| name | abstract-vs-concrete |
| description | Ladder-climbing primitive — given a claim at one rung of the abstraction ladder, propose the paired move at the other rung. When V's high (frameworks, positioning, theoretical claims), abstract-vs-concrete demands one paragraph of concrete instantiation. When V's low (specific tasks, examples, single cases), it asks for the framework-level statement. Triggers — "is this too abstract?", "ground this for me", "what's the example?", "what's the principle?", "lift this", "drop this". Tightly paired with distill but operates on the vertical axis (rung-climbing) where distill is horizontal (convergence). NOT a generic "give me an example" tool — the discipline is *paired moves* between rungs, not infinite expansion. Callable by Fry / thinking-partner / V directly. |
| tags | ["ai-os","skills","conceptual-analysis","fry","thinking","abstraction-ladder"] |
| last-verified | 2026-07-20T00:00:00.000Z |
Abstract vs Concrete
A transform, not a conversation. Input: a claim at one rung of the abstraction ladder. Output: the paired move at the other rung — a concrete instantiation if V's high; a framework-level statement if V's low.
The abstraction ladder is from S.I. Hayakawa via the General Semantics tradition (also used by Bret Victor, Tim Urban, and clear-thinking writers generally). The discipline: high-rung claims need one concrete case to ground them; low-rung cases need one framework-level statement to locate them. Rung-climbing is the test that V's claim works at multiple rungs — if it doesn't, that's diagnostic.
When this fires
Explicit triggers:
- "Is this too abstract?"
- "Ground this for me"
- "What's the example?"
- "What's the principle here?"
- "Lift this" (low → high)
- "Drop this" (high → low)
- "Climb the ladder"
Contextual (Fry detects):
- V's claim is high-rung framework with no concrete instantiation in 5+ minutes of conversation. (Frameworks without instances are floating.)
- V's claim is low-rung specific case but the implicit pattern hasn't been named. (Specifics without principles are anecdotes.)
- V's moving between rungs without realizing it — switching mid-argument from high to low or vice versa, hiding a category error.
- V's stuck on the abstract claim and a concrete case would unstick — or stuck on the concrete and the abstract would.
Offer once: "Want to drop this to a concrete case?" or "Want to lift this to the framework-level statement?"
When this stays dormant
- V is intentionally working at one rung — the ladder-climb would derail.
- The claim is identity-claim and neither rung is testable (per
identity-vs-empirical.md).
- V already has both rungs and is just naming both for clarity.
- The "concrete" version would be a personal anecdote that distracts from the analytical work.
Inputs
Required:
- A claim at one rung — high (framework, principle, theoretical) or low (specific case, example, single instance)
Optional:
- Direction (V can specify "lift" for low→high or "drop" for high→low; skill detects if not specified)
- Domain context (some ladders are domain-specific — what's high in cognitive science is different from what's high in branding)
The core move
- Detect the rung. Read the input claim. Is it high (framework, principle, abstraction) or low (instance, case, example)? If unclear, surface the ambiguity — the claim may be middle-rung and need ladder-clarification first.
- Pick the direction. If high, drop. If low, lift. (V can override.)
- Make the paired move.
- HIGH → LOW (drop): Produce ONE concrete case where the high-rung claim shows up. Specific. With detail.
- LOW → HIGH (lift): Produce ONE framework-level statement that this case is an instance of. The principle, named.
- Test the pairing. Does the high claim survive the concrete case? Does the concrete case fit the framework cleanly? If either answer is no, that's diagnostic — the claim doesn't work at multiple rungs, and the diagnosis is the value.
Output format (canonical contract)
## Abstract-vs-concrete on [V's input claim, one line]
**Detected rung:** HIGH | LOW | MIDDLE-AMBIGUOUS
**Direction:** DROP (high → low) | LIFT (low → high)
**Paired move:**
[For DROP: one concrete case. Specific, with detail. ~50-150 words.]
[For LIFT: one framework-level statement. The principle, named, located in tradition where possible. ~30-80 words.]
**Pairing test:**
- Does the high claim survive the concrete case? YES | NO | PARTIAL
- Does the concrete case fit the framework cleanly? YES | NO | PARTIAL
**Verdict:**
[ONE of:]
- LADDER-INTACT — the claim works at both rungs; V can move between them freely
- HIGH-FLOATS — the framework doesn't survive concrete instantiation; V needs to refine the framework or honestly call it gestural
- LOW-DOESN'T-FIT — the case doesn't fit the framework V tried to lift it to; V needs a different framework or to recognize it's sui generis
- MIDDLE-AMBIGUOUS — V's claim is at an indeterminate rung; clarify rung before continuing
**Recommended next move:**
[ONE sentence — what V does given the verdict.]
Strict rules:
- Detected rung: one of three options. If MIDDLE-AMBIGUOUS, the skill ends there with the diagnostic — clarify rung, then re-run.
- Paired move: ONE concrete case OR ONE framework statement. Not 3 examples. Not a list. The discipline is one paired move, not exhaustive expansion.
- Pairing test: explicit YES/NO/PARTIAL on both questions.
- Verdict: one of four options.
- Cap output at ~250 words. Brevity is the point — the ladder-climb is a quick diagnostic, not an essay.
Agent-callable mode
When called by Fry or another agent, skip scaffolding. Begin with ## Abstract-vs-concrete on [claim]. End after **Recommended next move:**. The bundle is the entire output.
Anti-patterns
- ❌ Multi-example expansion. "Here are three concrete cases..." — cut to one. The discipline is paired-move, not list-of-examples.
- ❌ Anecdote drift. When dropping high-to-low, the concrete case must be analytically clean — not "here's a story from my life." Personal cases are fine if they're load-bearing for the analysis; not fine if they're decorative.
- ❌ Ladder-skipping. If V's claim is at rung 3 and you drop to rung 5 (skipping rung 4), the move doesn't test the framework — it bypasses the rungs that matter. Drop one rung at a time.
- ❌ Framework inflation. When lifting low-to-high, don't reach for grand-theory if a mid-rung framework fits. "This is an instance of [niche pattern]" beats "this is an instance of how reality works."
- ❌ Pairing-test hedging. "Sort of works..." — pick YES, NO, or PARTIAL. Hedging defeats the diagnostic.
- ❌ Pairing-test optimism. Don't manufacture YES verdicts. If the framework doesn't survive the concrete case, the verdict is HIGH-FLOATS — that's the useful finding.
Relationship to other skills
| Skill | Shape | Relationship |
|---|
distill | Horizontal convergence (many threads → one claim) | abstract-vs-concrete is the vertical counterpart — operates on rungs, not threads. Often paired: distill produces the claim; abstract-vs-concrete tests it works at multiple rungs |
distinguish | Disambiguate concepts (horizontal at same rung) | distinguish operates within a rung; abstract-vs-concrete moves between rungs — orthogonal |
silent-assumption | Surface unstated dependencies | abstract-vs-concrete may reveal a silent assumption (the framework assumed an instance type that doesn't exist) — hand off |
negative-claim | Commitment trace | negative-claims at one rung may not hold at another — useful pairing |
falsify (via critical-thinking Mode 4) | What would prove this wrong? | abstract-vs-concrete is upstream: ground the claim at both rungs, then falsify the appropriate-rung version |
Common pipeline: distill → abstract-vs-concrete → falsify or silent-assumption. distill produces the claim; abstract-vs-concrete grounds it at multiple rungs; the appropriate-rung version is then testable.
What this does NOT do
- Generate multiple examples or instances (one paired move only)
- Settle "is this true at the abstract level" (use falsify)
- Run V's framework through every possible concrete case (one drop, then stop)
- Climb arbitrary rungs (one rung at a time; skipping defeats the diagnostic)
- Manufacture LADDER-INTACT verdicts when the claim genuinely doesn't survive