LinkedIn Launchpad speed run
This doc covers sections 1-3 of the "LinkedIn AI Launchpad" playbook in one pass, so you can post on Wed Oct 28. The playbook itself suggests one or two sections a week. The headings and checkboxes follow the playbook's toggles. [BEN] = only you can answer. [CHECK] = a fact Ray could not confirm. Citations are vault file:line.
Ben's 4 actions
- Paste your current headline and About into section 2. Nobody has seen the live profile: the 10-02 review hit a sign-in wall (
06-reference/2026-10-02-welsh-linkedin-os-applied.md:83), and you changed the headline yourself that evening (2026-10-02-site-spec.md:32).- Pick a headline: H1, H2 or H3.
- Post Day 18, "Never let an agent grade its own homework", on Wed Oct 28. First answer its 5 confirms (sent by iMessage 2026-10-27 ~16:40 ET;
ship30/2026-10-26-week3-drafts.md, Day 18). A fresh critic re-ran it on 10-27 (round 2 ITERATE, 5 exact edits applied; critic: PASS once applied); the paste-ready version is in iMessage.- This week: pick a cadence (2, 3 or 5 a week; section 3) and keep or bench the business-owner posts (Days 6, 8-12). Both change what posts after Wednesday. Day 14 (9 [BEN]) is next on Oct 29.
1. Create a personal brand category you can own
The positioning already exists. This section only records it in the playbook's fields.
[x] Unlock your information advantage
| Playbook field | From the vault |
|---|---|
| Jobs | Principal Consultant at phData (2026-10-02-site-spec.md:20; a retitle ("FDS/FDE" in the source; the vault never expands it) is likely [CHECK]) |
| Skills | Claude Certified Architect – Foundations (CCA-F), 850 of 1000 (01-projects/phdata/cert-progress.md:63). SnowPro Specialty: Gen AI (GES-C02), 854 (cert-progress.md:198, :200) |
| Experiences | Ran questions at a Claude Code activation session with about 30 advanced users (ship30/2026-10-07-locked-topics-and-question-bank.md:19) [BEN: taught alone or co-presented? It was a client activation call (locked-topics:29): OK to cite publicly?]. Uses his own agent as the case study (locked-topics:15). Had a model score 2,500 public businesses from Starter Story on 5 jobs (ship30/2026-10-07-week1-drafts.md:212) [BEN: confirm you ran this scoring] |
| Hobbies | [BEN] |
| Goals | Credibility that helps phData sales. "The score is phData pipeline, not followers" (2026-09-29-nov30-credibility-targets.md:13) |
| Information advantage | Ray's summary: you see agents from two seats. You consult on enterprise data platforms, and you run your own agents, whose failures become the posts (week3 Days 17-18) |
| 5 industries to serve | [BEN]. Healthcare and Life Sciences was the 09-29 pick. You dropped it on 10-05 (~/.claude/projects/-Users-ray/memory/project_ambition_bigger_problem_network.md:19). The same note names your professional exposure as B2B SaaS and D2C e-commerce, and your pull as programmable engineering. Start from those two? |
| Who benefits | VP- and Director-level data leaders; phData reps and partners (nov30-credibility-targets.md:32-33; VOICE.md:19) |
| Problem solved | "None of these were model failures. They were operations failures, the kind your data platform team already knows how to catch." (VOICE.md:26) |
| Advantage matrix | [BEN] optional, skip for now |
| 3 content pillars | The locked Ship 30 topics (locked-topics:14-17). P1 Agents that actually work. P2 Agents + data platforms (the "phData lane"). P3 Which jobs to give agents first |
[x] Position yourself in a niche
| Prompt input | Value |
|---|---|
| Industry | Enterprise data and AI consulting [CHECK: Ray's wording] |
| Expertise | "I write about agents in production (on Snowflake + Claude)." (02-sops/DESIGN-denominator.md:40) |
| Constraint | "cannot talk about clients directly and cannot sell the same services. Thought leadership is fine." (site-spec:19). No "hire me" calls to action and no client anecdotes (VOICE.md:38) |
| 30 micro-niches / 20 adjacent | [BEN] skip for now |
2. Maximize your profile
[ ] Update your bio header (headline, 220 characters max)
Current headline: [BEN: paste]
Do not reuse the 10-02 option A (welsh:84). It names Healthcare and Life Sciences, which you dropped. Ray skipped the "I help {who} {do what}" frameworks because they read as selling a service (site-spec:19, welsh:69).
| # | Framework | Headline | Chars |
|---|---|---|---|
| H1 | V3 Credible Talker | I write about agents in production (on Snowflake + Claude). Principal Consultant @ phData. Claude Certified Architect – Foundations and SnowPro Specialty: Gen AI certified. | 172 |
| H2 | V5 Specialist | Build notes on agents in production: what I built, what broke, what I'd do next | Snowflake + Claude | Principal Consultant @ phData | 132 |
| H3 | V2 Aspirational Brand | Principal Consultant @ phData | I write The Denominator: the common factor in agents that work, on Snowflake + Claude | 117 |
Sources: H1 = DESIGN-denominator.md:40 + certs. H2 = nov30-credibility-targets.md:23. H3 = tagline, DESIGN-denominator.md:135. Use H3 only after thedenominator.dev launches; it was "kept dark until launch" (site-spec:62) [CHECK: launched?]. Ray's pick: H1. It uses the line you already approved, and the certs are the strongest proof you can name.
[ ] Polish your profile image
Use the squiggle headshot, your stated favorite (site-spec:55; file per DESIGN-denominator.md:41) [CHECK: same image LinkedIn shows now?]. Skip BetterPic.
[ ] Update your banner image and strapline (10-15 words)
- B1, strapline: "The common factor in agents that work." (
DESIGN-denominator.md:135) + thedenominator.dev after launch [CHECK]. - B2, authority: "Agents in production on Snowflake + Claude | What broke, what I'd fix | Follow for build notes"
- Visual: blue graph paper
#EAF0FD, one marker-blue#3B82F6underline (site-spec:24-26). Leave the phData logo off unless brand policy allows it (welsh:85). The playbook wants metrics on the banner. Ray left them out because none is ready yet (WRITING-rdco.md:32).
[ ] Write your About section (2,600 characters max)
Current About: [BEN: paste]
Draft pending Ben's current profile text. Framework #4, Contrarian Insight. It runs about 1,250 characters, shorter than the playbook's 250-300 words, to keep your short sentences (VOICE.md:28). It ends on a question (VOICE.md:42).
Getting an agent to work in a demo is easy now. Keeping it working on a Tuesday is still hard.
My own agents have failed in three ways. One reported it was blocked on me when one look at a file would have shown it wasn't. One ran for a day and a half after a config file quietly changed its tool list. One checked its own work and called it clean. None of these were model failures. They were operations failures, the kind your data platform team already knows how to catch.
I'm a Principal Consultant at phData. I write about agents in production on Snowflake and Claude: what I built, what broke, and what I'd do next.
What I believe, with moderate confidence:
- Agents need your definitions more than your data warehouse. If the definition of gross retention changes between requests, so does the agent's answer.
- The builder should not be the reviewer. Give a second agent the finished work and the spec, and nothing else.
- Agents can lift people up instead of pushing them out.
Credentials: Claude Certified Architect – Foundations (850 of 1000) and SnowPro Specialty: Gen AI (854). I also ran an advanced Claude Code session on agentic software development for about 30 practitioners. [BEN: wording per the Experiences row; cut it if the client-session source rules it out]
If you run agents in production, what broke first for you?
Fact trail: blocked = week3 Day 17 (line 32). Tool list = ship30/2026-10-05-topic-bank-v1.md:61. Self-check = week3 Day 18 (line 60). Operations failures = VOICE.md:26. Definitions = site-spec:22. Reviewer = week3 line 74. Belief = DESIGN-denominator.md:138. Session = locked-topics:19 [CHECK: taught alone or co-presented?]. [BEN] Add "Views are my own." as the last line? Your 10-04 ruling keeps it once, in the about section (DESIGN-denominator.md:139), amending the 10-02 "drop everywhere" (site-spec:43). phData's disclaimer policy is still open (welsh:90).
[ ] Turn on Creator Mode
N/A. LinkedIn removed the toggle in March 2024, per third-party sources in welsh:49 [CHECK in settings].
[ ] Verify your LinkedIn account
[BEN] Verification needs your ID.
[ ] SEO-optimize your profile with Skills
Starter list: Snowflake · Snowflake Cortex · Claude · AI Agents · Model Context Protocol (MCP) · Large Language Models (LLM) · Semantic Layer · Data Engineering · Data Architecture · Data Governance. Run the playbook's Skills prompt after the About is final.
[x] Design your visual brand identity
Done: 02-sops/DESIGN-denominator.md. Skip the Gamma and Canva steps.
[ ] Featured (from section 7, pulled forward)
- The Day 18 post once it's live.
- The two cert badges: CCA-F on Credly (
cert-progress.md:63), Snowflake Gen AI from CertMetrics or Credly [CHECK: where it lives, and public?]. - thedenominator.dev after launch. Links go in Featured, not in post bodies (
welsh:68,:76).
3. Content strategy
[ ] Establish a posting cadence
The playbook says 3-5 days a week. Ray's 10-02 plan said about 2 posts a week within a 5-hour week (welsh:63; nov30-credibility-targets.md:15). Ray has no answer from you on record (question at welsh:91). [BEN] Pick 2, 3 or 5 a week. The calendar uses 5 a week, because the 18 drafts are already written. At 3 a week, the queue lasts 6 weeks.
Ship 30 note: the cohort ends Mon Nov 9 (ship30/2026-10-05-ship30-curriculum-map.md:14, :44). None of the 18 drafts has been posted [CHECK: Ray's last record is 10-18; confirm], so 30 pieces by then is out of reach. The cohort is free and has no "all 30" condition (curriculum-map:15), so the calendar just runs past Nov 9.
[x] Ideas for the content buckets
The prompt defines 3 buckets (Futurism, Your Story, Actionable Advice). Ray reads the toggle's "9 buckets" as 3 pillars × 3 buckets. Audiences: A1 VP/Director data leaders · A2 phData reps and partners · A3 data/platform engineers · A4 small-business owners (some drafts address them, but they are not the goal audience). All labels below are Ray's classification.
| Day | Headline | Pillar | Bucket | Angle | Aud. | Open tags |
|---|---|---|---|---|---|---|
| 1 | 4 questions engineers asked me about building software with agents | P1 | Actionable | Questions | A3 | 3 BEN |
| 2 | The spec your agent writes is disposable. Keep the tests and the why. | P1 | Actionable | Tips | A3 | 3 BEN |
| 3 | Your agent forgets everything between sessions. Mine reads a knowledge base. | P1 | Story | Personal story | A3 | 4 BEN |
| 4 | Is spec-driven development overkill for a bug fix? Shrink the spec, don't skip it. | P1 | Actionable | Questions | A3 | 5 BEN |
| 5 | When an agent reviewer finds something outside the spec, send it back with a test | P1 | Actionable | Steps | A3 | 2 BEN |
| 6 | 4 reasons the one job agents can't do is your whole business | P3 | Futurism | Reasons | A4 | 1 BEN |
| 7 | The 4 steps I used to score 2,500 small businesses on 5 jobs | P3 | Story | Steps | A3 | 1 BEN |
| 8 | 3 reasons admin is the first job to hand to an AI agent | P3 | Actionable | Reasons | A4 | 3 BEN, 1 CHECK |
| 9 | How small business owners can hand outreach to agents without spending their trust | P3 | Actionable | Tips | A4 | 2 BEN, 2 CHECK |
| 10 | How creative agency owners can stop selling hours and sell judgment once agents carry production | P3 | Futurism | Benefits | A4 | 2 BEN |
| 11 | How physical-product owners can give agents the back office, cut admin hours, and protect quality and repeat customers | P3 | Actionable | Benefits | A4 | 3 BEN, 3 CHECK |
| 12 | Most people think a one-person business can't run on agents. The data says some could. | P3 | Futurism | Stats | A4 | 3 BEN, 3 CHECK |
| 13 | A model grading itself is not a measurement | P1 | Actionable | Mistakes | A1 | 3 BEN, 6 CHECK |
| 14 | I passed two AI certs by studying my wrong answers | P2 | Story | Lessons | A3 | 9 BEN, 2 CHECK |
| 15 | The 4 checks a business must pass before I name it | P3 | Story | Steps | A1 | 4 BEN, 5 CHECK |
| 16 | The docs changed while I was studying | P2 | Futurism | Examples | A3 | 7 BEN, 3 CHECK |
| 17 | 3 checks an agent should run before it says "blocked on you" | P1 | Actionable | Steps | A3 | 3 BEN, 2 CHECK |
| 18 | Never let an agent grade its own homework | P1 | Story | Mistakes | A1 | 5 BEN, 5 CHECK |
Headlines are from the day tables: ship30/2026-10-07-week1-drafts.md:15-23, 2026-10-19-week2-drafts.md:15-21, 2026-10-26-week3-drafts.md:15-16. Ray counted the tags on 10-27.
Gaps
- P2, the phData lane, is thin. Only 2 of 18 drafts are P2, and both are about certs. No Snowflake build note exists yet; the tally is 0/6 (
nov30-credibility-targets.md:23). - No P1 Futurism. P2 has one Futurism draft (Day 16).
- Audience drift. Days 6 and 8-12 address business owners, not data leaders (
nov30:32-34). [BEN: keep or bench?] - Overlaps. Day 1 previews Days 2-5. Days 13 and 18 share a thesis, so the calendar puts about 3 weeks between them. Day 7 says 2,500 and Day 15 says 2,566. Both are right: 2,566 deduplicated, 66 dropped for suspect revenue, 2,500 clean (
week1-drafts.md:212). Make Day 15 say 2,500 scored. - Day 1 is stale. Edit "this week I taught" and "every day for the next 30 days" before it posts.
[ ] Fill your 30-day calendar (Wed Oct 28 → Thu Nov 26)
Weekdays only. All 18 drafts go first, rotating pillars where possible. Four open slots follow. Every post is a text post unless the row says otherwise.
| Date | Post | Pillar · Bucket |
|---|---|---|
| Wed Oct 28 | Day 18, grade its own homework | P1 · Story |
| Thu Oct 29 | Day 14, two AI certs | P2 · Story |
| Fri Oct 30 | Day 7, scoring 2,500 businesses | P3 · Story |
| Mon Nov 2 | Day 17, "blocked on you" | P1 · Actionable |
| Tue Nov 3 | Day 16, the docs changed | P2 · Futurism |
| Wed Nov 4 | Day 6, the one job agents can't do | P3 · Futurism |
| Thu Nov 5 | Day 1, 4 questions (edit first) | P1 · Actionable |
| Fri Nov 6 | Day 8, admin first | P3 · Actionable |
| Mon Nov 9 | Day 2, disposable spec (Ship 30 ends) | P1 · Actionable |
| Tue Nov 10 | Day 9, outreach and trust | P3 · Actionable |
| Wed Nov 11 | Day 3, knowledge base | P1 · Story |
| Thu Nov 12 | Day 10, agencies sell judgment | P3 · Futurism |
| Fri Nov 13 | Day 4, shrink the spec | P1 · Actionable |
| Mon Nov 16 | Day 11, back office | P3 · Actionable |
| Tue Nov 17 | Day 5, send it back with a test | P1 · Actionable |
| Wed Nov 18 | Day 12, one-person business data | P3 · Futurism |
| Thu Nov 19 | Day 15, 4 checks before I name it | P3 · Story |
| Fri Nov 20 | Day 13, a model grading itself | P1 · Actionable |
| Mon Nov 23 | Open: What a data leader should measure first when agents arrive: consistency before speed (topic-bank-v1.md:63) |
P2 · Actionable |
| Tue Nov 24 | Open: Agents don't need your data warehouse. They need your definitions. (site-spec:22). Document post: diagram + "what broke" list (welsh:76) |
P2 · Futurism |
| Wed Nov 25 | Open: My agent ran for a day and a half with its tools missing (topic-bank-v1.md:61) |
P1 · Story |
| Thu Nov 26 | Open: a Snowflake build note: built, broke, next [BEN: which build?]. Thanksgiving: [BEN] post or skip | P2 · Story |
The remaining topic-bank ideas (topic-bank-v1.md:55-66) stay on the bench. To close the P2 gap sooner, move the Nov 23-24 ideas into the Nov 4-6 slots.
Later (sections 4-7)
- 4. Attract the right audience. Run the pains/fears prompt for A1 data leaders, then hook-test each post before you publish it.
- 5. Use the algorithm. Post carousels and document posts to raise dwell time, and review analytics weekly. The first review point is topic-bank Day 23.
- 6. Build relationships. Write comments by hand on posts from about 15 named data leaders and reps (
welsh:77). The playbook's ~200 connection requests a week is your call. Ray flagged it as off-goal. - 7. Convert readers. The Denominator subscribe page is the opt-in once the site launches. The call to action stays "read the build notes," with no service offer (
site-spec:19).