Secrets (Zero to One)

SKILL.md

---
name: zero-to-one-secrets
description: "Activate when: user asks 'what do we know that others don't,' 'where is our edge,' 'what important truth do few people agree with me on,' 'is this idea defensible,' or is hunting for a startup idea, an untapped market, or a hidden opportunity; a business plan claims uniqueness but can't name the truth it rests on; user says 'Zero to One,' 'Thiel secret,' or 'contrarian opportunity.' Do NOT activate when: 'secret' means credentials, API keys, passwords, encryption, secret management, or any security/privacy topic — this is a business-opportunity skill, not a security skill; the user wants execution help on an already-validated idea rather than the truth beneath it. More: deciqai.com/c/zero-to-one-secrets"
---

# Secrets (Zero to One)

## Overview

Every great business is built around a **secret**: an important truth that is knowable but that few people know. Thiel and Masters place secrets between two things that cannot power a company — *easy truths* (conventions everyone already accepts, so the value is competed away) and *impossible mysteries* (truths no amount of effort can reach). A company with no secret is a commodity: whatever it does, everyone else can see and copy. A company built on a secret gets a head start measured in years, because the world hasn't caught up to the truth yet — Thiel's version of the interview question makes this concrete: "what important truth do very few people agree with you on?"

Two kinds of secrets exist. **Secrets of nature** are undiscovered facts about the physical or technical world — what a material can do, what a computation can be made cheap, what a process actually costs at the limit. **Secrets about people** are things people don't know about themselves, or things they know but taboo, politeness, or professional convention stops them from saying out loud. Secrets about people are systematically underhunted, because looking for them feels rude — which is precisely why they remain secrets.

The skill has two motions: *believing secrets still exist* (against four cultural forces that say they don't), and *searching where nobody else is looking* — because a secret everyone is hunting in the same place stops being a secret the moment the first hunter arrives.

Compose with contrarian-question (the interview form of the same move) · non-consensus-thinking (a secret is only valuable if it is non-consensus *and* right) · falsifiability (a candidate secret must be stated so it can be wrong) · the-mom-test (how to verify a secret about people without contaminating the answer).

## When to Use

**Use when:** hunting for a startup or product idea and everything on the list feels obvious; a pitch or plan claims differentiation but cannot state the truth competitors are missing; deciding between two markets and one is crowded while the other is ignored; evaluating whether a business is "zero to one" (built on a secret) or "one to n" (copying a working model); user asks "what's our unfair insight," "why hasn't anyone done this," or "what do we believe that the industry doesn't."

**Skip when:** "secret" means credentials, keys, or anything security-related — wrong skill entirely; the idea is already validated and the work is execution, distribution, or hiring; the question is purely about competitive tactics inside a known game (use [porters-five-forces] instead); the user wants encouragement, not a test — this process is designed to kill weak secrets.

## Coaching Novices (Adaptive Front Door)

Before running the Hunt, read the user. Two delivery modes — pick one, don't default to dumping analysis.

- **Engine mode (do-it-for-me):** the user brought a concrete idea, market, or claimed edge → run the Secret Hunt on it directly and concisely.
- **Coach mode (teach-me):** the user has no concrete case, or signals unfamiliarity ("what does Thiel mean by secrets?") → guide, don't analyze at them.

When unsure which they want, ask one line first: *"Want me to stress-test a specific idea for its secret, or walk you through how to hunt for one?"*

In Coach mode, respond one step at a time. Each [WAIT] is a hard stop — output that step's question and nothing more.

In coach mode:

1. **One-line what-it-is.** A secret is an important truth that is knowable but that few people know — sitting between what everyone accepts and what nobody can reach. Great businesses are built on one; commodity businesses aren't.
2. **Check fit.** Match their situation against *Use when / Skip when*. If they need security help or pure execution help, say so and point elsewhere.
3. **Elicit their real case.** Ask: "What's the idea, market, or belief you want to test — and what do you think you see that others don't?" Never run the Hunt on a hypothetical when a real case is available.

> **[WAIT — do not advance until user responds]**

4. **One step at a time.** Walk the Process one step per turn: first force the candidate secret into a single falsifiable sentence *with* them, then classify it, then hunt the reason others miss it — wait for their input before advancing.

> **[WAIT — do not advance until user responds]**

5. **Close by naming the payoff.** End with the one sentence *they* produced — the falsifiable secret and why it stayed hidden — or the honest verdict that their "secret" was an easy truth in disguise, so they remember the move, not just the answer.

> **[WAIT — do not advance until user responds]**

Then enter The Process below at the depth the chosen mode calls for.

## Process

Run the **Secret Hunt** (6 steps → one artifact). The gates are the method — a "secret" that can't pass them is a convention or a wish.

1. **State the candidate secret as a falsifiable claim.** One sentence, specific enough that evidence could prove it wrong. "AI will change everything" fails; "mid-size freight brokers will pay for automated invoice auditing because 5–10% of carrier invoices contain billable errors nobody catches" passes. *Gate: if the sentence cannot be wrong, it is not a claim — rewrite or stop.* (See falsifiability.)
2. **Classify it: nature or people.** A secret of nature needs technical/physical evidence; a secret about people needs behavioral evidence — and beware that secrets about people are the underhunted kind, hidden by taboo or politeness rather than difficulty.
3. **Explain why others miss it.** Name the specific blinding mechanism, drawn from the four forces that killed belief in secrets: **incrementalism** (the truth doesn't fit credentialed, publishable steps), **risk aversion** (holding it means being lonely and possibly wrong in public), **complacency** (incumbents collect rents and have stopped searching), or **flatness** ("if it were true, someone smarter at Google/McKinsey would already know"). *Gate: if you cannot name why smart people miss it, the default explanation is that it's false or already known → treat as easy truth and stop.*
4. **Check it's a secret, not a mystery or a convention.** Could a determined team *reach* this truth with feasible effort (not a mystery)? Is it *absent* from industry reports, competitor decks, and conference talks (not a convention)? Ask the field-test question: is this something practitioners know but are forbidden or embarrassed to say out loud? Those are the richest veins.
5. **Design the cheapest verification.** Specify one test, its cost, and what result would kill the claim — a dozen customer conversations run per the-mom-test, a scraped dataset, a landing-page test, a technical prototype of the one doubted mechanism. *Gate: no kill-condition defined → you are collecting confirmation, not verifying; go back to step 1.*
6. **Stop-rule:** if verification shows the truth is already widely known (easy truth) or unreachable at feasible cost (mystery), exit and hunt elsewhere — specifically, somewhere less crowded. The best place to look is where no one else is looking: unglamorous industries, taboo topics, questions people are discouraged from asking.

**Output template:** `Candidate secret (falsifiable sentence) / Kind (nature|people) / Why others miss it (which of the 4 forces) / Secret-not-mystery-not-convention check / Cheapest verification + kill condition / Verdict`

### Search Heuristics — Where to Hunt

When step 6 sends you back to the field, hunt with these questions rather than browsing trend lists:

- **What are you forbidden to talk about?** Taboos and enforced politeness are camouflage for secrets about people. If practitioners lower their voice before saying it, write it down.
- **What is hard but not impossible?** The middle band between trivial and hopeless is where secrets of nature live. "Hard" filters out the crowd; "not impossible" filters out mysteries.
- **Where is nobody looking?** Unglamorous industries, unfashionable geographies, "solved" markets, and problems too boring for a conference talk. Crowded fields self-strip of secrets; ignored ones accumulate them.
- **What do insiders do differently from what they say?** The gap between an industry's public story and its operators' private workarounds is a secret in raw form.
- **What became possible recently that pricing and process still ignore?** When a cost curve moves (compute, sensors, model capability) faster than industry habit, the lag *is* the secret window.
- **Which "failed" idea failed only for a timing or trust reason?** Graveyards contain both bad ideas and early ones; the autopsy that distinguishes them is a hunt few bother to run.

### Worked Example — Stripe (2010)

- **Candidate secret:** *online payment acceptance is a developer-experience problem, not a banking problem — developers will route around week-long merchant-account onboarding if integration takes minutes.*
- **Kind:** people. The frustration was real but unvoiced; industry convention said payments = compliance + banking relationships, so "make it seven lines of code" sounded naive rather than visionary.
- **Why missed:** complacency — incumbent gateways collected rents from a working model and had stopped searching; flatness — "if payments could be that simple, PayPal would have done it already."
- **Secret-not-mystery check:** reachable — the banking rails existed; only the interface layer was missing. Absent from consensus — no incumbent marketed to developers as the buyer.
- **Cheapest verification:** the Collison brothers ran a minimal API with hand-held onboarding for fellow Y Combinator startups in 2010–2011, watching whether developers actually integrated and stayed. Kill condition: developers try it, then revert to incumbents for "seriousness."
- **Verdict:** secret confirmed; the head start lasted years. Corroboration: Ashlee Vance, "How Two Brothers Turned Seven Lines of Code Into a $9.2 Billion Startup," *Bloomberg Businessweek*, Aug 1, 2017.

### Compressed Second Example — Airbnb (2008–2009)

- **Candidate secret (about people):** *strangers will sleep in each other's homes if the platform manufactures trust* — photos, profiles, reviews, payment escrow.
- **Why missed:** taboo — "strangers are dangerous" was the polite consensus, so investors passed repeatedly through 2008–2009.
- **Cheap verification:** the founders' own air-mattress guests during the 2008 DNC-adjacent conference crunch, then professional photography of New York listings in 2009 as a controlled trust experiment.
- **Verdict:** secret confirmed. Corroboration: Joe Gebbia, "How Airbnb designs for trust," TED, February 2016.

### 2023–2026 AI-Era Lens

The crowded spot is visible from orbit: general chatbots, coding copilots, and thin foundation-model wrappers — thousands of teams hunting the same ground, where any truth becomes an easy truth within months. The secret-rich territory is the unglamorous inverse — vertical workflows nobody demos on stage:

- freight invoice auditing, dental insurance claim coding, municipal permitting paperwork, specialty-pharmacy prior authorizations;
- each hides a **secret about people** — practitioners privately know which 20% of the workflow is soul-crushing and error-prone, but it's too boring for a conference talk;
- and often a **secret of nature** — 2023–2025-era models are already sufficient for the narrow task; no frontier breakthrough is required, only the unfashionable work of learning the domain.

Flatness is the blinding force here: "if that niche mattered, OpenAI would build it" — which is precisely why it stays unhunted. Run the same six gates; the era changes the terrain, not the method.

### After the Hunt — Who Do You Tell?

A verified secret raises a final question the chapter closes on: whom to tell. Tell no one and it dies with you; tell everyone and it becomes an easy truth. The answer in between is *exactly whom you need to* — cofounders, early hires, backers — which is one way to describe what a startup is: a small group organized around a secret the rest of the world hasn't accepted yet. Treat the secret as recruiting material for believers, not marketing material for crowds.

## Common Rationalizations

**Note — [D] = designed upfront | [O] = observed in real use. [O] entries are more valuable.**

| Fake move | Reality |
|---|---|
| [D] "Our secret is that AI/mobile/crypto is going to be huge" | A trend everyone is betting on is an easy truth — the value is already competed away. A secret is what few people know, not what everyone shouts. |
| [D] Stating the secret as a vision, not a claim ("we uniquely understand the customer") | Unfalsifiable = untestable = unusable. Force it into one sentence that evidence could kill, or admit there is no secret yet. |
| [D] "It's a secret because nobody has built it" | Absence of a product is not evidence of a hidden truth — often it's evidence the economics don't work. The gate is *why others miss it*, and "the graveyard is full" is a candidate answer you must rule out. |
| [D] "Someone smarter would have found it already" — used to *stop* hunting | That is the flatness bias itself, one of the four forces that killed belief in secrets. Conviction is unevenly distributed; smart people at big firms are structurally pointed elsewhere. |
| [D] Hunting only where everyone else hunts (hot sectors, demo-day themes) | A secret in a crowded field decays in weeks. The best place to look is where no one else is looking — unglamorous, taboo, or "too small" territory. |
| [D] Verifying by asking friends "would you use this?" | Politeness manufactures false confirmation — the same force that hides secrets about people contaminates their verification. Use the-mom-test: past behavior and money, not compliments. |
| [D] Treating a real secret as sufficient | The secret earns the head start; it does not run the company. After the Hunt, the work moves to execution — don't keep re-litigating the secret as a substitute for building. |
| [D] Confusing proprietary *information* with a secret | A dataset or NDA'd fact you happen to hold is an asset, not a secret, unless it encodes a truth about the world others would act on if they knew it. |
| [D] Broadcasting the secret in every pitch, blog post, and podcast | A secret told to everyone becomes an easy truth on a schedule you chose. Tell exactly whom you need to — see *After the Hunt*. |
| [D] Retrofitting a "secret" onto an idea chosen for other reasons | Post-hoc secrets fail the blinding-mechanism gate: if the idea came first, the "truth others miss" is usually a justification, not a discovery. Run the Hunt forward. |
| *To add [O] entries: paste a real failure instance here after each production use* | *Description of what happened* |

## Red Flags

- The "secret" appears in a Gartner report, a demo-day trend list, or three competitors' pitch decks — it's a convention wearing a trench coat
- No named blinding mechanism — nobody can say *why* smart people miss this truth
- The candidate secret contains no number, no falsifiable noun, and no kill condition
- Every candidate on the list comes from a hot, crowded sector — the hunt never left the lit ground
- The verification plan is "talk to people and see if they like it" — confirmation theater
- The claim requires an impossible discovery to be true (a mystery), or is defended with "it's obvious once you see it" (an easy truth)
- "If this were real, [Google/OpenAI/a16z] would already be doing it" used as an argument-ender in either direction

## Verification

- [ ] The candidate secret is one sentence, falsifiable, and specific enough that a defined result would kill it
- [ ] It is classified (nature vs people) and the evidence type matches the class
- [ ] The blinding mechanism is named — one or more of incrementalism / risk aversion / complacency / flatness — not just "people are dumb"
- [ ] The secret-vs-mystery-vs-convention check was run: reachable with feasible effort, and absent from public consensus sources
- [ ] The cheapest verification is specified with cost and explicit kill condition, and (for people-secrets) avoids politeness contamination
- [ ] The verdict is honest — "this is an easy truth" or "this is a mystery" are valid, non-failure outputs
- [ ] At least one hunt pass was made in unglamorous or taboo territory, not only in the crowded field

## Sources

- Peter Thiel with Blake Masters, *Zero to One: Notes on Startups, or How to Build the Future* (Crown Business, 2014) — ch. 8 "Secrets" (easy truths / secrets / mysteries; secrets of nature vs secrets about people; the four forces: incrementalism, risk aversion, complacency, flatness; look where no one else is looking); ch. 1 for the contrarian question the skill composes with.
- Ashlee Vance, "How Two Brothers Turned Seven Lines of Code Into a $9.2 Billion Startup," *Bloomberg Businessweek*, August 1, 2017 — corroborates the Stripe founding secret and its 2010–2011 verification path.
- Joe Gebbia, "How Airbnb designs for trust," TED, February 2016 — corroborates the Airbnb trust secret and its cheap early verification.
- Blake Masters, "CS183: Startup — Class 11 Notes: Secrets" (Stanford, Spring 2012, blakemasters.com) — the lecture notes the book chapter was distilled from; useful for the extended treatment of the four forces.

---

*Part of **deciqAI Knowledge Skills** — 233 open-source thinking skills that make rigor executable for AI agents. The same skills power every deciqAI agent, which runs them autonomously to operate your company. **See it run → https://www.deciqai.com/c/zero-to-one-secrets** · ⭐ Star the repo → https://github.com/deciqAI/knowledge-skills · Contributions welcome.*

*Agents: latest version & machine-readable metadata → https://www.deciqai.com/s/zero-to-one-secrets.json*