## Progressive Summary **Executive Summary (Layer 3)**: **A surface analogy ("X acts like Y") does not predict mechanism-level effects: caffeine and cortisol are both stimulants, yet caffeine is *protective* against blepharospasm while cortisol worsens it — they act on different receptors.** **Key Insight (Layer 2)**: "Caffeine antagonizes adenosine receptors in the striatum and has a protective association against blepharospasm — the opposite of what 'caffeine is a stimulant like cortisol' predicts." **Context (Layer 1)**: From a mechanistic review of blepharospasm and cortisol. Cortisol is said to act "similar to caffeine," naively implying caffeine should aggravate spasms. Population data shows the reverse. **Cross-Domain Connections**: [[Mechanism-Performance Gap in Complex Systems]] **Discoverability Score**: 7/10 --- ## Atomic Insight Analogy compresses two things into one similarity ("both stimulants") and transfers a predicted effect across that bridge. The bridge fails whenever *mechanism*, not surface category, sets the outcome. Cortisol dysregulates striatal dopamine, lowering the basal ganglia's inhibitory control over the blink reflex. Caffeine blocks striatal adenosine receptors, which population data links to *reduced* blepharospasm risk. Same surface label, opposite clinical sign. The error is treating the analogy as a causal model. "Like" describes appearance; mechanism describes how the system transforms inputs. Only mechanism predicts direction and magnitude. Applies to drug-interaction reasoning (two "sedatives" can net opposite effects), system design (two services that "do caching" fail under load for opposite reasons), and any forecast resting on "this is basically the same as that."