# Go Deeper

> Push past the obvious first answer on any topic, take, or draft by drilling three "why" layers down until a non-obvious insight appears. Use whenever the user types /go-deeper, says a take feels shallow, generic, or "something everyone already says", asks for a more original angle or a contrarian view, or shares an idea/draft and asks "what am I missing" or "make this more interesting". Also use when the user's own analysis stops at the first explanation.

- Skill: `rohankhullar24-oss/go-deeper` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add rohankhullar24-oss/go-deeper`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rohankhullar24-oss/go-deeper/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: rohankhullar24-oss (https://skillmd.com/u/rohankhullar24-oss)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/rohankhullar24-oss/go-deeper

---


# Go Deeper

The first answer to any question is the one everyone gives. Depth is not longer writing — it is asking "why is that true?" enough times that you arrive somewhere the reader hasn't been. This skill runs that drill.

## The drill

### Layer 0 — Name the surface take
State the obvious version of the idea in one sentence — the thing a competent generalist would say. Label it plainly as the starting point. (If the user gave a draft, extract its central claim as Layer 0.)

### Layer 1 — The mechanism (why is it true?)
Explain *what actually causes* the surface take to be true. Not evidence that it's true — the machinery behind it. Incentives, constraints, human behavior, economics, physics. If Layer 1 could be written by someone who has only read headlines about the topic, dig again.

### Layer 2 — The mechanism behind the mechanism (why does THAT exist?)
Ask why of Layer 1. This is where takes usually stop being interchangeable: the second why forces you into history, second-order effects, or the question of who benefits from the current arrangement. Useful prompts at this layer:
- Who profits from this staying true?
- What had to happen for this to become normal?
- What does everyone in this system know that outsiders don't?

### Layer 3 — The break point (where does it stop being true?)
Every insight has edges. Find them:
- Under what conditions does the surface take invert?
- Who is the exception, and what makes them exempt?
- What is quietly changing that will make this take wrong in 3 years?

The break point is usually the most original content in the whole drill — it's the part that can't be pattern-matched from other writing on the topic.

### The payoff — So what?
Compress the drill into one **earned insight**: a single sentence that would sound like a hot take on its own, but which the three layers now justify. Then state what the reader should *do or see differently* because of it. An insight without a consequence is trivia.

## Output format

ALWAYS use this structure:

```
**Surface take:** [the obvious version — one sentence]

**Layer 1 — why it's true:** [mechanism]

**Layer 2 — why that exists:** [incentives / history / second-order]

**Layer 3 — where it breaks:** [inversion, exception, or expiry date]

**The earned insight:** [one sharp sentence the layers justify]

**Rewritten take:** [2-4 sentences: the user's original idea, restated
from the deepest layer that survived scrutiny — ready to use]
```

## Quality bar

- **Each layer must add information, not adjectives.** If a layer just restates the previous one with more conviction, it doesn't count — dig again or admit the well is dry.
- **No fake depth.** If the honest answer is that the surface take is simply correct and complete (some things are!), say so at whatever layer the digging stops producing new information, and make the payoff about *why the topic resists depth* — that itself is often interesting.
- **Stay falsifiable.** Prefer claims someone could disagree with over mystical profundity ("it's really about trust"). If the earned insight can't be wrong, it isn't an insight.
- **Flag speculation.** Layers 2 and 3 often involve inference. Mark the difference between "this is documented" and "this is my read" so the user doesn't publish a guess as a fact — or offer to fact-check the load-bearing claims if accuracy matters for publication.

## Example (compressed)

**Topic:** "Consistency is the key to growing on LinkedIn."
- **Layer 1:** The feed rewards accounts that generate reliable engagement data — consistency gives the algorithm more samples to learn who your content serves.
- **Layer 2:** Platforms optimize for *predictable* inventory because advertisers buy predictability, so the system structurally favors habitual creators over occasional brilliance.
- **Layer 3:** It breaks at the extremes — past a quality floor, one exceptional post outperforms 90 mediocre consistent ones, because outlier content is what the algorithm tests on new audiences.
- **Earned insight:** Consistency isn't rewarded for discipline; it's rewarded because you're supplying the platform with predictable ad inventory — and the reward stops the moment consistency becomes an excuse for average work.