diff --git a/src/puzzles/cognitive-transfer-overview.md b/src/puzzles/cognitive-transfer-overview.md
new file mode 100644
index 0000000..c9735e5
--- /dev/null
+++ b/src/puzzles/cognitive-transfer-overview.md
@@ -0,0 +1,51 @@
+# Cognitive Transfer Puzzles
+
+Cognitive transfer puzzles operate on a **Learn → Apply** structure rather than Gather → Synthesize. Players encounter a system, rule set, or behavioral pattern in one context, then transfer that knowledge to solve problems elsewhere. The core challenge isn't collecting pieces—it's recognizing that information learned in situation A applies mechanistically to situation B.
+
+These puzzles test whether players observe systems deeply enough to extract reusable rules. The "aha" moment comes not from finding new information, but from realizing **old information has new applications**.
+
+## Core Characteristics
+
+| Trait | Description |
+|-------|-------------|
+| Learning Phase | Player observes or experiments to discover rules/patterns |
+| Transfer Distance | How different the application context appears from the learning context |
+| Abstraction Level | Whether rules are concrete (this button = that light) or abstract (patterns mirror relationships) |
+
+## Cognitive Transfer Taxonomy
+
+Cognitive transfer divides into five distinct mechanisms based on what transfers and how:
+
+### Direct Pattern Types
+
+- **[Pattern Learning / Knowledge Transfer](puzzles/pattern-learning.md)** - Learn mechanical rules in one domain, apply identical logic to different target
+
+- **[Symbol Code Translation](puzzles/symbol-code-translation.md)** - Visual symbol recognition and mapping as extended pattern learning
+
+### Abstract Reasoning Types
+
+- **[Metaphor-to-Literal Bridges](puzzles/metaphor-literal.md)** - Abstract concept reasoning made concrete through game mechanic
+
+### Observation-Based Types
+
+- **[Sensory Exploitation](puzzles/sensory-exploitation.md)** - Exploit NPC perceptual limitations after observing thresholds
+
+- **[Observation Replay](puzzles/observation-replay.md)** - Watch sequence once, reproduce exactly in player context
+
+## Design Distinctions
+
+Cognitive transfer puzzles differ from other categories by their emphasis:
+
+| Vs. Other Types | Key Difference |
+|-----------------|----------------|
+| **Multi-Faceted Plan** | MFP synthesizes disparate requirements; cognitive transfer applies unified rules across contexts |
+| **Meta-Construction** | Meta-construction chains outputs sequentially; cognitive transfer uses parallel application of learned system |
+| **Brokerage** | Brokerage trades items along networks; cognitive transfer trades knowledge across domains |
+
+## Common Design Failures
+
+**Observation vs. Replay**: Teaching players to memorize a specific sequence rather than understand underlying rules creates observation replay, which feels like rote memorization instead of genuine learning.
+
+**Insufficient Transfer Distance**: If the application context looks identical to the learning context, players don't experience cognitive transfer—they recognize surface similarity rather than rule abstraction.
+
+**Hidden Learning Opportunities**: Players must have clear opportunities to learn the system before being asked to apply it. No tutorial means no fair transfer.