
| 16 | Sun Oct 25 | Single-sentence openers (6 types): use one at the start of every section | The docs changed while I was studying | Opener type "moment in time". The lesson: study material lags the product, so read the current docs and note the gap. Study method only, no exam content. | [[2026-10-02-post3-snowpro-genai-cortex-v1]] ("What changed since I studied") |
| 17 | Mon Oct 26 | Subheads every ~100 words that mirror the headline promise exactly (live: Data-Driven Niche Selection) | 3 checks to run before you believe an agent that says "blocked on you" | Subheads: Check #1 / #2 / #3. Stale notes become false blockers, so re-check status at the source. | post2 draft; verify-blockers lesson |
| 18 | Tue Oct 27 | 1/3/1 rhythm under each subhead | Never let an agent grade its own homework | Route review through a fresh reviewer with no context. Confirmation bias is real for models too. | [[2026-05-19-verification-as-independent-worker-pattern |
| 19 | Wed Oct 28 | Short-form post (X/LinkedIn): 1 line / 3-5 bullets / 1 line, and it must pass the Screenshot Test | Write the test before you let the agent build | Three bullets: the test comes from a source older than the code; the agent can't move the goalposts; review gets cheaper. | plan → tests → implementation practice |
| 20 | Thu Oct 29 | X thread: lead-in, one point per tweet, TL;DR tweet, call-to-action (CTA) tweet (live: Comments-As-Content) | Agent context is a file, not a chat | One tweet per point: a markdown vault, skills, git history, why a longer prompt loses. The CTA points to the site. | Kutay note "RDCO itself" mapping |
| 21 | Fri Oct 30 | LinkedIn long-form post: lead-in of 250 characters or fewer, numbered points, CTA in a self-comment, or edited in about 4 hours after posting | More context makes your agent worse | Long sessions rot. Hand big documents to a sub-agent and keep the summary. | [[2026-04-15-thariq-claude-code-session-management-1m-context]] |
| 22 | Sat Oct 31 | Re-write the topic you enjoyed most (any format) | Placeholder: The semantic layer is your agent's dictionary | Replace with whichever of Days 1-21 he most enjoyed writing. The placeholder assumes the definitions line wins. | [[2026-10-02-agents-need-your-definitions-v1]]; [[2026-06-15-data-engineering-weekly-ai-agents-data-foundations]] |
| 23 | Sun Nov 1 | Double down on one of the top 3 performers, ranked views < likes < comments < shares (any format) | Placeholder: 5 questions to ask before you put an agent in production | Fill from the real numbers. The placeholder assumes the agent-ops line performed: who checks it, what it can touch, how you see it stall, the cost of being wrong, how you roll back. | post2 draft + build notes (Ray-selected synthesis) |
| 24 | Mon Nov 2 | Same proven topic, change exactly one variable (audience, approach or outcome) (live: AI Agent Distribution) | The boring part of agents is the whole job | Same topic as Day 23, approach changed from Questions to Lessons. Restarts, logs, auth and alerts decide uptime, not the model. | post2 draft "operations failures" |
| 25 | Tue Nov 3 | Double down on a proven audience (explicit or implicit) | My agent ran for a day and a half with its tools missing | Aimed at whichever audience replied most by then (default: data platform engineers). One config file quietly changed the tool list. | post2 draft; synced-plugin-settings lesson |
| 26 | Wed Nov 4 | Core narrative: one of his 3-5 recurring stories | Agents lift people up. They don't replace the fifth job. | Core narrative candidate: "I run an always-on agent at home and watch what it can't do." Restate the series belief with one worked example. | series plan core belief |
| 27 | Thu Nov 5 | X bio audit for the 4 mistakes, then ship any format (live: Leveraging Your Library) | What a data leader should measure first when agents arrive | Measure consistency of answers before speed of answers. Side action: audit the X bio. | [[2026-10-02-agents-need-your-definitions-v1]] ("Consistency") |
| 28 | Fri Nov 6 | LinkedIn bio rewrite ("I talk about {Topic}" + {Audience} + credibility), then ship any format | I publish my counterpoint in every post. Here's why. | Moderate-confidence claims build more trust than certainty. Side action: the bio rewrite. Coordinate with the profile drafts in [[06-reference/2026-10-02-welsh-linkedin-os-applied]]. | [[VOICE]] tone #1 |
| 29 | Sat Nov 7 | Pin your best work (X pin + LinkedIn Featured), then ship any format | Data for good: the denominator decides the story | Same numerator, different base, different headline, so pick the base out loud. Doubles as a lead-in to the site name. Side action: pin the best piece of the month. | [[2026-10-04-diamandis-laws-and-data-for-good]] |
| 30 | Sun Nov 8 | Final reflection essay (any format). Ray's suggested prompts: what you learned, what you'll keep writing, what you were afraid to share | What 30 essays taught me about my own ideas | Which topics pulled replies and which fell flat (fill from real numbers), plus the 3 ideas that become Denominator posts. | days 1-29 |
What changed from the unverified version
- Day 1 is now the required announcement essay. "Definitions" moved to Day 2, because it fits the 2-Year Test.
- Every headline on Days 8-12 is rebuilt with the course's headline formula, adding one layer per day.
- Days 19-21 now use the course's three format days: short-form, X thread and LinkedIn long-form.
- Days 22-25 are placeholders that depend on real performance data, because those days' rules require it.
- Two old rows merged. "What 28 days showed me" and "The 3 ideas I'm keeping" are now the Day 30 reflection, which keeps the count at 30.
- Day 14 and Day 15 swapped topics: the cert-learning story fits the zero-to-hero hook, and the 4-step check fits Prep the Page.
Notes for Ben
- Days 6-11 follow The Fifth Job spine in order, so the daily essays can seed the long posts.
- Rows marked "Ray-selected" are Ray's picks, awaiting your read.
- Posting: the evergreen course says publish on X plus the Typeshare Social Blog. The 2026 cohort adds LinkedIn and Substack. You post, and you share each day's link in the 🚢 Ship 30 community category.
- No refund rule was found. This cohort is free with AI Writing Skool. See the curriculum map.
Alternates (swap in any day)
- Headline: "Your Digital Brain is a semantic layer" · Angle: Ship 30's "Digital Brain" (a library of your own stories the AI draws from so it doesn't make things up) is The Denominator's thesis applied to writing: AI invents unless you give it shared definitions. · Format: Atomic Essay (≤300 words). · Source: Dickie Bush's Oct 4 Ship 30 launch letter (X Article, https://x.com/dickiebush/status/2106763420053774713); [[2026-10-02-agents-need-your-definitions-v1]]. · Added 2026-10-05, Ray-selected, awaiting his read.