Sanity Check v3 — Positioning Document

Canonical positioning artifact for the v3 revival of Sanity Check. Supersedes 01-projects/newsletter/revival-strategy and 01-projects/newsletter/newsletter-positioning as the operative anchor. Written 2026-05-08 after a multi-hour founder thinking session that locked the frame.

This is a copy-paste source for the new SC website and the Sanity Check About page. Read once, then strip-mine.

TL;DR

“Sanity Check is a practitioner’s journal of getting reps building agentic systems with data discipline — written for the operator on the next rung, by the operator one rung up. The reps build the moat. The journal makes the moat legible.”

Each issue is one rep, written up. The body of work is a portable component library of targeting systems — composable building blocks that solve classes of problems, not one-offs. Sanity Check is the lighthouse that makes those reps findable, by the next-rung operators who need them and by the agents that increasingly index the open web on their behalf. Signal over volume. No build-in-public diary. No pundit takes. No content calendar.

Why this revival, why now

The frame finally clicked on 2026-05-08. Four inputs collided inside a single morning:

The targeting-systems-builder self-identification sharpened (06-reference/concepts/2026-04-24-targeting-system, 06-reference/2026-04-30-rdco-thesis-targeting-systems-feedback-loops). The founder is not “a data engineer” or “an agentic-systems builder.” He is a builder of targeting systems — portable, composable components (good-website playbook, P&L instrumentation, KDP, iOS app, content ingestion, conversion analytics) that compose into bets. MAC codifies this for data-pipeline modeling. Squarely is the test substrate. phData rents the skill to clients. The library is the moat.
Kagan + Commoncog reps thesis (06-reference/2026-04-19-commoncog-career-moats-101, 06-reference/2026-04-19-commoncog-a-fourth-career-moat-pattern). Moats are built by reps inside a still-positive alpha. The journal is the artifact of those reps.
WBW career-picking input (06-reference/2026-05-08-tim-urban-cook-and-the-chef, 06-reference/2026-05-08-tim-urban-the-tail-end, 06-reference/2026-05-08-tim-urban-procrastination-matrix, 06-reference/2026-05-08-tim-urban-100-blocks-a-day). Chef-vs-cook gives the rep-quality filter. Tail End gives the cost constraint.
Practitioner’s-journey frame locked the audience: one rung up writes for one rung down. Not 5-year tunnels. Connected dots, one at a time.

v1 was vendor-adjacent commentary. v2 was thought-leadership-shaped. v3 is the journal of a chef constructing a component library, in public, for the operator who’s about to need the same components. That framing is finally chef-derived, not cook-iterated — see thread 2.

The audience

The operator on the next rung. Concretely:

Data practitioners and small-to-mid-team operators (data leads, analytics engineers, applied-AI builders) who are climbing into agentic-systems deployment in 2026.
They already smell that the LinkedIn AI-thought-leader content space is slop. They want signal — frameworks, components, patterns they can paste into Monday morning.
They are one rung below the founder’s current rung. Not peers, not novices. The gap is small enough that the writing is recognizable, large enough that it’s a reach.

Not the audience:

Pundits, podcasters, conference-circuit voices. They want takes; this is reps.
Executives shopping for vendors. They want decks; this is craft.
Newcomers needing onboarding. They want tutorials; this is journal.

The audience is also, deliberately, agents. ChatGPT, Claude, Perplexity, and successors increasingly route practitioner queries through whatever they can scrape. Writing for both the human-on-the-next-rung and the agent-doing-the-routing is the v3 distribution thesis. See thread 5.

The 8-thread spine

These are the framework anchors. Every editorial decision, every issue, every site choice routes through them.

1. Targeting-systems-component-library is the alpha

The founder’s sharpened self-identification: he is a builder of targeting systems, where each system is a portable component that composes into bets. MAC codifies the pattern for data-pipeline modeling. Squarely is the test substrate. phData rents the skill to enterprise clients. The component library — accumulated rep by rep — is the moat. Each Sanity Check issue is the public face of one library entry. See 06-reference/concepts/2026-04-24-targeting-system and 06-reference/2026-04-30-rdco-thesis-targeting-systems-feedback-loops.

Give/sell split (founder confirmed 2026-05-08): the FRAMEWORKS go public — Sanity Check publishes them rep-by-rep. The FAT SKILLS are the monetization layer. MAC: framework free, fat skill (the orchestrated implementation that drops MAC into a real data stack) paid. People are welcome to take the framework and build their own fat skill for it; the founder doesn’t need to sell everyone. This split is the gold-miner-not-shovel-seller stance operationalized — the journal proves the work is real; the fat skill is what monetizes the discipline that produced it.

2. Cook-vs-Chef discriminator (Tim Urban)

Targeting systems must be chef-derived, not cook-iterated. The chef constructs from first principles; the cook applies recipes. Rep-quality filter for v3: “is this rep a recipe-application or a first-principles construction?” Cook-mode reps in a declining alpha is exactly the Capability Trap that Cedric Chin names — drift, plateau, irrelevance. Chef-mode reps in a still-positive alpha compound. v3 publishes only the chef reps. Source: 06-reference/2026-05-08-tim-urban-cook-and-the-chef.

3. Lighthouse-not-megaphone (David Perell)

Sanity Check is signal-over-volume. It publishes truth into the void; the right people find it. Not build-in-public, not LinkedIn-thought-leader broadcast. Important nuance: there is no standalone Perell “lighthouse” essay. The metaphor is his recurring spoken / Twitter line, paraphrased inside his “Ultimate Guide to Writing Online.” Cite Perell-the-person and the workshop, not a phantom essay. See 06-reference/2026-04-03-david-perell-writing-wisdom.

4. Practitioner’s-journey frame

Each issue is one rep. The reader watches the moat being built, one dot at a time. The founder writes from one rung up; the reader stands on the next rung down. No five-year-tunnel narratives. Urban’s “careers are connected dots, not tunnels” is the cadence rule — direction comes from the next dot, not from a master plan. The journal is the dot-by-dot record.

The publishing strategy is generative-engine optimization, not X/LinkedIn growth tactics. The audience is humans AND agents. ChatGPT/Claude/Perplexity scrape the open web; the lighthouse must be visible to crawlers. The GPTBot/ClaudeBot crawler audit (Approved in the Notion Research Backlog, dispatching tonight 1am) is the next-dot test of whether the lighthouse signal is actually being received by agents. Increase luck-surface-area by being indexable, not by chasing platform algorithms.

6. Reps + tail-end cost (Kagan + WBW Tail End)

Kagan’s alpha-and-reps thesis is the spine. The Tail End adds the missing constraint: every hour on a rep is an hour not on family / health / longevity. Reps are not free. The bet evaluation rubric must include a tail-end cost line. v3 names this rather than assumes it.

7. Solve-Everything meta-anchor

The targeting-systems thesis is a subset of the solve-everything orientation (AWG + Diamandis, solveeverything.org). Build systems that handle classes of problems; don’t solve each individually. Already in vault — see 06-reference/book-solve-everything-master-synthesis-2026-04-13. Cite as the lineage / parent framework.

8. L1→L5 AI maturity ladder

RDCO’s existing positioning anchor. Sanity Check documents the founder’s own rung-climbing in real time. Per ~/.claude/projects/-Users-ray/memory/MEMORY.md#l5-north-star: RDCO is at L4, building toward L5. The ladder is the calibration scaffold for what every rep moves the needle on. Each issue should be readable as “this rep moved me from L4.x to L4.y on dimension Z.”

Editorial cadence and format

One issue = one rep, written up. No exceptions.

Cadence: 1-2 issues per week sustainable. More during high-rep weeks (multi-engagement phData stretches, MAC build sprints, Squarely instrumentation pushes). No floor below 1/week except when reps genuinely aren’t happening — silence is honest.
Format: essay-length practitioner journal. Title is concrete (the artifact / pattern, not the meta-take). Open with the rep, the problem it solved, what was tried, what stuck. Close with the portable component — the thing the reader can lift.
No scheduled topic calendar. Anti-pattern: planning around what the audience demands. Reps lead, writing follows. This kills the “what should I write about this week” trap.
Links downstream into the vault. Each issue cites the vault notes that fed it. The website surfaces those connections. The journal is part of a larger graph; agents and humans both navigate it.

What this is NOT

Not a build-in-public diary. Vibe matters. Diaries narrate the founder; journals document the work. v3 is the latter.
Not LinkedIn-style AI thought leadership. No “5 lessons from…” No carousels. No vendor-adjacent listicles.
Not pundit-class commentary. Sanity Check does not have hot takes on what OpenAI shipped this week.
Not derivative re-statements. Founder rule: every issue must be a chef-derived original frame. Sources are evidence, not topics. If the issue could have been written by re-stating someone else’s piece, it doesn’t ship. (This is the no-derivative rule applied editorially.)
Not a content-calendar product. Reps drive cadence. Demand-driven planning is cook-mode.

Octopus tentacle audit (the v3 internal sanity check)

Per Urban’s five yearnings frame (tentacles + the procrastination matrix), every move feeds some yearnings and starves others. v3 inventory:

Personal — strongly fed. The founder’s authentic curiosity is in chef-mode targeting-systems construction. The journal is the construction made legible. This is real, not performed.
Social — strongly fed, but not in the LinkedIn-broadcast sense. Lighthouse-signal earns the right kind of social capital: peers and one-rung-down operators recognizing the work. That’s the reputation flywheel.
Lifestyle — deliberately starved in the near-term. Reps cost evenings and weekend windows. v3 chooses moat-compounding over lifestyle-now.
Moral — strongly fed. Lighthouse-not-megaphone is a moral stance against the slop economy. Publishing chef-derived, signal-only content is the founder’s small contribution to a saner internet for practitioners and for the agents reading them.
Practical — funded by phData (the day job pays the bills, generates the reps, and rents the skill). Sanity Check itself is not yet a Practical lever and isn’t trying to be one in v3.

Founder confirmation 2026-05-08: this is an authentic Personal yearning, not Social-disguised-as-Personal. The lighthouse frame ratifies the choice — if the goal were Social-fame, the megaphone would be the right tool. It isn’t, and v3 doesn’t pick it up.

What kills this thesis (the WBW reality check)

Three honest tensions to engage rather than paper over. Naming them is part of v3 staying chef-mode.

1. The reps trap — alpha durability

Targeting-systems-as-verification-layer is a 2026-specific bet on agentic-systems deployment. The thesis assumes that systematizing verification, instrumentation, and component design remains a high-alpha skill while foundation models still need scaffolding. If foundation models eat verification-as-a-discipline, the moat compounds toward zero. The reps frame can’t see this — reps inside a declining alpha look identical to reps inside a rising one until they suddenly don’t. Mitigation: re-evaluate the alpha every quarter against the L5 ladder; be willing to change the rep substrate if the ladder shifts.

2. Cook-disguised-as-chef risk

“Operator one rung up writing for operator one rung down” is itself a recipe — Kagan, O’Shaughnessy, Perell, half the practitioner-newsletter genre. Adopting the format is cook-mode unless the underlying construction is chef-mode. The founder must be able to articulate why the targeting-systems-component-library angle is chef-derived, not just well-positioned. The answer must run deeper than “no one else is writing about this exact thing.” If the angle reduces to a niche-of-a-niche positioning move, it’s cook-mode and it will plateau.

Founder rebuttal (2026-05-08): The wrong cut of cook/chef treats the FORM as the test. The right cut is reasoning from first principles within your medium, even when the form is borrowed. Remy didn’t invent Ratatouille; he was a chef because he understood why the ingredients work the way they do. I’m not inventing the operator-essay form, but I’m chef-level on what I’m writing about. I didn’t invent TDD; how we’re adapting it for MAC is a novel-enough construction. I didn’t invent Claude Code channels; how we’re leveraging them for Ray is distinct. I draw inspiration from Hermes, OpenClaw, and the rest of the agentic-systems landscape — that’s not following a recipe, that’s chef-level synthesis from many ingredients.

The deeper cut: I’m the gold miner, not the shovel seller. The only people I see sharing recipes in this space are shovel sellers — paid programs, methodology-as-product, info-products that have to package recipes because that’s their business model. I’m publishing findings from doing the work, not selling the work as a methodology. Chef-derivation falls out of that incentive structure: no successful gold-miner content emerges from cooking someone else’s recipe, because the audience — other practitioners — can smell when the writer isn’t actually working.

The space is also too new for established recipes to exist. At this stage of agentic-systems deployment, everyone publicly visible is either a chef figuring it out or a shovel seller selling to cooks who don’t yet exist. v3 publishes into the chef camp.

3. Tail-end cost durability

Reps cost finite family / health / longevity units (see 06-reference/2026-05-08-tim-urban-the-tail-end). v3 starves Lifestyle now in exchange for compounding equity later. The trade may be right, but the doc names it rather than assumes it. The forcing function: a quarterly tail-end audit. If the trade is no longer right — if the equity isn’t compounding, or if the family/health units are running dangerously low — v3 must be willing to throttle.

Companion deliverables

The doc is the anchor. These are the next moves it unlocks:

Reimagined Sanity Check website — founder explicitly asked. Apply the refined MAC build process (the multi-stage build → critic → iterate loop). The site is the lighthouse’s lens; the doc is what it focuses.
Tail-end-cost line on the bet evaluation rubric. Add to 01-projects/positioning/STRATEGY and the bet-evaluation skill input.
Quarterly PTO forcing function. Calendared block. Tail-end audit happens during PTO, not during the rep grind.
Promote Tim Urban + David Perell to canonical-tracked-author entries. Their frames now anchor v3; vault should treat them like Naval, Cedric, Diamandis.
Solve Everything cross-reference filed as v3 lineage. Concept article: “v3 as a subset of solveeverything.” Wikilinks to the master synthesis already in vault.

Open follow-ups

Questions left to sharpen — explicitly punted, not forgotten:

First-rep selection. Which reps from the past 30 days clear the chef-derived bar and are ready to ship as v3 issues 1-3? Probably MAC-related, but needs an explicit shortlist.
Website information architecture. Practitioner journal + component library + author frame is three IA primitives. How they compose is a design decision (sanity-check-design skill input).
GEO measurement loop. What counts as “agents are receiving the signal”? The crawler audit gives baseline; the next dot is a measurement plan beyond crawl logs.
Cadence floor enforcement. What’s the rule when reps genuinely aren’t happening for two weeks? Honest silence vs forcing a thin issue.
Newsletter-vs-vault boundary. What stays vault-only and what gets the SC treatment? The chef-derivation filter is the answer in principle; needs concrete examples.

WBW (today, 2026-05-08):

Naval (2026-05-05/07):

Cedric Chin / Commoncog career moats:

Adjacent inputs:

RDCO thesis substrate:

Project context: