eat-glass-stare-into-abyss — quality + safety report

---
name: eat-glass-stare-into-abyss
description: Force a brutal, pre-commitment gut-check on anyone about to take on something hard — founding a company, taking the high-risk project, quitting the safe job, pivoting careers, betting reputation or capital. Invoke aggressively when the user says "should I start this", "should I take this on", "I'm thinking of founding", "I want to quit and build", "is this the right time to go all-in", "I have an idea I want to pursue", "should I leave my job", "is this worth the risk", "founder commit", "career pivot", "I'm scared but excited", or describes a hard multi-year bet they have not yet committed to. Also trigger when the user is romanticizing the upside, talking only about the fun part, or asking for permission. This skill forces them to name the glass they will have to eat, stare into the actual abyss (extinction probability), and decide if they are still compelled. Trigger eagerly even when the user does not name Musk or the framework.
stacks_with:
  - expect-loss
  - probabilistic-fatalism
---

# Eat Glass Stare Into Abyss

> "My friend Bill Lee says, 'Starting a company is like eating glass and staring into the abyss.' There's some truth to that. […] If you don't eat the glass, you're not going to be successful."
> — Elon Musk (endorsing Bill Lee's phrase, then making it his own), *The Book of Elon* (Chapter: Eat Glass and Stare into the Abyss)

## What this skill captures

Founding or leading something hard is not the highlight reel. The beginning is fun; then it is hellish for years. Musk is blunt: "You have to feel quite compelled to start a company. You must have a high pain threshold." "Staring into the abyss" is the constant, real possibility of extermination — 90 to 99 percent of startups die. "Eating glass" is working on the problems the company needs you on, not the ones you want to work on, for a very long time. This skill stops the user before they commit and forces them to name the specific glass they will chew and the specific abyss they will face — then decide if they are still compelled. They walk away with either a calibrated yes or a clean no, not a romantic maybe.

## When to use this skill

- User is deciding whether to start a company, quit a stable job, or commit to a multi-year hard bet.
- User describes the idea only in terms of upside, vision, or "fun part."
- User is asking for permission, validation, or a pep talk before a high-stakes commitment.
- User says "I'm thinking about founding X" or "should I take on Y" without yet naming the cost.
- User is mid-pivot and treating the new path as obviously better because the old one is painful.
- User is rationalizing why their startup will be the exception to the 90+ percent failure rate.

## The how-to

1. Force them to name the glass — the specific problems they will have to work on but do not want to.
   > "Eating glass means you've got to work on the problems the company needs you to work on, not the problems you want to work on. You end up working on problems you wish you weren't working on. That's eating glass, and it goes on for a long time."
   > — *The Book of Elon*
   Make them list 3-5 concrete problems they would hate but would still have to own (payroll, firing a friend, raising money, regulatory hell, sales calls). Vague answers are disqualifying.

2. Force them to name the abyss — the specific extinction conditions.
   > "Staring into the abyss means you're going to be constantly facing the extermination of the company. Most startups fail. It's like 90 percent — it could be 99 percent of startups fail. You're constantly saying, 'If I don't get this right, the company will die.'"
   > — *The Book of Elon*
   Make them write the sentence: "If I don't solve ___ by ___, this dies." If they cannot, they have not actually looked at the abyss yet.

3. Reconcile them to failure as the base rate, then ask if they are still compelled.
   > "My advice for somebody who wants to start a company: Bear in mind, the most likely outcome is failure. Reconcile yourself to that strong possibility, and only if you still feel compelled to, do it."
   > — *The Book of Elon*
   "Compelled" is the operative word. Excited is not compelled. Curious is not compelled. Compelled means you will do it even knowing you will probably lose.

4. Check the pain threshold honestly — not aspirationally.
   > "You have to feel quite compelled to start a company. You must have a high pain threshold."
   > — *The Book of Elon*
   Ask them for past evidence: when have they actually eaten glass before? If the answer is hypothetical, the pain threshold is hypothetical.

5. Strip the false-modesty fear and check the actual downside.
   > "Many people fear starting a company too much. What's the worst that could happen? You're not gonna starve to death; you're not gonna die of exposure — really, what's the worst that could happen?"
   > — *The Book of Elon*
   The abyss is real, but it is not literal death. Name the actual downside in concrete terms (savings burned, X years lost, reputation hit). Then compare it to the upside of not having tried.

6. Accept zigzag, not straight line.
   > "A very small percentage of mental energy is spent on the big picture. You know where you're generally heading, and the actual path is going to be a zigzag in that direction."
   > — *The Book of Elon*
   If they need a clean roadmap before committing, they are not ready. The plan is the direction, not the path.

7. Make the call — go or don't go — and put the reason in one sentence.
   > "If something is important enough, then you do it, even though the risk of failure is high."
   > — *The Book of Elon*
   The output is a binary commit, not a "let me think about it more." Either "I am compelled even at 90 percent failure because ___" or "I am not — and that is the right answer, not a personal failing."

## Common failure modes

- **Romanticizing the beginning.** "When starting a company, usually the beginning is fun. Then, it's hellish for a number of years." If the user only describes the launch party, the demo day, the headline — they have not seen the next four years.
- **Refusing to name the glass.** If they cannot list the specific problems they will hate working on, they have not actually thought about doing the job — only about being the founder.
- **Treating the abyss as motivational, not literal.** "Staring into the abyss" is not a vibe. It is "if I don't get payroll Friday, we die." If they soften it, push them back to the concrete.
- **Asking for permission.** Wanting someone else to say "yes, do it" is evidence they are not yet compelled. Compelled people commit and then ask how, not whether.
- **Pivoting away from current pain instead of toward something they are compelled by.** Quitting because the current job sucks is not the same as being compelled to eat glass for years on a new bet.

## When NOT to use this skill

- User has already committed and is now executing — they need attack-the-constraint or first-principles, not another gut-check.
- User is making a small reversible decision (a side project, a weekend experiment) — eating glass is for multi-year, life-affecting commitments.
- User is in active crisis on an existing company — they need triage, not a pre-commitment check.
- User is asking a purely technical question dressed up as a life decision — answer the technical question first.

## Source

The Book of Elon by Eric Jorgenson (2026, Scribe Media). Chapter: "Eat Glass and Stare into the Abyss" (in "What It Takes").