Pattern Matching

Humans are pattern-seeking machines. We see faces in clouds, conspiracies in randomness. The brain compresses overwhelming input into actionable understanding by matching new situations against stored templates. Every perception, every judgment, every decision begins with the same operation: does this match something I've seen before? If yes, apply the stored response. If no, flag it as novel and allocate scarce attention. The mechanism trades accuracy for speed — and across most of evolutionary history, the trade has been successful.

In valid domains — chess, radiology, firefighting — pattern matching is expertise. The grandmaster recognises board positions in a glance. The radiologist spots the anomaly before articulating why. The firefighter senses structural danger before the conscious mind processes the evidence. These domains have stable, repeating structures and clear feedback. The pattern library captures genuine causal structure. Pattern matching produces expert-level accuracy at speed.

In invalid domains — stock picking, hiring from resumes, venture capital — pattern matching is superstition. The patterns feel like insight. They function as bias. VCs pattern-match on "founder-market fit": the right background, the right pedigree, the right narrative. Sometimes it works. Often it's survivorship bias — the pattern library was built from the founders who succeeded, not from the full distribution of attempts. The companies that produced the largest returns — Google, Airbnb, Tesla — did not match existing patterns. They broke them. The VC who relies on pattern matching will consistently fund companies that look like past successes. They will systematically miss the companies that look like nothing that has succeeded before — which is exactly the category that produces the largest returns.

Amazon's "working backwards" forces evidence over pattern. Start with the customer press release. What would success look like? Then work backward to the requirements. The method prevents the team from pattern-matching against "what we've built before" or "what the industry does." It forces engagement with the specific problem rather than defaulting to the general template. The key: know when patterns are real — when the domain has structure — and when they're noise — when the domain is random or the structure has shifted.

The meta-skill is not pattern matching. It is knowing when to trust pattern matching and when to override it. In stable environments with repeating structures — chess, medicine, military tactics — pattern matching is the expert's superpower, compressing thousands of prior encounters into split-second recognition. In novel environments with structural breaks — technological paradigm shifts, new market categories — pattern matching is the expert's blindspot, confidently applying yesterday's templates to tomorrow's problems. The master is not the person with the largest pattern library. It is the person who knows which patterns in their library are still valid and which have expired.